Short Term Load Forecasting using Artificial Neural Network and Time Series Methods: A Case Study of Bishnumati Feeder in Balaju Substation, Nepal

Abstract

Electrical load forecasting takes a crucial role in planning, controlling and operation of electric power system. The accuracy of actual and forecasted load be very essential for economically effective control and efficient operation. Short term electric load forecasting takes an importance role in power system operation and planning of Utilities Company. Short term load forecasting (STLF) be always one of most critical, sensitive and accuracy demanding factors of the power systems. An accurate short term load forecasting improves not only the systems economic viability but also its safety, stability and reliability. The researcher presented in this works support Artificial Network and Time Series Methods techniques in forecasting of short term load. This analysis represents a load forecast using hourly, daily and weekly load of Bishnumati Feeder of Balaju Substation, by using artificial neural network & time series methods. Load forecasting be very important and crucial for efficient and necessary operation for electric power system. This can be taken by obtaining the most effective forecast which help in minimizing the risk in decision making and changes the prices of operation of the electric power system. Then the comparison of artificial neural network and time series method for short term load forecasting that can be used out in this thesis using actual time electric power in this feeder. Moving average, exponential smoothing are analyzed in excel and ANN are analyzed in MATLAB software toolbox. The analysis to be done for the hour-to-hour operation to day to day of the soon be completed of the Bishnumati feeder. The ANN and time series methods, different analyzing models were used for forecasting. For validation of this model it can be used to forecast of an 11kV Bishnumati Feeder of Nepal for November 10, 2018 and the output is compared to Conveaant University of Nigeria such as Multilayer feed forward Model. The proposed model of ANN for Bishnumati Feeder gives weeekly MAPE of 3.67% which is lower than the MAPE value obtained by the ANN models in Conveant Univesity of 8.37%, but it is within the limit of acceptance. Again, the value of %MAPE for moving average and exponential smoothing be 9% & 6.22% in Bishnumati Feeder which is lower than the MAPE value obtained by the models of Conveant University of 10.3% & 8.31%% but it is within the limit of acceptance.