38

Energy Engineering

Vol. 107, No. 4

2010

PLC-based Load Management in Steel Rolling Mills

Ashok S, Senior Member IEEE ABSTRACT This article presents a physically-based load model and formulation for continuous process industries to use in implementing industrial load management. The formulation utilizes an integer programming technique for minimizing electricity costs by scheduling the loads and satisfying the process, production, and maximum demand constraints. The case study of a typical steel rolling mill, with the proposed model, shows that savings of about 5.21% in electricity costs are possible with optimal load scheduling under TOU tariff. After scheduling the loads, real time implementation of load management action is investigated using a programmable logic controller. Keywords: Load management actions, load flow control, steel mill, timeof-use tariff (TOU) INTRODUCTION Demand for electricity has increased with the advancement of technology and living standards. Most of the electric utilities throughout the world are facing difficulties in meeting the increasing demand from different consumer sectors, at all times. For example, the power system of India is experiencing an energy shortage of 9%, with a peak demand deficit of about 15.2% [1]. As the demand is widely varying, the utility must run generation units that are sufficiently rated to meet the demand. Especially during peak periods, the utility has to increase generation capacity and operate costly peak-generating units. Initiatives are usually introduced by the utilities to smooth the system load curve and thus delay (or avoid) the installation of extra capacity. Typically, customers

39

are made aware of this either by means of price signal or through load shedding. Electricity costs can be minimized by taking advantage of incentives and favorable pricing offered by utilities in order to encourage consumers to use energy in such a way and at suitable times so as to enable the utility to manage load patterns. By making the best use of these incentives, it is possible to achieve significant savings in production costs, with no adverse effect on product quality and productivity. The set of options available for load management in industry includes process rescheduling or load shifting, machinery interruption/ restart cycle, energy storage, captive power, and automation. The choice of each option must be weighed against the rate system or financial agreement in effect and the technological constraints posed by the production process. The industrial sector consumes about 41% of the total electrical energy generated on a worldwide basis [2]. Since industries consume a significant proportion of the total electrical energy generated, load management in the industrial sector assumes an important role in peak demand management. Load shifting, one of the simplest methods of load management, is to reduce customer demand during the peak period by shifting the use of appliances and equipment to partial-peak and offpeak periods. No loads are switched off but only shifted or rescheduled; hence, the total production is not affected [3]. Load management (LM) is a specific method of controlling the peak load in the network in order to produce a constant demand. For applying LM techniques to the industrial sector, a detailed modeling and optimization of the industrial loads is needed, including the complexities and constraints of the process. Industrial load management (ILM) activities are aimed at the economic reduction of an electric utilitys demand during peak hours without affecting the specified production. ILM applications have been reported for utilities using interruptible load control schemes [4]. For iron and steel industries, electricity comprises about 30% of total production costs. A mathematical model formulation coupled with an optimization framework for optimal process schedule of the steel plant, for a specified tariff, minimizing the total operating cost and satisfying production, process flow, and storage constraints has been reported[5]. A methodology for collecting end-use demand data for devising demand-side management programs in the commercial sector has

40

Energy Engineering

Vol. 107, No. 4

2010

been reported [6]. As the model developed is aimed at peak-demand reduction of commercial sector sites like hotels and does not cover the intricacies of industrial loads, it is inadequate for ILM applications. A methodology developed to achieve peak utility load reduction in batch processes of an industrial brewery has been reported [7]. However, the optimization model which was developed for batch process, cannot be directly applied to continuous process industries. An optimization formulation using mixed integer programming for load side demand control has been reported [8]. As it does not include storage and process constraints, it cannot be directly applied to process industries. A general approach to solve the optimal contracting capacity for a petrochemical plant with an in-house cogeneration system has been discussed [9]. A methodology utilizing one- and two-way management systems for load control and distribution automation has been discussed [10]; it is reported for the entire distribution system of a particular area. Many methods of load scheduling during peak hours of operation have been utilized for efficient use of generation during peak periods. Some commercial customers in the load management program having refrigeration units controlled by PLC are reported [11]. As a continuation of the physical models and methodologies proposed in the literature, a load model with an optimization formulation for load scheduling is proposed in this article. The model can be applied to determine the optimal operating strategies for industries of continuous-type processes. The formulation utilizes an integer linear programming technique [16] and considers the process, production, process sequence, and maximum demand constraints. Hence, it can be extended to any type of continuous process industry, as the model proposed is a generalized one. Real-time implementation methodology is investigated in this article. MATHEMATICAL MODEL The formulation is based on discrete time representation for continuous process loads. For the planning horizon, one day is split into N intervals with equal time durations of t hours. The decision variable I indicates whether the equipment is either ON or OFF in a particular interval. The decision variable

41

I mjk = )

1, if m th equipment processing j th product in interval k 0, otherwise (1)

Electric power input in kW to equipment m in any interval k when it is processing jth product is W mjk = "^ A m * U mjkh/mjk , (2)

where Am, is the rated capacity of the equipment in kW, Umjk, is the utilization of the mth equipment in the interval k when it is processing the jth product, and mjk is the efficiency of the equipment m in the interval k. The energy consumed in kWh by the equipment m in the kth interval when processing the product j is E mjk = / W mjk * I mjk * t
k=1 N

(3)

The objective function minimizing the monthly operating cost is

e Min / / / 6 E mjk * C k @ * H + C d * MD o
k=1 m=1 j=1

(4)

Where, Ck = H = Cd = M = J =

Cost of energy (charge per kWh) for the interval Number of working days in the month; MD charge (charge/kVA/month), Particular item of equipment; Total number of products

Production constraint to keep the total production Qj of a product j in the planning horizon is
k=1

/ Pmjk * I mjk Q j, 6 m = M

(5)

where Pmjk is production (discharge) in quantity for the machine m for the product j in kth interval.

42

Energy Engineering

Vol. 107, No. 4

2010

Total production of the plant QT is / Q j QT

J

j=1

(6)

Availability of raw material for production is ensured by / / R mjk * I mjk R T, 6m = 1

J N

j=1 k=1

(7)

where Rmjk is the quantity of raw material required for the mth equipment for the product j in the kth interval, and RT is the total raw-material available for the day. Maximum demand constraint to avoid penalty from a utility when processing a product in any interval is ensured by
M

m=1

/ S mjk * I mjk MD k

8)

where Smjk is the demand due to mth machine when processing jth product in kth interval, and MDk, is maximum demand imposed by the utility to industry. Until one product is completed, the same machine should not be allocated to any other product in a particular interval. In order to prevent allocation clashes, the following constraint is included. / I mjk = 1
J

j=1

(9)

Process sequence constraint in between machines is ensured for a specified product in an interval by I mjk I (m + 1)jk 6m > M.

(10)

43

ALGORITHM Steps in the load-shifting algorithm are as follows: 1. 2. 3. 4. Read the load data at different time instants; contract maximumdemand (CMD) or any preferred maximum-demand limit (PMD) at TOU rates. Calculate the maximum demand (MD) at each time instant and cost of electricity for the day. Shift the loads or group of loads to offpeak periods from peak periods. Check for constraints: i. Process sequence constraint ii. Allocation clash constraint (machine can process only one process at a time) iii. Maximum demand constraint iv. Production constraint v. Raw Material constraint, etc. Do steps 3-4 for all possible loads and for all the time instants in a repetitive manner such that (cost of electrical energy for the day) is a minimum.

5.

The above algorithm has to be implemented in any optimization software package for the search of an optimized schedule; the flow chart of the above algorithm is shown in Figure 1. CASE STUDY The industrial units are broadly divided into two categories, namely continuous process plants and industrial batch-type manufacturing plants [12]. Rolling mills comes under the first class, where the ingot rods are the raw material to the mill and the output is steel rods with different diameters. The entire process will run continuously, with full-time raw material feeding and continuous collection of the steel rods at the output section. Plant Description The plant [13] has an installed capacity of producing 6000 MT of

44

Energy Engineering

Vol. 107, No. 4

2010

steel bars in a month, operating in two shifts. Present monthly average production is 5000 MT (approximately 84% of installed capacity) with different types of products, the bars varying in diameters of 8mm, 10mm, 12mm, 16mm, 20mm, and 25mm. The plant has a contract demand of 1500 KVA, with daily energy consumption of 15MWh. Specific energy consumption varies, depending on the products processed. The average specific energy consumption of the plant is 150kWh/MT. Process Flow Plant loads have the nature of continuous process. A process flow diagram of the plant operation is shown in Figure 2. Major Plant Equipment Mill loads mainly consist of motors; the load on these motors is calculated by considering the utilization (U) and efficiency of operation (). Small loads in the mills are grouped as a single load. Based on the observed utilization of loads, efficiency Figure 1. Flowchart parameters are considered as per the standard manufacturers characteristics. The major equipment details provided below are generalized for all products, even those with different rates of consumption in the actual case; also, the production

45

Figure 2. Process Flow Diagram

rate specified is generalized for all, although there are actually different production rates for different products. Specifications of the major equipment are shown in Table 1.

46

Energy Engineering

Vol. 107, No. 4

2010

Table 1. Major Equipment Details

Load In Got Pusher(L1) Blower(L2) Roller Tables(L3) Roller Mill 1(L4) End Cutters(L5) Roller Mill 2(L6) Roller Mill 3 (L7) Dc Drives(L8) Pinch Rolls (L9) Shear(L10) Tail Brake Drives(L11) Twin Channels(L12) Rake Drive(L13) Roller conveyor(L14) Align Motors 1(L15) Cold Shear(L16) Cooling Pumps (L17)* Mill Water Pumps(L18)*

Am 22.2 37 33.3 750 25.9 600 375 300 30 110 44 11 60 14.8 22 18.5 381.3 51.8

U .36 .54 .53 .70 .46 .56 .26 .80 .53 .70 .63 .80 .73 .54 .54 .64 .52 .50

E 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80 0.80

*For L17 and L18 the load varies depending on product. *Loading rate on each machine is 10 T/hr.

Production Requirement For the case study, the production requirement for an average day is about 180MT, consisting of all products (presently producing one product daily), and the total monthly production is about 5000 MT, including all the products. Plant monthly production of each product is listed in Table 2 for typical monthly data. After conducting the load management study, it is observed that instead of producing one product for an entire day, industry can produce many products on the same day, with the less energy-consuming products being processed in the peak hours such that the monthly electricity bill is reduced. It is observed that production can be scheduled in the following manner for a typical month.

47

21 2 2 2

days days days days

8mm, 20mm, 25mm 10mm 12mm 16mm

The industry runs 27 days a month for production, leaving 3 days for maintenance. The resulting monthly production is given in Table 2. Here the production of the 8mm and 25mm is drastically increased; however, these products are in high market demand. The optimal product scheduling for 8mm, 20mm, and 25mm products processed by critical loads, like rolling mills and DC drives, is shown for a typical day in Table 3.
Table 2. Production Details

SECTION 8 mm

Production in MT (Before Scheduling) 2944.4 381.0 421.0 457.5 799.0 90.0

Production in MT (After Scheduling) 3035.34 428.64 115.85 460.8 424.8

10 mm 12 mm 16 mm 20 mm 25 mm Total

*Data shown represent product number *Products 1, 5, 6 are treated as 8mm, 20mm, and 25mm respectively.

5092.9

1161.93 5627.4

Presently the plant is running under two shifts, due to utility restrictions. For the high tension industrial consumers, the utility (Kerala State Electricity Board-KSEB) [14] follows a differential pricing system for both energy and maximum demand, as given in Table 4. Tariff 1 is the prevailing tariff for the industry. The utility has already included the furnace loads in a power- intensive group and will charge as per the power-intensive tariff (Tariff 2) in the future.

48

Energy Engineering

Vol. 107, No. 4

2010

Table 3. Optimal Production Schedule for Typical Day

Machines Intervals 1 4 8 12 16 20 24 30 32 34 41 45

L4 1 1 1 1 1 1 1 6 1 6 5 1

L6 1 1 1 1 1 1 1 6 1 6 5 1

L7 1 1 1 1 1 1 1 6 1 6 5 1

L8 1 1 1 1 1 1 1 6 1 6 5 1

*Only selected loads and selected intervals are listed.

Tariff Rs/KVA Tariff 1 Tariff 2

Table 4. Tariff rates for load management study

MD Charge Rs/kWh 270 270

Energy charge rate 3.00 3.00

Differential 1:1.8:.75 1:3.6:.75

*Energy differential rate: normal, peak, off-peak *TOU Time partition: 6am-6pm: normal, 6pm-10pm: peak; 10pm-6am: off-peak ILM RESPONSE OF INDUSTRY The optimization model as per Eq. 4 is developed based on the equipment and process data. The corresponding integer programming [15] formulation consisting of 5184 decision variables and 6057 constraints is solved using Unlimited LINGO [16]. Results of load scheduling operation under two different energy tariffs, Tariff 1 and Tariff 2, are shown in Table 5. It can be seen that

49

under the prevailing tariff (Tariff 1), the load rescheduling or production rescheduling will result in an annual saving of Rs.9.92 lacks (2.65 %) in the electricity cost. The response to load scheduling operation under the power-intensive tariff (Tariff 2) is more encouraging. Monthly electricity cost gets reduced from Rs.38.94 lacks to Rs.37.01 lacks. It results in an annual saving of Rs.23.16 lacks (5.21%).
Table 5. Comparison of Load scheduling under different tariffs Description Existing Operation Load Scheduling Operation Tariff 1 Tariff 2 Tariff 1 Tariff 2 Peak Demand MDpeak (KVA) 1500 1500 1500 1500 Electricity Charge (Rs. Lacks./Month) 31.18 38.94 30.36 37.01 Annual Saving (Rs. Lacks) 9.92 23.16 Annual Saving (%) 2.65 5.21

After scheduling, the optimal response load curve is compared with the existing 8mm load curve, shown in Figure 3.

Figure 3. Comparison of load curve before and after scheduling with 8mm load

50

Energy Engineering

Vol. 107, No. 4

2010

The above model can be directly applied to other power- intensive continuous process industries with controllable loads. In some cases peak demand reduction is also possible; in such cases if the process industries reschedule their process according to the optimal schedules developed, the utility can achieve a significant reduction in peak demand. IMPLEMENTATION USING PLC A PLC (programmable logic controller) is the most obvious choice for the industrial control application. Eighteen machines need to be controlled, so PLC is a viable option in the industrial environment. The above savings can be achieved in real time by PLCs, and most of the industries already have them for their processes to run smoothly. So it is easy to implement industrial controls by using PLC; also, a graphical user interface can be developed with less investment cost. The optimized scheduling of the machines is obtained by solving the industry model in Unlimited LINGO Optimization Software. LINGO functionality can be used in any Microsoft Windows-based applications by calling up the DLL file. The optimization model developed in LINGO is interfaced with a graphical user interface (GUI) in Visual Basic. LINGO Solver is used as a back-end solver for optimization of the problem. The optimized results from LINGO are called to visual basic by calling LINGO Dynamic Linked Library (DLL). Scheduling of the machines and products can be represented in GUI in proper format using labels and a button in the visual basic. Changes of machine state are represented by changing the color of labels; here a machine can be either on or off. In Visual Basic the change of color is transferred to any other Microsoft Windows-based application by using dynamic data exchange (DDE). This feature is used to generate control signals for machines through a PLC. Allen Bradley PLC Programming Language supports the DDE; the control signals generated from the LINGO are given to PLC by activating an output coil in PLC. The corresponding output coil can easily control the machine by connecting these control signals to a relay and switching control panel of the machine. Control signals are generated according to the optimized schedule. Control signals given to PLC output coils are shown in Figure 4, where only three output coils are shown.

51

Figure 4. Control signals generated in PLC

This methodology is implemented in the Allen-Bradley SLC 5/03 Series C Programmable Logic Controller (PLC), using 18 output and input coils. Here the main interlink communication between PLC and the Visual Basic is through a Dynamic Data Exchange (DDE) server, available in RSLinx Classic. APPLICATIONS This article represents a physically-based load model and formulation that could be applied to any continuous process industry for implementing industrial load management by a load shifting technique for electricity cost minimization. The optimization results are represented in GUI in proper format in Visual Basic, which provides a user friendly interface in an industrial environment. In the increasingly automated industrial sector, real time implementation of load management actions by using Programmable Logic Controllers, which is investigated in this article, is also very important. CONCLUSION An optimization formulation based on load models incorporating equipment and constraints on process, production, raw materials, and

52

Energy Engineering

Vol. 107, No. 4

2010

maximum demand has been developed for LM in continuous process industries. The model represents continuous-type loads and is capable of determining the industry response under different tariff structure. Most of the industries in India are unaware of the potential of load scheduling, with the time varying tariff structure, in minimizing their electricity bill. The case study for a typical rolling mill shows that a reduction of total electricity cost is possible by optimal process scheduling. The optimal schedule under the TOU tariff results in a 5.21% saving in electricity bills as compared to present working conditions. Under TOU tariff, an optimal load schedule can also result in peak savings. The utility can achieve a significant peak demand reduction if industries reschedule their process in response to the TOU tariff. The optimization tool developed in this article facilitates this by helping industries to determine the optimal response, and proposed methodology helps to implement load management actions in real time. Investigative studies are conducted for the possibility of developing an industrial load management controller. ACKNOWLEDGEMENTS The authors would like to thank M/s Peekay Rolling Mills, Calicut, for providing the plant data. References
[1] [2] [3] [4] [5] [6] [7] [8] [9] Ministry of Power, Government of India, Feb-2008, AnnualRep.[Online]. Available:http://www.powermin.nic.in. C.A. Babu and S. Ashok Peak Load Management in Electrolytic Process industries IEEE Transactions on Power Syst. vol. 23, no. 2, 399-405, May 2008. Ashok, S and R. Banerjee, Industrial Load Management, Int. Journal on Applied Energy, Vol. 66 (2),2000, pp105-111. C.S. Chen and J.T. Leu, Interruptible load control for Taiwan power company, IEEE Trans. Power Syst., vol. 5, no. 2, pp. 460465, May 1990. S. Ashok Peak-load management in steel plants Int. Journal on Applied Energy, Vol.83, 2006 pp413424. U. Atikol, A demand-side planning approach for the commercial sector of developing countries, Energy, vol. 29, pp. 257266, Feb. 2004. D. Mignon and J. Hermia, Peak utility load reduction in batch processes operated periodically and under uncertainty, Comput. Chem. Eng., vol. 20, pp. 249263, Mar. 1996. Z. Iou, R. Kumar, J. Sottle, and J.C. Yingling, An MILP formulation for load side demand control, Elect. Mach. Power Syst., vol. 26, pp. 935949, 1998. T.-Y. Wu, S.-S. Shieh, S.-S. Jang, and C.C.L. Liu, Optimal energy management

53 integration for a petrochemical plant under considerations of uncertain power supplies, IEEE Trans. Power Syst., vol. 20, no. 3, pp. 14311439, Aug. 2005. Jack F. Morris, Frank J. Kern, Earl F. Richards Distribution Automation for the Association of Missouri Electric Cooperatives-A Statewide Evaluation of Load Management IEEE Trans. on Industry Applications, Vol. 24, No. 5, pp.781-791, September/October 1988. Robert Roman, Robert Wilson Commercial Demand Side Management Using A Programmable Logic Controller IEEE Trans. Power Syst., vol. 10, no. 1, pp. 376379, 1995. S. Ashok and R. Banerjee, An optimization mode for industrial load management, IEEE Trans. Power Syst., vol. 16, no. 4, pp. 879884, Nov. 2001. Data Collected from M/s Peekay Rolling Mills, Calicut, India. Kerala State Electricity Board (KSEB), Extra High Tension Tariff Trivandrum, India, 2006. Frederick S. Hillier and Gerald J. Lieberman, Introduction to Operations Research, Industrial Engineering Series, McGraw-Hill International Edition, 1995 LINDO Systems, Inc., Hyper LINGO (Release 9), LINDO, Chicago

[10]

[11] [12] [13] [14] [15] [16]

ABOUT THE AUTHOR S. Ashok received M.Tech and Ph.D. degrees from the Indian Institute of Technology (IIT) Delhi, New Delhi, India, and the IIT Bombay, Mumbai, India, respectively. His research interests are peak demand management, energy modeling, captive power, power system protection, and hybrid energy systems. Dr. Ashok is a senior member of the IEEE power engineering society. He is an assistant professor at NIT Calicut. Email: [email protected]