Introduction of Power Planning (5) {07 FORECASTING MODELLING he Foreca: There are many factors which influence the prediction of load and their influence vary from ‘area to area and from country to country. The impact of any factor on load of a utility needs to be erly examined before building a forecasting model. The factors found to affect a variety of ies’s load are time dependent, weather dependent, random, and other. ‘Time dependent factors : ‘ Power systems exhibit a time dependent pattern of electri are regular, irregular or random in nature. load demand. At ti 1es, these factors Regular pattern is exhibited during the time of day, day of week and week of the year, and yearly growth. ‘+ Irregular pattern is exhibited on holidays, weekends, special days etc. and load requirements tend to differ on these days than on other days. Sometimes, load requirements do not follow any pattern because of weather or other factors © Electric load requirements tend to depend on work rest style of our set up as there can be different possibilities of electric power consumption if people are at home during the day than if they are away at work. This implies that load patterns are different on weekdays and weekends, with the, possibilities of 2-4 groups, namely, weekdays, weekends, pre and post respectively. + Ananalysis of past data can reveal two or more pattern of load consumption of a weak. On the same lines, load consumption also differs on holidays, special holidays, preceding and following the weekends, and special days of national or social importance which may require excessive lighting loads ete. * The impact of these holidays and special days on load demand would depend on the extent of public participation, impact on industrial activity, and state level celebrations requiring excessive lighting load. There are seasonal variations in hourly or daily load, due to change in daylight hours, change in heating to cooling load or vice-versa, typicality of load pattern of some months etc. From the past date (typically 2-5 years) periods in a year can be divided into time scales (hourly, daily etc.) which exhibit an established load curve and others with a comparatively variable load curve. Weather dependent factors : * Weather is one of the principle causes of load variations as it affects domestic load, public lighting, commercial loads etc. Therefore, its is essential to choose relevant weather variable and model their influence on power consumption. Principal weather variables found to affect the power consumption included temperature, cloud cover, visibility, and precipitation. * The first two factors affect the domestic/office (e.g, heating, cooling) loads, whereas the others affect lighting loads as they affect daylight illumination.Power System Planning ¢————————__ Average temperature is considered to be the most significant dependent factor that influen, loads variations. However, temperature and load are not linearly related, and variations ig temperature in one temperature range may not have any effect on the load, whereas in temperature ranges and or other seasons a 1°C change can change load demand by over ong percent. This non-linear relation is further complicated by the influence of humidity and by the effects of extended periods of extreme heat or cold spell. Random factors : * There are random phenomena which affect load consumption and can cause large errors iy load forecast._ «itis difficult to accurately model their actual impact on load demand. These include schoo} holidays, factory strikes, and influence of popular TV programn = ‘Influence of these phenomena can be studied if past data on these occurrences are available, Other factors : Other factors that influence the demand include, (Effects of DSEs (Distributed generating devices) (ii) Effects of rate tariff (time of day pricing, change in industrial tariffs) (ii) Change over to winter time or summer time. Impact of these factors in past data should be identified. The model should be selected based on these factors and other considerations, and should be fitted to the data. Before use, the model should be checked to discover possible lack of fit or any inadequacy, and necessary correction should be applied as required. 17.2 Forecasting M (a) Regression model © This functionally relates load to other economic, competitive or weather or weather variables and estimates an equation using the least squares technique. Relationship are primarily analyzed ~ Statistically, although and relationship should be selected for testing on a rational ground. Regression analysis involves the necessity of using judgment along with statistical analysis whenever forecasting takes place. . * Regression of time series date is a common occurrence in utilities where tracking importatt measures of performance on a weekly, monthly or quarterly basis is conducted. As auto correlation is common problem in such studies, an understanding of this condition and its cur’ becomes vital if the results of such analyses are to be valid in the decision making process. (b) Econometric model *~ Aneconometric models a system of interdependent regression equations that describes eners) + sales. The parameters of the regression equations are usually estimate simultaneously. AS tule, these models are relatively expensive to develop. However, due to the system equations inherent in such models, they wilLbetter express the casualties involved than an ordinary regressios equation and hence, will predict turning points more accurately.To Introduction of Power Planning —_———__@) (c) Strategic forecasting + Strategic forecasting is becoming increasingly important and involves the explicit examinati of the factors and issues affecting future . It recognizes the impact that policy decisions _ can have on future Toads. + Thisrequires details of consumer operations, their current and potential’ demand for electricity, their competitiveness in the market place and their options with respect to production processes. + Switching altematives, energy conservation technologies etc. + Inthe industrial sector, this implies combining elements of the econometric approach with the technology detail found in end use/process models. Strategic models must be capable of doing _ additional information to help shape the future demand. Mathematical modelling simulation : + In modelling, the total load is considered to be the sum total of various components due to various factors. * These factors need to be measured and interrelated with load requirements. Thus, this technique individual modelling of each load type, and identifying their interrelationship to arrive at future Joad requirements. * This is mathematical modelling. Mathematics is a language that allows us to represent physical problems in a form that a computer can understand. + The strength of a method lies in the accuracy of the results it gives. Errors in predicted loads are found mainly in peak periods, transitional phase (from peak to off peak and vice-versa), and on weekends and special days. * Inextrapolation, future load is treated as an extension of the past and the load curve based on past data is suitably adjusted to reflect growth trend. Thus, this technique involves the detection of trends in the past data for various parameters, fitting a trend curve which could be a straight. line, a parabola, exponential or a polynomial of other orders or a mix of the above and finding coefficients of theses curves as given below, Straight line Y=a+ bx Parabola Y=a+ bx + cx S.curve Yra+ bx t+ c+ de Exponential Y=be* Modified exponential Y=a + be* Logistics Y= Wa + be) Where Y is a variable to be fitted, x is time in assigned frame (in day, week, year etc.) and @.),c,d are coefficients to be calculated. Extrapolation could be deterministic or probabilistic, with the accuracy of results quantified using statistics (i.e. standard deviation, variance etc.) in the letter case The mathematical model for domestic, commercial and other sectors have been determined by the CEA as given below.____ _ Introduction of Powwor Planning —_—_—_@ ecurity analysis, wi §) vii) Scheduling of pumped storage units. vii) Lond management scheduling, (ix) Optimization of fuel stocking, Utilities use past normalized data, weather data, and information on known random phenomena Jike popular TV programmes, school vacations, factory strikes etc., for short-term forecasting of, (i) Peak load conditions for system in a day. (ii) System load at various intervals of time (half hour /hour) in a day. (ii) Hourly or half-hourly energy requirements. (iv) Individual bus load prediction. (v) A few minutes to several hours ahead forecast and is useful in utility's systems operations to deal with economic load dispatching & security assessment. 1.6.5 Forecasts Techi The need to understand the proper use of forecasting techniques has increased as the computing capability has moved out on the hands of the experts in to those of the users in an organization. Forecasting continues to gain in importance due to the increasing scarcity of electrical energy along with the availability of lower cost and more powerful computing equipment and softwares. Examples Here techniques used are called Deterministic and Statistical. Deterministic techniques are further classified as extrapolating, econometric, end use and strategic. For example 1. For extrapolation : Sheer's formula is used which is based on the hypothesis that far every one hundredfold increase-In per-capita generation, half will reduce the rate of growth of power generation. The following relation was developed after studying load growth in a number of countries. formula is used iteratively to forecast — consumption growth for each year with the preceding value used to forecast the next year’s growth. 2. Im the end use method : The consumption of each category is projected, based on expected changes in production (industrial), traction, irrigation, water works and sewerage pumping etc. ‘This technique is adopted where sufficient data regarding the programme for future is available.Powor Systom Planning @——————_—— 3. Trend method : Is suitable in case of other sectors such as domestic, commercial and public lighting, For example, an exponented trend using energy consumption data in India the calculated regression equation is shown below: i Y = =3411.39+ 8555.05 x €0.0988% X = time in years with 1950-5 las base year, Y = GWh requirement for the year Trends identified nowadays are: (i) _ Industria to information society (ii) National to world economy (ii) Short-term to long-term thinking (iv) Centralization to decentralization (v)_ Either/or to multiple options Time series analysis : Is a good technique involving the necessity of using sound judgments along with an analysis of past history. The history of past loads might be forecasted by a utility using a time series analysis program, which used monthly data and yields an analysis of trend, cyclical variation, seasonal variation, and irregular movement. A recomposition of these four factors in to future months would involve considerable judgementas to the future course of the cyclical and irregular elements and, if these elements were well formulated, would produce usable forecast of electrical energy demand. 5. _Moving average : Here each point of a moving average of a time series is the arithmetic or ‘weighted average of a number of consecutive points of the series, where the number of data points is chosen so that the effects of season or irregularity of both are eliminated, A minimum. of two years of past energy consumption is desirable, if seasonal effects are present. Otherwise, k there will be less data. (Of course, more the history, the better). The moving average must be | specified. 6. Trend projections : This technique fits a trend line to a mathematical equation and then projects it into the future by means of this equation. There are several variations, the slope } characteristic method, polynomials, logarithms, and so on. Trend analysis is the study of the behavior of a process in the past and its mathematics modeling so. Future behavior can be extrapolated from it. Two general approaches followed for trend analysis are: (i) The fitting of continuous mathematical functions through actual data to achieve the least overall error, known as regression analysis. (ii) The fitting of a sequence of discontinuous lines or curves to the data. The second approach in the short term forecasting. A time varying event such as power system load can be broken down into the following four major components.