Energy Conversion and Management 298 (2023) 117768

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Renewable energy design and optimization for a net-zero energy building

integrating electric vehicles and battery storage considering grid flexibility
Jia Liu a, b, Huijun Wu a, *, Huakun Huang c, Hongxing Yang d
a
School of Civil Engineering, Guangzhou University, Guangzhou, China
b
Key Laboratory of Building Safety and Energy Efficiency of Ministry of Education, Hunan University, China
c
School of Computer Science and Cyber Engineering, Guangzhou University, Guangzhou, China
d
Department of Building Environment and Energy Engineering, The Hong Kong Polytechnic University, Kowloon, Hong Kong, China

A R T I C L E I N F O A B S T R A C T

Keywords: This study proposes a design management and optimization framework of renewable energy systems for
Net-zero energy buildings advancing net-zero energy buildings integrated with electric vehicles and battery storage. A building load data
Electric vehicle integration augmentation model is developed to obtain the annual hourly load profile of a campus building based on the on-
Grid flexibility management
site collected data adopting the Gate Recurrent Unit neural network. A grid-protective energy management
Design optimization
strategy of the solar photovoltaic (PV) power and storage system is proposed with novel assessment indicators
including PV utilization ratio, load match ratio and grid flexibility factor. Multi-objective optimizations are
conducted to identify optimal sizing of the renewable energy system and its interactive impact on balanced
techno-economic performance. The research results indicate that the net-zero energy building achieves optimum
performance with the sizing configuration of 1050 kW rooftop PV power, 300 electric vehicles and 450 kWh
batteries. The vehicle-to-building interaction introducing vehicle discharge improves the load coverage (+12.08
%), grid flexibility (− 29.63 %), annual electricity bill (− 18.70 %) and levelized cost of energy (− 6.24 %). The
optimum net-zero energy building achieves good techno-economic-environmental feasibility regarding the load
coverage (+16.22 %), grid flexibility performance (− 58.48 %), annual electricity bill (− 27.86 %), decarbon­
isation benefits (− 34 times) and vehicle degradation. The developed design management and optimization
framework of the renewable energy system provides a possible pathway for the typical campus building towards
zero-carbon operations, and it also offers guidance and reference for stakeholders to develop similar carbon­
–neutral buildings.

in 2022 including tax credits for rooftop solar PV, electrical vehicles and
home batteries [4]. China plans to install nearly half of the global new
1. Introduction renewables capacity addition from 2022 to 2027 [5]. China also stipu­
lates to promote on-site renewable energy application in urban build­
1.1. Background –Necessity to develop renewable energy supply for ings, where an installation capacity of 50 GW solar PV power is targeted
buildings for all new public buildings and factories [6]. Renewable energy is
enjoying an increasing global market driven by policy incentives and
The world experienced the impact of a severe global energy crisis technology development [7] with an annual average rising demand ratio
caused by the COVID-19 pandemic and international conflict wars, of 4.7 % between 2010 and 2020 [8].
resulting in soaring energy prices affecting all energy-consuming sectors The building sector contributes to around 33 % of global final energy
[1]. Renewable energy is the forefront of policy in response to the twin consumption in 2020, where about 15.5 % of the building energy use is
crises of rising energy costs and inflation attributed to its reliable, stable supplied by renewables [9]. The energy consumption in buildings of top
and affordable characteristics [2]. The EU plans to phase out fossil fuel ten regions in 2020 is shown in Fig. 1 contributing to a global proportion
imports with an investment of US$ 222 billion during 2023 to 2027, and of about 67 % [9]. It can be found that the building energy consumption
aims to install rooftop solar photovoltaic (PV) on all public buildings by varies greatly around the world with about 40 % in Canada and only 5 %
2025 [3]. The US allocates US$ 370 billion for all energy demand sectors

* Corresponding author.
E-mail address: [email protected] (H. Wu).

https://doi.org/10.1016/j.enconman.2023.117768
Received 2 August 2023; Received in revised form 23 September 2023; Accepted 12 October 2023
Available online 21 October 2023
0196-8904/© 2023 Elsevier Ltd. All rights reserved.
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Nomenclature Egrid to load_valley energy from grid to cover building load in valley
hours, kWh
Acronyms Egrid out_flat grid imported energy during flat time, kWh
CE carbon emissions Egrid out_peak grid imported energy during peak time, kWh
DOD depth of discharge Egrid out_valley grid imported energy during valley time, kWh
EB electricity bill Eload annual electrical load of the building, kWh
EV electric vehicle Enet grid import_flat net grid imported energy during flat hours, kWh
FiT feed-in-Tariff Enet grid import_peak net grid imported energy during peak hours, kWh
GFF grid flexibility factor Enet grid import_valley net grid imported energy during valley hours, kWh
GRU Gate Recurrent Unit EPV annual generation from solar photovoltaic system, kWh
LCOE levelized cost of energy EPV to battery energy from solar photovoltaic supply to charge battery
LMR load match ratio storage, kWh
NPV net present value EPV to EVs energy from solar photovoltaic supply to charge electric
O&M operation and maintenance vehicles, kWh
PRV present value EPV to grid_flat energy from solar photovoltaic supply to utility grid in
PV solar photovoltaic flat hours, kWh
PVUR PV utilization ratio EPV to grid_peak energy from solar photovoltaic supply to utility grid in
RE renewable energy peak hours, kWh
EPV to grid_valley energy from solar photovoltaic supply to utility grid in
List of symbols valley hours, kWh
c cycle number of electric vehicles EPV to load energy from solar photovoltaic supply to cover building
CEFflat carbon emission factor of grid electricity in flat hours, kg/ load, kWh
kWh GFFpeak grid flexibility factor of peak time
CEFpeak carbon emission factor of grid electricity in peak hours, kg/ GFFflat grid flexibility factor of flat time
kWh GFFvalley grid flexibility factor of valley time
CEFvalley carbon emission factor of grid electricity in valley hours, i real discount rate
kg/kWh k one replacement time of a specific component
DegEVs annual degradation rate of electric vehicles K total replacement times of a specific component
Ebattery to load energy from battery storage to cover building load, kWh lres residual lifetime of a specific component
EEVs to load energy from electric vehicles to cover building load, kWh n a specific year
Egrid to EVs_peak energy from grid to charge electric vehicles in peak Y service lifetime of the renewable energy system, years
hours, kWh Rz range of z-th cycle in the depth of discharge profile of
Egrid to EVs_valley energy from grid to charge electric vehicles in valley electric vehicles
hours, kWh tarr,gi arriving time of gi-th electric vehicles
Egrid to load_flat energy from grid to cover building load in flat hours, υ degree of freedom
kWh γ annual rising rate of local grid electricity
Egrid to load_peak energy from grid to cover building load in peak hours, δPV annual degradation rate of solar photovoltaic system
kWh

Fig. 1. Energy consumption in buildings of top ten regions around the world.

2
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

in India, Russian and Korea. The building energy consumption in the US Germany, France and the UK. It is necessary to study the techno-
is the maximum, where majority is provided from fossil fuels and about economic-environmental feasibility of integrating EVs with renewable
14 % from renewables. Renewables are responsible for about 20 % of the energy systems for application in buildings, given its low emission, high
building energy use in China. International Energy Agency expects that efficiency, and particularly increasing role in the power market.
all buildings must possess renewable energy strategies by 2050 towards
zero-energy and zero-carbon standards [10].
The current constitution of power unit installation in Guangdong 1.3. Literature review
Province of China in 2020 is shown in Fig. 2 with a total installation
capacity of about 140,755 MW [11]. It can be found that majority of the The design optimization and feasibility analysis of renewable energy
local power is supplied by coal currently accounting for about 47 %, and and storage systems for net-zero energy buildings has attracted much
followed by gas turbine at about 19 %. The clean power production attention in the academia motivated by the sustainable, affordable and
contributes to about 33 % of the total installation with 11 % from nu­ low-carbon characteristics of renewable energy [21]. The renewable
clear, 7 % from hydropower, and 4 % for both on-grid solar energy and energy design for power supply to buildings has been studied regarding
on-grid wind power. The local government commits to accelerating the the energy management and system evaluation. The feasibility of inte­
high-quality development of renewable energy to support the clean grating EVs with renewable energy systems for building applications has
power network with a targeted provincial installation capacity of 79 been evaluated considering the EV battery degradation, EV charging loss
million kW by 2025 [12]. It highlights the necessity to proceed net-zero and stochastic driving schedules. Design optimizations of renewable
energy buildings including utilizing solar PV in public buildings, fac­ energy systems have been conducted for optimum solutions to achieve
tories and infrastructures. It is therefore important to develop renewable balanced performance of the technical, economic and environmental
energy applications in buildings in modern cities in response to the aspects.
global and local momentum towards net-zero energy pathways [13]. Renewable energy design for power supply to buildings has been
investigated by researchers regarding the energy management [22,23]
and techno-economic-environmental feasibility evaluation [24,25].
1.2. Global application status of electric vehicles integrating renewable Specifically, the net-zero energy management of commercial buildings
energy in Hong Kong is investigated considering the future-oriented application
scenario with high renewable energy penetration. It is found that the
Electric vehicles (EVs) are enjoying an unprecedented market in proposed future-oriented strategy improves the renewables utilization
recent years driven by the urgent demand to decarbonize the transport and building load coverage evaluated by the self-consumption ratio and
sector representing about 33 % of global energy use and 20 % of global load cover ratio, respectively [22]. A two-stage real-time demand
carbon emissions in 2021 [14]. It is reported that direct electrification response framework is presented for a building multi-energy microgrid
via batteries is the most efficient, cost-effective and commercially system incorporating solar energy, wind power, geothermal and energy
available route to fully decarbonize the road transport [15]. Numbers of storage units. The case study on a building community shows that the
countries have issued plans on developing EVs regarding tax incentives system operational cost can be reduced by about 36.9 % ensuring a high
and stimulus packages. For example, Thailand plans to produce 0.25 flexibility [23]. A renewable energy system is developed including solar
million EVs by 2025 and 1.2 million EVs by 2036 [16]. The Philippines energy, biomass and geothermal for power supply to an Olympic
and the US governments provide fiscal incentives for EV manufactures Training Center with heterogeneous buildings in Finland. The technical
[17,18]. The environmental bonus is offered in France and Germany for and economic feasibility of the system is identified via developing
buying EVs [9]. It is projected that global EV sales would rise from 10.5 innovative scenarios [24]. The renewable energy design for zero-energy
million in 2022 to almost 27 million in 2026 [19] sharing a crucial buildings and communities is studied with battery storage and hydrogen
market. It can be found that enhancing EV technologies and increasing vehicle storage, where the grid flexibility is evaluated by a novel grid
renewables share is critical to decarbonize the transport sector [20]. penalty cost model. The indicators of self-consumption ratio and load
The global investment in EVs and relevant charging infrastructure cover ratio are adopted to evaluate the utilization efficiency of renew­
surged in recent years as shown in Fig. 3 [8]. China kept at the leading able power supply and building load coverage, respectively [25].
role in investing EVs during the last five years accounting for about 59 % Increasing number of literatures have evaluated the feasibility of
in 2022 totaling US$ 234 billion. The US invested about US$ 57 billion integrating EVs with renewable energy systems for building applications
in 2022 for EVs as the second largest investor with a global proportion of considering the EV battery degradation [26,27], EV charging loss [28]
about 14 %. The rapid growth of EV investment is also observed in and stochastic driving schedules [29]. Specifically, a rule-based energy

Fig. 2. Constitution of power unit installation in Guangdong Province.

3
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 3. Investment of electric vehicles of major countries.

management strategy is proposed to study the performance of a micro- case studies in three Saudi cities to optimize the lifecycle cost and carbon
grid renewable energy system with battery EVs and fuel cell EVs. The emissions. It is reported that higher investment on battery storage is
battery degradation is considered from the perspective of energy needed to combat extreme weather events for the hybrid PV and wind
throughput adopting the equivalent aging cycles, finding that the power system [33].
degradation of battery EVs can be reduced by 13 % via introducing fuel Based on the above literature review, research gaps can be found that
cell EVs [26]. The impact of charging scenarios on cycling degradation few studies have presented the renewable energy design optimization of
of EVs is explored for integration with solar-powered buildings net-zero energy buildings proposing the grid-protective energy man­
including charging forms, charging limits and commuting distances. A agement strategy with novel assessment indicators, such as PV utiliza­
case building community in Sweden indicates that smart charging with tion ratio, load match ratio and grid flexibility factor. The quantitative
small charging limits enables reduced peak energy exchange with slight impact of vehicle-to-building interaction on EV degradation and system
impact on battery degradation [27]. The feasibility and design of a performance still needs to be clarified considering stochastic driving
residential building with rooftop PV and EV charging is studied applied schedules for net-zero energy building applications. Furthermore, the
in a typical Malaysian house. The economic and environmental perfor­ techno-economic-environmental feasibility of integrating static battery
mance of two scenarios with mono and thin film PV systems integrating storage with net-zero energy buildings powered by the PV-EV system is
EV charging is clarified considering a simplified EV charging loss [28]. A seldom studied by conducting multi-objective optimizations for the
mixed integer linear programming building energy management system trade-off of PV utilization, grid flexibility and lifetime energy cost.
is proposed to manage a grid-connected smart building integrating
renewable energy supply, EVs and energy storage. The stochastic driving 1.4. Research scope and contribution
schedule of EVs is considered and the authors reported that EVs perform
better for grid integration and load coverage in the case of requesting no To fill the above research gaps, this study proposes a design man­
grid power from 9 am to 2 pm [29]. agement and optimization framework of developing renewable energy
Optimization sizing of renewable energy systems has been investi­ systems for net-zero energy buildings integrating EVs and battery stor­
gated to obtain optimum solutions for achieving favorable performance age. A building load data augmentation model is developed to obtain the
regarding the technical [30,31], economic [32] and environmental [33] annual hourly load profile of the net-zero energy building based on the
aspects. Specifically, the real-time energy optimization of a smart home on-site collected data adopting the Gate Recurrent Unit (GRU) neural
with PV-wind power supply and EV integration is conducted to optimize network. A grid-protective energy management strategy is developed for
the system cost, thermal comfort, and battery and EVs charging/dis­ an office building equipped with rooftop PV panels, EVs and static
charging. The applicability of the proposed model is identified via batteries proposing novel assessment indicators. Multi-objective opti­
comparison with other models in different scenarios and weather con­ mizations are conducted to explore optimum sizing of the PV-EV-battery
ditions [30]. A robust and effective multi-objective modified firefly al­ system for balanced tecno-economic performance applied in the net-
gorithm is proposed to optimize the hybrid battery and hydrogen zero energy building. The research scope and contribution is explained
storage for a real PV system installed in a grid-connected warehouse, as below:
and the PV self-sufficiency ratio is adopted to evaluate the on-site uti­
lization of PV power generation. The research results show that (1) This study develops a transient energy design and management
hydrogen storage in the long term can benefit the storage degradation, model with rooftop PV panels, EVs and static battery storage to
power efficiency and system economy [31]. The technical and economic achieve net-zero energy operation for a practical office building.
feasibility of developing net-zero energy hotel building in Egypt is The grid-protective energy management strategy is presented to
investigated powered by the multi-energy system including PV, wind maintain grid power flexibility proposing a novel evaluation in­
and biogas generators. The sizing of the multi-energy system is opti­ dicator applicable for time-of-use operations totaling the grid
mized via minimizing the net present cost and levelized cost of energy, flexibility factor of peak, flat and valley hours. The novel in­
where net grid purchase is considered for grid integration assessment dicators of PV utilization ratio and load match ratio are also
[32]. The impact of weather variability on the optimal design sizing of proposed to evaluate PV power utilization efficiency and effective
renewable energy systems applied in office buildings is identified via

4
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

building load coverage under the grid-protective energy man­ flowchart shown in Fig. 4. A building load data augmentation model is
agement strategy. firstly developed to obtain the annual hourly load profile of a typical
(2) This study clarifies the quantitative impact of vehicle-to-building campus building based on the partially collected load data adopting the
interaction on the EV degradation and performance of the net- GRU neural network. The transient design simulation model of the net-
zero energy building via comparison between the baseline case zero energy building integrating rooftop solar PV, EVs and static battery
without EV discharging and the case allowing bidirectional storage is then established proposing a grid-protective energy manage­
charging/discharging. The stochastic mobility schedules of EVs ment strategy to maintain high grid flexibility. Novel assessment in­
in five groups are considered including stochastic arriving time dicators are proposed for net-zero energy buildings under time-of-use
and parking time, and the annual degradation rate of EVs is operations including the PV utilization ratio, load match ratio and grid
estimated based on the Rainflow Counting algorithm. flexibility factor. Afterwards, multi-objective optimizations are con­
(3) This study develops multi-objective optimizations of the net-zero ducted to explore optimal system configurations for balanced perfor­
energy building for optimum sizing configurations of PV panels, mance on the PV utilization, grid integration and lifetime system cost.
EV numbers and battery capacity for achieving balanced perfor­ The sensitivity analysis is presented to identify the interactive impact of
mance on renewable energy utilization, grid flexibility and lev­ system configurations on techno- economic performance of the net-zero
elized cost of energy. The interactive impact of component sizing energy building.
configurations on the system performance is identified via Finally, the techno-economic-environmental feasibility of three net-
sensitivity analysis. The technical, economic, environmental zero energy building cases is analyzed and compared including the
feasibility of utilizing the PV-EV-battery system in the net-zero baseline net-zero energy building case with unidirectional EVs for
energy building is investigated. charging only (Case 1), the net-zero energy building case with direc­
tional EVs for charging and discharging (Case 2), the optimum net-zero
2. Methodology energy building case with optimum configurations (Case 3). The pro­
posed transient design management and optimization framework of the
This study aims to develop renewable energy design management renewable energy system provides a possible pathway for the typical
and optimization framework for a net-zero energy building, with the campus building to achieve zero-carbon operations. The comprehensive

Fig. 4. Flowchart of renewable energy design and optimization for a net-zero energy building.

5
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

techno-economic-environmental feasibility analysis offers significant ④ A new dataset is generated incorporating the hourly load and
guidance and reference for relevant stakeholders to develop carbon­ weather data of the collected hourly data of 9 months and predicted
–neutral buildings. hourly data of 10 days (i.e., 01 Jan. to 10 Mar. & 01 Jun. to 31 Dec.).
⑤–⑧ The obtained dataset is used to train the new load prediction
2.1. Weather profile and electrical load of the net-zero energy building model with GRU neural network model to predict the hourly load of
the next 10 days (i.e., 11 Mar. to 20 Mar.). The new dataset can be
The net-zero energy building is located in Guangzhou with favorable then generated including hourly data and weather data of the
solar resources as shown in Fig. 5 obtained from the weather data in collected hourly data of 9 months and predicted hourly data of 20
TRNSYS 18 based on the Meteonorm data from the local weather station days (i.e., 01 Jan. to 20 Mar. & 01 Jun. to 31 Dec.).
[34]. The local monthly horizontal solar radiation varies between 62.67
kWh/m2 and 167.85 kWh/m2 with the minimum in February and Finally, the hourly load data of the missing three months (i.e., 01
maximum in July. The daily average temperature is within the range of Mar. to 31 May) can be predicted after numbers of the above loop
13.29 ◦ C–28.56 ◦ C with higher temperature in May to October in the execution. The validity of the proposed building load data augmentation
hot-summer and warm-winter region. model is ensured via manually removing the real load data of one month
A typical administrative building with six floors in Guangzhou Uni­ from the dataset to construct a validation set, and then use the validation
versity is adopted as the research object to study the feasibility of set to validate the prediction performance of the model. It is also verified
installing PV-EV-battery systems to achieve net-zero energy operations. via comparison with the on-site manually collected monthly load data
The total rooftop area of the building is about 4876 m2 with 3900 m2 obtained from the logistics department. The detailed code and data of
available for installing PV panels considering a 20 % area for necessary the building load data augmentation model are provided in the sup­
rooftop facilities and maintenance [35]. The planned car parking plementary material.
number in the underground of the building is 162 and ground parking is The annual hourly building load data are therefore generated as
also available around the building. The location and layout of the outputs of the building load data augmentation model including
building is shown in Fig. 6. collected hourly load of 9 months and predicted hourly load of 3 months
A building load data augmentation model is developed to obtain the as shown in Fig. 8. It is indicated that the total annual load of the
full annual hourly profile of the net-zero energy building adopting the building is about 687,611.04 kWh with the maximum monthly load at
GRU neural network based on the partially collected data from the lo­ about 300,41.68 kWh in February and maximum monthly load at about
gistics department in the university. The framework of the building load 104,441.30 kWh in September. The monthly load from May to October
augmentation model is shown in Fig. 7. The inputs of the model include is relatively higher than other months in accordance with the local
collected hourly load data of 9 months (from 01 Jan. to 28 Feb & 01 Jun. weather.
to 31 Dec. 2022), collected monthly load data of 12 months (from Jan. to
Dec.), and annual hourly weather data of the location. The hourly
2.2. Renewable energy system for the net-zero energy building integrating
building load data from 01 Mar. to 31 May are missing due to unex­
electric vehicles and static battery storage
pected breakdown of data collectors, and it is obtained by the developed
data augmentation model in following steps:
Three typical net-zero energy building cases are developed with
different system configurations as shown in Table 1. Case 1 is developed
① Firstly, the load prediction model is trained with the GRU neural
as the baseline net-zero energy building case with a 700 kW rooftop PV
network model based on the collected hourly load and weather data
system to keep an annual energy balance with the building load and EV
of 9 months (i.e., 01 Jan. to 28 Feb. & 01 Jun. to 31 Dec.).
travelling load. EVs are integrated for charging only to ensure daily
② The hourly load data of the next 10 days (i.e.,01 Mar. to 10 Mar.)
cruise, and the EV number is 162 determined by the planned under­
are then predicted with the trained load prediction model.
ground parking spots. Case 2 is developed with the same PV capacity
③ The hourly load data and weather data of the next ten days (i.e.,
and EV number with Case 1, while the bidirectional charging and dis­
01 Mar. to 10 Mar.) are combined as inputs of the following pre­
charging of EVs is allowed in Case 2 to study the impact of vehicle-to-
diction batch.
building interaction. Case 3 is developed with optimum configurations
on the PV capacity, EV number and static battery to study the impact of

Fig. 5. Monthly solar radiation and daily temperature of the building location.

6
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 6. A typical net-zero energy campus building with renewable energy system.

Fig. 7. Building load data augmentation model with GRU neural network.

static battery storage for integration with the renewable energy system. single module [37]. The weather conditions of dry bulb temperature,
(1) Solar PV power generation beam radiation, sky diffuse radiation, ground reflected diffuse radiation
The rooftop PV system is installed in the building as the power supply on a titled angle of 23◦ similar to the local latitude are adopted as inputs
given the advantageous solar resources in the location. The electrical to determine the photovoltaic source and current–voltage characteris­
performance of rooftop PV panels is simulated adopting TRNSYS Type tics. The performance of the multi-module PV array is extrapolated
103b operating at the maximum power point [36]. An empirical based on the results of the single module equivalent circuit. The
equivalent circuit model consisting of a direct current source, diode and installation number of PV panels of net-zero energy cases is determined
a resistor is employed to predict the current–voltage characteristics of a to keep an annual energy balance with the building demand and EVs

7
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 8. Electrical load of the net-zero energy building in 2022.

Table 1 ( (υ− 2)/2

) tarr,gi • e− tarr,gi /2
System configurations of three cases for the building. f tarr,gi = υ/2 (1)
2 • Γ(υ/2)
System PV capacity EV number Battery capacity ∫∞
configuration where Γ(•) is the Gamma function defined as Γ(x) = 0 tx− 1 e− t dt, x
> 0, tarr,gi is the arriving time of gi-th EVs, υ is the degree of freedom with
Case 1 - 700 kW 162 (Charging only) 0
unidirectional a value of 4 [29].
EV The parking hours of EVs integrated with the net-zero energy
Case 2 - 700 kW 162 (Charging/ 0 building are described following the normal distribution with a mean of
bidirectional discharging)
8 h and a variance of 4 h assuming that most EVs would park in the
EV
Case 3 - optimum Optimization Optimization variable Optimization
building for 8 h [29]. It is calculated that the probability distribution of
variable (Charging/ variable EVs parked in the building can be described as Fig. 9.
discharging) The cycling degradation process of EVs is a cumulative fatigue pro­
cess, where the annual degradation rate calculates the sum of the spe­
cific degradation rate from all cycles. It is difficult to identify the number
cruise load. The installation number of PV panels of the optimal case is
of cycles from the dynamic depth of discharge (DoD) dataset of EVs
determined by the multi-objective optimization considering the acces­
because of irregular charging and discharging operations. The Rainflow
sible installation area in the rooftop.
Counting algorithm is adopted to effectively extract the number of
(2) Electric vehicle integration with stochastic characteristics
equivalent cycles from the DoD dataset of EVs, as an effective algorithm
and degradation evaluation
originally developed for calculating fatigue cycle numbers during a load-
Electric vehicles (EVs) can be integrated with the building as mobile
time history [41]. The effectiveness of the Rainflow Counting algorithm
energy demand and energy supply units. The electrical performance of
in modelling battery degradation has been validated in the battery dy­
EVs is modeled using TRNSYS Type 47 corresponding to a simple energy
namic stress test [42] and related research [27,43]. The flowchart of the
balance to specify the dynamic EV state of charge varying with the input
Rainflow Counting algorithm is shown in Fig. 10 based on the MATLAB
charging and discharging power. The detailed parameters are defined
platform. Firstly, the hourly DoD dataset of EVs is used as inputs ob­
according to a commercialized EV model of “Tesla Model S 75” with a
tained from transient simulations of the renewable energy system in
rated capacity of 75 kWh, a cruise range of 360 km, and a maximum
TRNSYS. Then data preprocessing on the DoD dataset is conducted
charging power rate of 0.22C [38]. The daily travelling distance of EVs
including peak-valley reduction by identifying slop reversals, recon­
integrated with the building is about 17.4 km/day based on the local
struction begin with maxima and peak-valley reduction to ensure
traffic report [39], and the vehicle consumption is 0.201 kWh/km. The
smooth reconstruction. The preprocessed DoD dataset is then passed to
maximum charging state of charge of EVs is 0.95 with a charging effi­
the Rainflow Counting algorithm to extract the ranges of DoD and cycle
ciency of 0.9 [40], and the minimum discharging state of charge is 0.2 to
information of each range [44]. N is the length of DoD counting the
ensure safe operations and one-day travelling.
number of the remaining points of the dataset, and j is a specific point in
The stochastic mobility schedule of EVs is considered including
the dataset. Finally, the degradation of cycle-life for each cycle is ob­
stochastic arriving time and parking time for energy integration with the
tained and summed up to estimate the annual degradation rate of EVs
net-zero energy building. The official working hours during weekdays in
(DegEVs) according to Eq. (2) as below:
the university is 8:30 am to 4:30 pm, and more vehicles arrive at the
building at around 8:00 am. It is assumed that EVs parked in the building ∑c
Cycle of Rz
DegEVs = B × 100% (2)
are classified into five groups with different arriving time and parking z=1 A × (Rz )
hours to make energy charging and discharging. The daily arriving time
of EVs integrated with the net-zero energy building is predicted in the where Rz is the range of z-th cycle in the DOD time-series profile of
form of chi-square distribution with the probability density function EVs. c is the total number of cycles. A and B are empirical parameters
expressed as Eq. (1): determining rated cycle-life at different DoD ranges at the value of 512.2
and − 1.9928, respectively [27].

8
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 9. Probability distribution of EVs parked in the net-zero energy building.

Fig. 10. Rainflow Counting algorithm for calculating degradation rate of EVs.

(3) Static battery storage time is covered by battery storage in Case 3 when battery storage is
The static battery storage is integrated with the renewable energy integrated. The unsatisfied load is then covered by EVs with extra
system to study the techno-economic-environmental feasibility applied electricity storage in Case 3 and Case 2 when bidirectional charging of
in the net-zero energy building. The minimum and maximum state of EVs is adopted. The residual load is finally covered by the utility grid.
change of the lithium-ion battery is 0.15 and 0.95, respectively. The The utility grid is also responsible for the necessary electricity demand
charging and discharging rate of the static battery is the same as EVs at of EVs to ensure daily cruise.
0.22C. The installation capacity of the static battery storage is deter­ During flat time, less grid import and less grid export is preferred to
mined by the multi-objective optimization as explained in Section 2.3. maintain high energy autonomy of the net-zero energy building and
(4) Grid-protective energy management strategy of the renew­ therefore relying less on the utility grid. Surplus PV power is used to
able energy system charge the parked EVs and then charge static battery storage (only in
The grid-protective energy management strategy of the renewable Case 3 integrated with static battery storage). The unused PV power is
energy system is proposed to maintain higher grid flexibility based on finally exported into the utility grid. The static battery storage (Case 3)
the local time-of-use schedule of the power sector as per Fig. 11. The and EVs (Case 2 and Case 3) can be discharged to supply power for the
rooftop PV generation is firstly supplied for the building load in all pe­ unmet building load during flat time when PV power is not enough. The
riods and cases. The detailed energy flow then differs with time and grid acts as the final energy backup for the building to ensure power
system configuration. During peak time with high grid pressure, the reliability for the building.
utility grid prefers more energy export and less energy import. Surplus During valley time, more grid import and less grid export is preferred
PV power is directly exported into the utility grid to support the grid to absorb more grid electricity to avoid frequent generator starts and
with relatively high demand in this period. The unmet load during peak stops in the power generation sector. Surplus PV power is used to charge

9
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

E grid to load E grid to EVs E PV to grid peak with RE E grid to load E grid to EVs E PV to grid flat with RE E grid to load E grid to EVs E PV to grid valley with RE
E grid to load E grid to EVs peak no RE E grid to load E grid to EVs flat no RE
E grid to load E grid to EVs valley no RE

Fig. 11. Grid-protective energy management strategy of the renewable energy system.

EVs and static battery storage (Case 3) before being exported into the charge static batteries for energy storage.
utility grid. The residual building load exceeding the PV generation is
directly covered by the utility grid, and the grid power is also utilized to

Fig. 12. Flowchart of multi-objective optimization of the renewable energy system.

10
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

2.3. Multi-objective optimization and evaluation indicators of the (2) Load match ratio evaluating building load coverage
renewable energy system for the net-zero energy building In order to relieve the utility grid, the building load imports power
from the utility grid during valley periods after covered by the PV sys­
The multi-objective optimization of the renewable energy system is tem, rather than discharging EVs or batteries. Therefore, the load match
conducted based on the coupled simulation and optimization platform ratio (LMR) is proposed to evaluate the effective load coverage of the
to explore optimum sizing configurations and balanced techno- net-zero energy building including load covered by the PV system, static
economic performance for the net-zero energy building as per Fig. 12. battery storage, EVs and utility grid in valley hours as shown in Eq. (4):
The transient simulation model of the net-zero energy building is
EPV to load + Ebattery to load + EEVs to load + Egrid to load
firstly established based on the TRNSYS 18 platform. Three key sizing (4)
valley
LMR =
Eload
parameters are selected as optimization variables as defined in the
parameter tree including the PV capacity ranging from 500 kW to 1500 where Ebattery to load calculates the load coverage by static battery
kW at a step of 50 kW, the EV number ranging from 100 to 300 at a step storage, kWh. EEVs to load calculates the load coverage by EVs, kWh. Egrid to
of 20, the battery capacity ranging from 200 kWh to 1000 kWh at a step load_valley calculates the load coverage by the utility grid in valley hours,
of 50 kWh. Three important indicators are adopted as optimization kWh. Eload is the annual electrical load of the building, kWh.
objectives extracted from the TRNSYS model including the PV utiliza­ (3) Grid flexibility factor evaluating grid integration
tion ratio, grid flexibility factor and levelized cost of energy. A dimensionless indicator of grid flexibility factor (GFF) is proposed
The Non-dominated Sorting Genetic Algorithm II (NSGA-II) is to evaluate the grid integration performance of the net-zero energy
adopted as the evolutionary optimization algorithm based on the jEplus building powered by the PV-EV-battery system compared with the
+ EA platform given its robustness and efficiency as well as smooth baseline case without renewable power supply. It calculates the grid
integration with TRNSYS [45]. It improves the adaptive fit of candidate flexibility factor of peak hours, flat hours and valley hours under the
populations as a robust and elite multi-objective genetic algorithm with grid-protective time-of-use energy management strategy. During peak
fast non-dominated sorting approach, fast crowded distance estimation hours, GFFpeak is defined as the ratio of net grid import with renewable
procedure and simple crowded comparison operator [46]. The multi- supply to grid import without renewable supply as per Eq. (5). The
objective optimization starts with creating the initial population with GFFpeak value less than 1 is preferred indicating lower net grid import
random sampling within the full searching range of 1329 solutions. It during peak hours compared with the baseline without renewable
then runs the TRNSYS simulation on each solution and evaluates the supply.
optimization objectives with stochastic ranking to generate and store
(Egrid to load peak + Egrid to EVs peak − EPV to grid peak )with RE
new populations in the archive [47,48]. The solutions are selected for GFF peak = (5)
(Egrid to load peak + Egrid to EVs peak )no RE
reproduction considering a crossover rate of 0.9 and a mutation rate of
0.05 to create new variables based on the Pareto archived global elitism where Egrid to load_peak is grid imported energy to cover building load in
strategy [49]. The design optimization process is terminated when the peak hours, kWh. Egrid to EVs_peak is grid imported energy to charge EVs in
optimization objectives are satisfied to obtain the Pareto optimal set, peak hours, kWh.
and convergence is also maintained in the optimization process to reach During flat hours, GFFflat is defined as the ratio of total grid exchange
the desired level of stability where the population is stable compared to with renewable supply to grid import without renewable supply as per
the previous one [50]. There are totally 1359 solutions evaluated to Eq. (6). The GFFflat value less than 1 is preferred indicating lower total
compare the optimization objectives and 423 solutions are selected in power exchange between the net-zero energy building and the utility
the Pareto optimal set indicating the condition under which no grid compared with the baseline. Namely, the net-zero energy building
improvement on one objective can be made without making worse on all with a lower GFFflat value has higher energy autonomy during flat pe­
other objectives. The Pareto optimal solutions provide important guid­ riods, which is preferred by the utility grid.
ance to design the renewable energy system applied in the net-zero
(Egrid to load flat + Egrid to EVs flat + EPV to grid flat )with RE
energy building with trade-off in the technical and economic perfor­ GFF flat = (6)
(Egrid to load flat + Egrid to EVs flat )no RE
mance. The final optimum solution is achieved based on the decision
making strategy of the minimum distance to the utopia point method where Egrid to load_flat is grid imported energy to cover building load in
[22]. The sensitivity analysis is conducted to further investigate the flat hours, kWh. Egrid to EVs_flat is grid imported energy to charge EVs in
interactive impact of renewable energy configurations on the system flat hours, kWh. EPV to grid_flat is grid exported energy from PV generation
techno-economic performance. in flat hours, kWh.
The techno-economic-environmental indicators of the renewable During valley hours, GFFvalley is defined as the ratio of net grid import
energy system are presented to evaluate the feasibility of the proposed with renewable supply to grid import without renewable supply as per
energy system applied in the net-zero energy building. Eq. (7). The GFFvalley value higher than 1 is preferred indicating higher
(1) PV utilization ratio evaluating power generation utilization net grid import during valley hours compared with the baseline without
The grid-protective energy management strategy is adopted in the renewable supply.
net-zero energy building, where surplus PV generation is exported into
(Egrid to load valley + Egrid to EVs valley − EPV to grid valley )with RE
the utility grid after meeting the building load in peak hours to relive GFF valley = (7)
grid pressure. The PV utilization ratio (PVUR) is proposed to evaluate (Egrid to load valley + Egrid to EVs valley )no RE
the PV power utilization efficiency including PV power consumed by the where Egrid to EVs_valley is grid imported energy to charge EVs in valley
building load, static battery storage, EVs and utility grid in peak hours as hours, kWh. EPV to grid_valley is grid exported energy from PV generation in
explained in Eq. (3): valley hours, kWh.
EPV to load + EPV to EVs + EPV to battery + EPV to grid The total grid flexibility factor of the net-zero energy building con­
(3)
peak
PVUR =
EPV siders the grid flexibility factor of peak, flat and valley hours as per Eq.
(8):
where EPV to load is energy from PV generation to cover building load,
kWh. EPV to EVs is energy from PV generation to charge EVs, kWh. EPV to GFF = GFF peak + GFF flat − GFF valley (8)
battery is energy from PV generation to charge static battery storage, kWh. (4) Annual electricity bill and levelized cost of energy evalu­
EPV to grid_peak is energy from PV generation to be exported into grid in ating economic feasibility
peak hours, kWh. EPV is total annual generation from the PV system The annual electricity bill (EB) is calculated as per Eq. (9) to show the
considering an inverter efficiency of 0.95.

11
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

time-of-use cost of imported energy from the utility grid. of three cases. It considers different carbon emission factors for the time-
of-use operations according to the local power industry. The peak-valley
EB = Egrid out • Prpeak + Egrid outflat • Prflat + Egrid outvalley • Prvalley (9)
peak
carbon emission factor is clarified as more effective than the average
where Egrid out_peak, Egrid out_flat, Egrid out_valley is grid imported energy carbon emission factor for characterizing the carbon footprint of the
during peak, flat and valley time, respectively in kWh. It is regulated by power system. The specific decarbonisation potential results of three
the local government that the peak period of the power system is net-zero energy building cases are compared in Section 3.3.
10:00–12:00 and 14:00–19:00, the valley period of the power system is
CE = Enet grid importpeak • CEF peak + Enet grid importflat • CEF flat + Enet grid importvalley
0:00–8:00 and the residual time is the flat period. Prpeak, Prflat, Prvalley is
the unit price of grid electricity at 0.1587 US$/kWh, 0.0938 US$/kWh • CEF valley
and 0.0361 US$/kWh, respectively [51]. (11)
The lifetime levelized cost of energy (LCOE) calculates the net pre­
where Enet grid import_peak, Enet grid import_flat, and Enet grid import_valley is net
sent value (NPV) of the renewable energy and storage system at the scale
grid imported energy during peak hours, flat hours and valley hours,
of generated electricity over a 20-year lifetime as per Eq. (10-1). NPV is
respectively in kWh. CEFpeak and CEFflat is the carbon emission factor of
the net present value of the renewable energy and storage system as
grid electricity in peak hours and flat hours, both at 0.62 kg/kWh ac­
shown in Eq. (10-2) including present value of investment cost of key
cording to the local power industry [11]. CEFvalley is the carbon emission
components (i.e., PV panels, inverters and batteries), present value of
factor of grid electricity in valley hours, at 0.55 kg/kWh for the net-zero
electricity bills for grid imported energy (PRVEB) and present value of
energy building [11].
Feed-in-Tariff (FiT) cost of renewable energy installation (PRVFiT). The
investment cost of key components includes present value of initial cost
3. Results and discussion
(PRVini), present value of operational and maintenance cost (PRVO&M),
present value of replacement cost (PRVrep), present value of residual cost
3.1. Multi-objective optimization results of the renewable energy and
(PRVres). It is assumed that the investment cost of EVs is covered by
storage system
vehicle owners, while its electricity charging cost for daily cruise is
included. The local FiT policy for renewable energy installation is
Multi-objective optimization is further conducted to explore optimal
adopted including FiT for PV capacity installation (FiTpow) and FiT for
configurations of the renewable energy system for power supply to the
annual PV generation (FiTgen) during first 6 years [52]. The detailed
net-zero energy building adopting bidirectional power interaction with
definition of each parameter is explained in Table 2.
EVs and static battery storage. Fig. 13(a) shows the distribution of
NPV optimization objectives including the three-dimensional Pareto optimal
LCOE = ∑n=N (10-1)
EPV •(1− δPV )n− 1
surface and its projections, indicating an obvious trade-off among the
(1+i)n
searching objectives. Fig. 13(b) shows the distribution of optimization
n=1

NPV = PRV ini + PRV O&M + PRV rep − PRV res + PRV EB − PRV Fit variables corresponding to the Pareto optimal surface.
It can be found that most of Pareto optimal solutions are achieved
∑
n=Y ∑
k=K with relatively high EV numbers. The decision-making strategy of the
fmai • Cini 1 − d k•l lres
= Cini + n + Cini ( ) − Cini minimum distance to the utopia point method is adopted to determine
(1 + i) 1+i l
n=1 k=1 the overall optimum solution as the final optimum configuration of the
(1 − d)Y ∑ n=Y
EBa • (1 + γ)n− 1 net-zero energy building. Another three optimum solutions with best
• Y + − (FiT pow •CapPV + FiT gen
(1 + i) n=1
(1 + i)n performance on PVUR (in green color), GFF (in blue color) and LCOE (in
∑n=6 EPV • (1 − δPV )n− 1 cyan color) respectively are also highlighted to make a comparison.
• ) (10-2) The distribution of variables and objectives of the Pareto optimal set
n=1 (1 + i)n
is shown in Fig. 14. It can be found that majority of the optimal PV
(5) Annual equivalent time-of-use carbon emission evaluating capacity ranges within 600–1150 kW with an average of about 875 kW.
decarbonisation potential Most of the optimal EV number is within the range of 280–300. Majority
The annual equivalent time-of-use carbon emissions of the renew­ of the static battery storage lies in the dimension of 350–750 kWh with
able energy system for power supply to the net-zero energy building is an average of about 563 kWh. In terms of the system performance in­
calculated as shown in Eq. (11) to evaluate the decarbonisation potential dicators, the minimum LCOE of the Pareto optimal solutions is about
0.0940 US$/kWh with the minimum battery capacity and maximum PV
capacity. The maximum LCOE is about 0.2579 US$/kWh with less PV
Table 2 panels and more batteries, where the maximum PV utilization ratio is
Parameters for calculating economic indicator of the renewable energy system. achieved. The centralized area of LCOE is about 0.1219–0.1800 US
Components Cini (initial fmai (O&M ratio) l (lifetime), $/kWh. GFF of the Pareto optimal solutions is within the range of − 1.00
cost) year to − 3.43, indicating better performance in grid integration performance
PV 857 US$/kW 2% 20 of the renewable energy system integrating EV and static battery stor­
inverter 120 US$/kW 1% 10 age. It is indicated that PVUR of Pareto optimal solutions is higher than
battery 400 US 1% 5
0.81958, and majority of the solutions even arrive to approximately 1
$/kWh
Parameters Value Parameter Value indicating good PV power utilization efficiency. Specifically, nearly 92
n (a specific year) 1–20 dbattery (price 0.1015 % of the optimal solutions achieve a PV utilization ratio at over 90 %.
degression rate) Table 3 shows detailed results of the Pareto optimal solutions and
Y (system lifetime) 20 years dinveter (price 0.05 four optimum solutions including optimization variables and objectives.
degression rate)
δPV (PV degradation 1 %/year lres (residual lifetime) By component
It can be found that the overall optimum solution for the net-zero energy
rate) building to achieve balanced techno-economic performance is installing
i (real discount rate) 5.8 %/year γ (electricity rising 1.4 %/year 1050 kW rooftop PV power, 300 EVs and 450 kWh static battery storage.
rate)
k (one replacement 1-K FiTpow (FiT for PV 0.02886 US
time) installation) $/W
K (total replacement By FiTgen (FiT for PV 0.02165 US
times) component generation) $/kWh

12
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 13. Multi-objective optimization results of the PV and storage system for the building (a) optimization objectives, (b) optimization variables.

13
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 14. Distribution of variables and objectives of Pareto optimal for the building.

Table 3
Pareto optimal and four optimum solutions of the net-zero energy building.
Optimization solutions Optimization variables Optimization objectives

PV capacity, kW EV number Battery capacity, kWh PV utilization ratio Grid flexibility factor LCOE, $/kWh

Pareto optimal solutions 500–1500 140–300 200–1000 0.81958–0.99997 − 3.43 to 0.0940–0.2579

− 1.00
Overall optimum 1050 300 450 0.95638 − 2.68 0.1193
PVUR optimum 500 280 950 0.99997 − 1.55 0.2579
GFF optimum 1500 300 850 0.87923 − 3.43 0.1220
LCOE optimum 1500 140 200 0.81958 − 2.62 0.0940

3.2. Interactive impact of renewable energy system configurations on impact of renewable energy configurations on each system indicator.
techno-economic performance of the building Fig. 15 shows the interactive impact of installed PV capacity and battery
capacity on the techno-economic performance of the building, where a
Sensitivity analysis is conducted to further explore the interactive fixed EV number of 300 is adopted based on the optimization results. It is

Fig. 15. Interactive impact of PV and battery capacity on the system performance.

14
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

indicated that the system LCOE during a 20-year lifetime achieves better the techno-economic performance of the building is shown in Fig. 17,
results with a lower battery capacity and higher PV capacity. Because where a fixed PV capacity of 1050 kW is adopted. It can be found that the
the system electricity bill can be reduced and FiT can be increased with system LCOE is within the range of 0.1023–0.1523 US$/kWh when the
the rising PV capacity, and investment on battery storage can be reduced battery and EV capacity changes, and the minimum is achieved with the
with less battery installation. Specifically, a minimum LCOE is obtained minimum EV number (100) and minimum battery capacity (200 kWh).
at about 0.0952 US$/kWh with 200 kWh battery storage and 1500 kW Under the fixed PV installation of 1050 kW with more annual PV gen­
PV panels, while the configuration of 500 kW PV and 1000 kWh battery eration than annual demand, integrating more EVs and static battery
storage results in an LCOE of up to 0.2680 US$/kWh. GFF ranges from storage would obviously improve grid flexibility. Specifically, a mini­
about − 0.96 to − 3.44 when the installed PV panels and batteries mum GFF at about − 2.81 is obtained with the configuration of 300 EVs
change, reaching better results with more batteries and more PV panels. and 1000 kWh batteries. PVUR ranges from 0.8393 to 0.9647 with the
It is shown that higher PV and battery installations contribute to better change of EV numbers and battery capacity, reaching the maximum with
grid flexibility for the building. PVUR decreases with the increasing PV 200 kWh batteries and 300 EVs.
capacity, achieving good results (over 0.8790) under the grid-protective
time-of-use energy management strategy. PVUR is weakly sensitive with
the battery capacity, because few PV generation is charged into battery 3.3. Techno-economic-environmental performance of the renewable
storage under the strategy charging the static battery storage during energy system for the building
valley hours. It can be found that installing more PV panels in the
building contributes to improving the grid flexibility and system econ­ (1) PV utilization efficiency and building load match of the net-
omy, while decreasing the PV power utilization efficiency. Integrating zero energy building
more static battery storage in the building enables higher grid flexibility The PV consumption of the net-zero energy building is shown in
and lower system economy, and shows weak impact on the PV utiliza­ Fig. 18. The PV self-consumption ratio in three cases ranges from 49.71
tion efficiency. % to 59.93 %, while PVUR ranges from 95.54 % to 98.02 % considering
Fig. 16 shows the interactive impact of installed PV capacity and EV the effective utilization of PV energy export during peak hours. PVUR is
number on the techno-economic performance of the building, where a more applicable for assessing PV power utilization efficiency under the
fixed battery capacity of 450 kWh is adopted. It is indicated that LCOE grid-protective energy management strategy. Specifically, the PV self-
shows an obvious inverse correlation with the PV capacity, as the life­ consumption ratio in Case 1 is about 57.45 %, with most of the rest
time NPV of electricity bills and FiT decreases with the rising PV gen­ exported into the utility grid during peak hours to relive the grid
eration. LCOE shows a slightly positive correlation with the EV number, resulting in a PVUR at up to 95.54 %. The PV self-consumption ratio and
mainly because LCOE considers only the electricity cost for vehicle PVUR in Case 2 are slightly higher than Case 1 at about 59.93 %
cruise and assumes that the vehicle investment cost is covered by vehicle (improved by 4.33 %) and 98.02 % (improved by 2.60 %) respectively
owners. Therefore, a minimum LCOE is achieved at about 0.1036 US after introducing EV discharge. PVUR keeps at a high level at 95.64 % in
$/kWh with the configuration of 100 EVs and 1500 kW PV power. In the optimum case (Case 3) introducing more PV power and hybrid EV
terms of the grid flexibility, a minimum GFF at about − 3.34 can be and battery storage. Only 0.03 % of PV power is used to charge static
obtained with 1500 kW PV power and 300 EVs, indicating that inte­ battery storage, as the battery storage is charged by the utility grid
grating more PV panels and EVs contributes to higher grid flexibility. It during valley hours for high grid flexibility and economy under the grid-
should be noted that integrating more EVs contributes to higher grid protective energy management strategy.
flexibility when the PV capacity is relatively large, while it shows an The load coverage of the net-zero energy building is shown in Fig. 19.
inverse tendency when the PV capacity is relatively small. PVUR ranges The on-site load cover ratio of three cases ranges from 53.55 % to 76.69
from 0.79244 to 0.99997 with the varying PV and EV capacity, reaching %, while LMR ranges from 73.79 % to 96.11 % considering effective load
the maximum with the configuration of 300 EVs and 500 kW PV power. coverage from the utility grid during valley hours. LMR is more appli­
It can be found that EV integration with the building contributes to cable for assessing load coverage under the grid-protective energy
improving the PV power utilization and slightly increasing the system management strategy. Specifically, the building load is firstly covered by
LCOE, while its impact on the grid flexibility depends on the PV the PV generation with a load cover ratio at about 53.55 % in Case 1. The
capacity. load cover ratio is about 62.47 % in Case 2, increased by about 8.91 %
The interactive impact of static battery capacity and EV number on than Case 1 after introducing vehicle discharge for vehicle-to-building
interaction. Around 20.23 % of the building load is covered by

Fig. 16. Interactive impact of PV and EV capacity on the system performance.

15
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 17. Interactive impact of battery and EV capacity on the system performance.

Fig. 18. PV consumption of the net-zero energy building.

importing energy from the utility grid during valley hours in the net-zero ranges from 0.16 % to 1.65 % calculated by the Rainflow Counting al­
energy building cases (Case 1 and Case 2), resulting in an LMR of about gorithm, which is reasonable for short-distance EVs as reported in the
73.79 % and 82.70 %, respectively. LMR is increased by about 12.08 % literature [27]. The annual degradation rate of EVs in Case 1 with only
by introducing the vehicle-to-building interaction. The load cover ratio charging functions varies between 0.16 % and 0.66 % with the
and LMR in the optimum case (Case 3) is further increased to 76.69 % maximum in group 4 with more charging availability in parking
and 96.11 % respectively, when battery storage is integrated with the schedules. The annual degradation rate of EVs in Case 2 is within 0.29
building. About 16.69 % of the building load is satisfied by the static %–1.65 %, obviously higher than that in Case 1, because of the EV
battery storage, and most of battery charging energy is supplied by the discharging for the building load. The annual degradation rate of EVs in
utility grid during valley hours. Only 0.30 % of the building load is Case 3 is reduced compared with Case 2, as the static battery storage
supplied by the utility grid in peak time in Case 3 to achieve the best grid takes the major role in energy storage making use of the off-peak grid
flexibility. energy. The detailed distribution of EV SOC in group 4 is explained as an
(2) EV degradation of the net-zero energy building example as per Fig. 20(b). The EV SOC curve in Case 1 shows the dy­
Fig. 20(a) shows the annual degradation rate of EVs in five groups of namic distribution of EV charging by the PV system or utility grid and EV
three cases. It can be found that the annual degradation rate of EVs discharging for daily cruise, as EV integration with the renewable energy

16
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 19. Load coverage of the net-zero energy building.

Fig. 20. Comparison of EV degradation and SOC integrated with the net-zero energy building.

17
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

system in Case 1 is not allowed. It is calculated that the annual degra­ although the electricity bill and FiT in Case 3 is more profitable. LCOE of
dation of EVs in group 4 is about 0.66 % considering short-distance Case 3 is about 0.1193 $/kWh, slightly higher than Case 1 by 3.92 % and
cruises. It is shown that EV SOC in Case 2 is obviously more fluctuant Case 2 by 10.83 %, respectively.
with the annual EV degradation rate at about 1.65 %, increased by (5) Decarbonisation potential of the net-zero energy building
149.48 % compared with Case 1. The annual degradation rate of EVs in The annual equivalent time-of-use carbon emissions of the net-zero
Case 3 is about 0.72 %, reduced by about 56.50 % than Case 2, as the energy building supplied by the renewable energy system are calcu­
static battery storage makes complementary energy storage operations lated to identify the decarbonisation potential of three cases as defined
with EVs in Case 3. in Eq. (11) in Section 2.3. It can be found from Fig. 24 that the time-of-
(3) Grid flexibility of the net-zero energy building use equivalent carbon emissions of peak time in all three cases are
Fig. 21 shows that GFF in peak time of the optimum case (Case 3) negative, indicating that grid export from the PV system is more than
achieves the best performance at a value of − 1.92, lower by 1.07 than grid import during peak time under the effective grid-protective energy
Case 1 and 0.97 than Case 2, respectively. It is shown that GFF in flat management strategy. The equivalent carbon emissions of peak time in
time in the optimum case (Case 3) is about 0.25 with the best perfor­ Case 2 are less than Case 1 by about 11.80 % (19.06 tons) via adopting
mance, lower than Case 1 and Case 2 by 0.20 and 0.09, respectively. It is bidirectional power interaction between EVs and the building. The
shown that GFF in valley time of Case 2 and Case 3 is higher than 1 with equivalent carbon emissions of peak time in Case 3 with optimum
the value of 1.08 and 1.01, while GFF in valley time of Case 1 is slightly configurations are the minimum, by about 101.82 % (183.91 tons) less
lower than 1 at 0.91. GFF of the net-zero energy building adds GFF in than that in Case 2. The carbon emissions of flat time in Case 1 and Case
peak time and GFF in flat time and minus GFF in valley time, where a 2 are about 47.42 tons and 49.22 tons respectively, while it is negative at
lower value is preferred. It can be found that GFF of the net-zero energy about − 17.63 tons in Case 3 with a negative net grid import. The carbon
building is Case 3 is about − 2.68 with the best grid integration perfor­ emissions of valley time in all three cases are positive within the range of
mance, reduced by 1.38 and 0.99 than Case 1 and Case 2, respectively. 106.47 tons to 202.44 tons. The total annual equivalent carbon emis­
(4) Economic performance of the net-zero energy building sions of Case 1 and Case 2 are about − 7.67 tons and − 5.29 tons,
The annual electricity bill of three net-zero energy building cases is respectively. The decarbonisation impact in Case 3 is the most
compared including bills of peak time, flat time and valley as shown in outstanding at about − 179.72 tons, much less than that in Case 1 and
Fig. 22. It is indicated that Case 1 has the maximum electricity bill in Case 2, as more PV panels, EVs and batteries are integrated with the
both peak and flat periods with the highest net grid import, resulting in a building in the optimum case. Obvious decarbonisation benefits can be
maximum total annual electricity bill at 28.44 k$. The electricity bill in achieved for three net-zero energy building cases powered by the
Case 2 is about 23.12 k$ adopting EV discharging, reduced by about renewable energy system.
18.70 % than Case 1 with less grid import during peak and flat hours. It
can be also found that much less energy is imported during peak and flat 4. Conclusions
hours in Case 3 via introducing static battery storage, resulting in more
grid import in valley hours instead. The total annual electricity bill of the This study proposes a transient design management and optimization
optimum case (Case 3) is about 16.68 k$, reduced by 41.35 % and 27.86 framework of renewable energy systems to achieve net-zero energy
% than Case 1 and Case 2, respectively. It shows a great potential in buildings integrating EVs and static battery storage. A building load data
reducing electricity bills for building uses with the optimum PV system augmentation model is developed to obtain the annual hourly load
integrating battery and EV storage. profile of a practical campus building based on the on-site collected data
PRV and LCOE of the renewable energy system is evaluated for adopting the Gate Recurrent Unit neural network. A grid-protective
power supply to the net-zero energy building during a 20-year lifetime energy management strategy is presented for the renewable energy
time as per Fig. 23. It is indicated that NPV of Case 1 is about 1109.62 k system proposing novel assessment indicators including the PV utiliza­
$, where the initial cost accounts for about 61.63 % and FiT contributes tion ratio, load match ratio and grid flexibility factor for time-of-use
to about 9.94 %. LCOE of Case 1 is about 0.1148 $/kWh, and it is operations. The multi-objective optimizations are conducted to
reduced by about 6.24 % in Case 2 at 0.1077 $/kWh with reduced explore optimum sizing configurations of the renewable energy system
electricity bill. NPV of Case 3 with an increased PV capacity and battery for balanced tecno-economic performance. Important conclusions of the
storage is about 1729.64 k$, obviously higher than Case 1 and Case 2, present study are outlined as follows:

Fig. 21. Grid flexibility factor of the net-zero energy building.

18
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Fig. 22. Annual electricity bill of the net-zero energy building.

Fig. 23. Lifetime present value and levelized cost of energy of the net-zero energy building.

(1) Integrating more EVs with the net-zero energy building contrib­ contributes to improving the grid flexibility and system economy,
utes to achieving balanced performance on the power utilization while decreasing the PV utilization efficiency. Integrating more
efficiency, grid flexibility and lifecycle economy, with about 89 % static battery storage enables higher grid flexibility and lower
of Pareto optimal solutions close to the maximum EV number. A system economy, and shows weak impact on the PV utilization
good solar PV power utilization efficiency can be achieved with efficiency. Integrating more EVs improves grid flexibility when
the proposed grid-protective energy management strategy, with the PV capacity is relatively large with the annual power gener­
nearly 92 % of the Pareto optimal solutions achieving a PV uti­ ation higher than annual building demand, while it shows
lization ratio at over 90 %. The net-zero energy building achieves negative impact when the PV capacity is relatively small.
the best overall techno-economic performance with the system (3) The vehicle-to-building interaction introducing EV discharge in
configuration of 1050 kW rooftop PV power, 300 EVs and 450 the net-zero energy building (Case 2 vs. Case 1) improves the PV
kWh battery storage. utilization efficiency (+2.60 % on PV utilization ratio), load
(2) The sensitivity analysis identifying the interactive impact of coverage (+12.08 % on load match ratio), time-of-use grid flex­
renewable energy system configurations on techno-economic ibility performance (− 29.63 % on grid flexibility factor) and
performance of the net-zero energy building helps to guide the system economy (− 18.70 % on annual electricity bill and − 6.24
system design and management. Installing more PV panels % on levelized cost of energy). But it shows negative impact on

19
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.

org/10.1016/j.enconman.2023.117768

References

[1] International Energy Agency. Global energy crisis; 2023.

[2] Liu J, Ma T, Wu H, Yang H. Study on optimum energy fuel mix for urban cities
integrated with pumped hydro storage and green vehicles. Appl Energy 2023;331:
120399.
[3] European Commission. REPowerEU: Affordable, Secure and Sustainable Energy for
Europe; 2023.
[4] Senate Democrats. Summary: The Inflation Reduction Act of 2022; 2023.
[5] International Energy Agency. World Energy Balances 2020: Extended Energy
Balances; 2022.
[6] Chinese Ministry of Housing and Urban-Rural Development. 14th Five-Year’
Building Energy Efficiency and Green Building Development Plan; 2022.
Fig. 24. Carbon emissions of the net-zero energy building. [7] Liu J, Cao S, Chen X, Yang H, Peng J. Energy planning of renewable applications in
high-rise residential buildings integrating battery and hydrogen vehicle storage.
Appl Energy 2021;281:116038.
EV healthy (e.g., 1.65 % vs. 0.66 % on EV degradation in group 4 [8] REN21. Renewables Global Status Report 2023 Collection: Energy Demand
and decarbonisation benefits (− 5.29 tons vs. − 7.67 tons, +31.03 Modules; 2023.
% on carbon emissions). [9] REN 21. Renewables 2023 Global Status Report; 2023.
[10] International Energy Agency. GlobalABC roadmap for buildings and construction
(4) The optimum net-zero energy building integrated with hybrid EV 2020-2050: towards a zero-emission, efficient, and resilient buildings and
and battery storage achieves good techno-economic- construction sector; 2020.
environmental performance (Case 3 vs. Case 2) regarding the [11] Zuo J, Zhong Y, Yang Y, Fu C, He X, Bao B, et al. Analysis of carbon emission,
carbon displacement and heterogeneity of Guangdong power industry. Energy Rep
load coverage (+16.22 % on load match ratio), EV healthy (0.72
2022;8:438–50.
% vs. 1.65 % on EV degradation in group 4, time-of-use grid [12] Energy Administration of Guangdong Province of China. Implementation Plan for
flexibility (− 58.48 % on grid flexibility factor), electricity cost Promoting High-quality Energy Development in Guangdong Province; 2023.
[13] Liu J, Wang M, Peng J, Chen X, Cao S, Yang H. Techno-economic design
(− 27.86 % on annual electricity bill) and decarbonisation bene­
optimization of hybrid renewable energy applications for high-rise residential
fits (− 179.72 tons vs. − 5.29 tons, − 34 times on carbon emis­ buildings. Energ Conver Manage 2020;213:112868.
sions). But it results in an increment on the levelized cost of [14] International Energy Agency. Global CO2 emissions from transport by sub-sector in
energy of about 10.83 % (1.1193 $/kWh vs. 0.1077 $/kWh). the Net Zero Scenario, 2000–2030; 2022.
[15] Terkes M, Demirci A, Gokalp E. An evaluation of optimal sized second-life electric
vehicle batteries improving technical, economic, and environmental effects of
The proposed transient design management and optimization hybrid power systems. Energ Conver Manage 2023;291:117272.
framework of the renewable energy system provides a possible pathway [16] Amir J. Thai Government Announces EV Roadmap; 2020.
[17] Mckenzie B. Philippines: The Electric Vehicle Industry Development Act (EVIDA),
for the typical campus building towards zero-carbon operations. The Republic Act No. 11697, Lapses into Law; 2022.
comprehensive techno-economic-environmental feasibility analysis of­ [18] Kelly S, Renshaw J. U.S. EPA proposes revamp of biofuel program to include
fers significant guidance and reference for relevant stakeholders to electric vehicles; 2022.
[19] BloombergNEF. Electric Vehicle Outlook 2023; 2023.
develop carbon–neutral buildings. [20] Xu X, Hu W, Liu W, Du Y, Huang R, Huang Q, et al. Risk management strategy for a
renewable power supply system in commercial buildings considering thermal
CRediT authorship contribution statement comfort and stochastic electric vehicle behaviors. Energ Conver Manage 2021;230:
113831.
[21] Liu J, Chen X, Cao S, Yang H. Overview on hybrid solar photovoltaic-electrical
Jia Liu: Conceptualization, Methodology, Software, Formal analysis, energy storage technologies for power supply to buildings. Energ Conver Manage
Data curation, Visualization, Funding acquisition, Writing – original 2019;187:103–21.
[22] Liu J, Zhou Y, Yang H, Wu H. Net-zero energy management and optimization of
draft, Writing – review & editing. Huijun Wu: Project administration, commercial building sectors with hybrid renewable energy systems integrated with
Funding acquisition, Resources, Writing – review & editing. Huakun energy storage of pumped hydro and hydrogen taxis. Appl Energy 2022;321:
Huang: Software, Data curation, Investigation, Methodology, Valida­ 119312.
[23] Xu D, Zhong F, Bai Z, Wu Z, Yang X, Gao M. Real-time multi-energy demand
tion, Writing – review & editing. Hongxing Yang: Supervision, Inves­ response for high-renewable buildings. Energ Build 2023;281:112764.
tigation, Resources, Writing – review & editing. [24] Lü X, Lu T, Karirinne S, Mäkiranta A, Clements-Croome D. Renewable energy
resources and multi-energy hybrid systems for urban buildings in Nordic climate.
Energ Build 2023;282:112789.
Declaration of Competing Interest [25] Liu J, Chen X, Yang H, Shan K. Hybrid renewable energy applications in zero-
energy buildings and communities integrating battery and hydrogen vehicle
storage. Appl Energy 2021;290:116733.
The authors declare that they have no known competing financial [26] Wang B, Yu X, Xu H, Wu Q, Wang L, Huang R, et al. Scenario analysis,
interests or personal relationships that could have appeared to influence management, and optimization of a new Vehicle-to-Micro-Grid (V2μG) network
based on off-grid renewable building energy systems. Appl Energy 2022;325:
the work reported in this paper. 119873.
[27] Huang P, Tu R, Zhang X, Han M, Sun Y, Hussain SA, et al. Investigation of electric
Data availability vehicle smart charging characteristics on the power regulation performance in
solar powered building communities and battery degradation in Sweden. J Storage
Mater 2022;56:105907.
The authors do not have permission to share data. [28] Khan S, Sudhakar K, Bin Yusof MH. Building integrated photovoltaics powered
electric vehicle charging with energy storage for residential building: design,
simulation, and assessment. J Storage Mater 2023;63:107050.
Acknowledgement [29] Thomas D, Deblecker O, Ioakimidis CS. Optimal operation of an energy
management system for a grid-connected smart building considering
This work was financially supported by the National Natural Science photovoltaics’ uncertainty and stochastic electric vehicles’ driving schedule. Appl
Energy 2018;210:1188–206.
Foundation of China (52208100), the Natural Science Foundation of [30] Alzahrani A, Sajjad K, Hafeez G, Murawwat S, Khan S, Khan FA. Real-time energy
Guangdong Province (2023A1515010725) and the Opening Fund of Key optimization and scheduling of buildings integrated with renewable microgrid.
Laboratory of Building Safety and Energy Efficiency of Ministry of Ed­ Appl Energy 2023;335:120640.
ucation (2022BSEE003).

20
J. Liu et al. Energy Conversion and Management 298 (2023) 117768

[31] Le TS, Nguyen TN, Bui D-K, Ngo TD. Optimal sizing of renewable energy storage: a [42] Xu B, Oudalov A, Ulbig A, Andersson G, Kirschen DS. Modeling of lithium-ion
techno-economic analysis of hydrogen, battery and hybrid systems considering battery degradation for cell life assessment. IEEE Trans Smart Grid 2016;9:
degradation and seasonal storage. Appl Energy 2023;336:120817. 1131–40.
[32] Abdelhady S. Techno-economic study and the optimal hybrid renewable energy [43] Yousri D, Farag HEZ, Zeineldin H, El-Saadany EF. Integrated model for optimal
system design for a hotel building with net zero energy and net zero carbon energy management and demand response of microgrids considering hybrid
emissions. Energ Conver Manage 2023;289:117195. hydrogen-battery storage systems. Energ Conver Manage 2023;280:116809.
[33] Souayfane F, Lima RM, Dahrouj H, Dasari HP, Hoteit I, Knio O. On the behavior of [44] Alam M, Saha T. Cycle-life degradation assessment of Battery Energy Storage
renewable energy systems in buildings of three Saudi cities: winter variabilities and Systems caused by solar PV variability. In: 2016 IEEE power and energy society
extremes are critical. J Build Eng 2023;70:106408. general meeting (PESGM). IEEE; 2016. p. 1–5.
[34] Solar Energy Laboratory Univ. of Wisconsin-Madison. TRNSYS 18 a transient [45] Bengoetxea A, Fernandez M, Perez-Iribarren E, Gonzalez-Pino I, Las-Heras-Casas J,
system simulation program, Volume 8 weather data; 2017. Erkoreka A. Control strategy optimization of a Stirling based residential hybrid
[35] Pan D, Bai Y, Chang M, Wang X, Wang W. The technical and economic potential of system through multi-objective optimization. Energ Conver Manage 2020;208:
urban rooftop photovoltaic systems for power generation in Guangzhou. China 112549.
Energy Build 2022;277:112591. [46] Deb K, Pratap A, Agarwal S, Meyarivan T. A fast and elitist multiobjective genetic
[36] Solar Energy Laboratory Univ. of Wisconsin-Madison. TRNSYS 18 a transient algorithm: NSGA-II. IEEE Trans Evol Comput 2002;6:182–97.
system simulation program, Volume 4 mathematical reference; 2017. [47] Zhang Y. Use jEPlus as an efficient building design optimisation tool. In: CIBSE
[37] Duffie JA, Beckman WA. Solar engineering of thermal processes. John Wiley & ASHRAE technical symposium. Imperial College London; 2012.
Sons; 2013. [48] Zhang Y, Korolija I. Performing complex parametric simulations with jEPlus. In:
[38] Liu J, Zhou Y, Yang H, Wu H. Uncertainty energy planning of net-zero energy SET2010-9th international conference on sustainable energy technologies; 2010. p.
communities with peer-to-peer energy trading and green vehicle storage 24–7.
considering climate changes by 2050 with machine learning methods. Appl Energy [49] Zhang Y, Jankovic L. An interactive optimisation engine for building energy
2022;321:119394. performance simulation; 2017.
[39] Guangzhou Municipal Planning and Natural Resources Bureau. Guangzhou [50] Magnier L, Haghighat F. Multiobjective optimization of building design using
Transport Development Annual Report 2021; 2022. TRNSYS simulations, genetic algorithm, and Artificial Neural Network. Build
[40] Calise F, Cappiello FL, Dentice d’Accadia M, Vicidomini M. A novel smart energy Environ 2010;45:739–46.
network paradigm integrating combined heat and power, photovoltaic and electric [51] Guangdong Provincial Development and Reform Commission. Notice on Improving
vehicles. Energ Conver Manage 2022;260:115599. the Time-of-use Electricity Price Policy in Guangdong Province; 2021.
[41] Matsuishi M, Endo T. Fatigue of metals subjected to varying stress. Jpn Soc Mech [52] The People’s Government of Guangzhou Municipality. Notice on organizing the
Eng, Fukuoka, Japan 1968;68:37–40. application of subsidy funds for 2019 solar photovoltaic power generation projects;
2018.

21