Developing a mathematical model to predict the daily demand for electricity, based on weather parameters

Abstract

System Control Centre (SCC) of Ceylon Electricity Board (CEB) conducts short term (hour ahead, day ahead) and medium term (up to three years ahead) demand forecasting based on historic demand, seasonal patterns, time of day and regional sales forecast. However, there are no measures taken as yet to include the influence of weather conditions such as temperature, humidity, sky cover, wind speed, etc. in this forecasting exercise. Ambient temperature and humidity has become dominant parameters for electricity demand with the introduction of space cooling methods in recent history. The study focuses not only the influence of temperature and humidity to electricity demand of Sri Lanka but also the influence of wind speed and wind direction. Study focused to build up a linear model using hourly historical demand data and meteorological data of four consecutive years using IBM SPSS statistics V 21 software. Meteorological parameters were taken as the independent parameters and hourly demand data was taken as the dependent parameter. Correction factor was needed to include the effect of yearly demand growth, for a better correlation. Every demand data point was corrected based on the average demand growth (yearly) and time of day. Weekdays were taken as one set and Saturday and Sunday were taken separately. Model consists of 72 independent equations (24 representing a weekday, 24 for Saturday and 24 for Sunday). Correction factors were calculated for calendar holidays which have major influence on electricity demand. Model validation was done for historical weather data as well as forecasted weather data. Predicting average absolute error was under 9% decreasing more with the prediction date close to the real date. Model is recommended to be used for short term demand forecasting and power plant dispatching in Sri Lanka.