Development of time series based consumption forecasting models for determining profile coefficients in electrical energy distribution systems

The load profile coefficients represent the daily and annual electricity usage pattern for electrical energy consumers. It has become important to determine the load profile coefficients accurately and reliably in order to minimize the imbalance costs in the Electricity Energy Market. Reliable methods and sufficient measurement data are required to make accurate estimates. Load profile coefficients are determined on a daily, monthly, seasonal or annual basis. While trying to estimate the amount of consumption in our country, distribution companies try to evaluate the consumption of subscriber groups such as industry, residence, business, agricultural irrigation on an hourly basis. Since residential subscribers are billed monthly, consumption values are also determined at the end of the month. For this reason, there are difficulties in the hourly evaluation of profile coefficients. In addition, meteorological data such as temperature and humidity are not taken into account in the calculation of these profile coefficients to be used in the consumption estimation. Meteorological data is one of the important variables known to affect consumption. It is thought that the fact that meteorological data are not taken into account in the calculation of the profile coefficients will reduce the reliability of the estimation process. In this thesis, mathematical estimation models were developed using multiple polynomial regression analysis method to determine the load profile coefficients for Düzce, Turkey. First, hourly electrical energy consumption and meteorological temperatures were measured in 50 different residential subscribers, which were determined as a sample beforehand. Using the measured data, mathematical prediction models were produced and then load profile coefficients were determined using these mathematical prediction models. Electrical energy consumptions were estimated using the determined load profile coefficients and the estimation results were compared with the measurement data. Mathematical estimation models have been subjected to estimation error tests accepted in the literature and the performances of the models have been verified. According to the results obtained, it has been seen that mathematical forecasting models can predict loads with an accuracy of up to 96% depending on the changing meteorological conditions in the future and it has been suggested as a fast and practical method for calculating load profile coefficients. The study shows that the mathematical prediction models produced have obtained satisfactory results for the determination of load profile coefficients in Düzce, Turkey.