Short-term forecasting of electricity consumption

Abstract

There are many numbers of large electricity consumers who focuses on improving the accuracy of electricity consumption forecasts because of intensive electricity market development. Wholesale market for electricity and power require consumers to make electricity consumption forecasts at different time intervals. Increasing the accuracy of electricity consumption forecasts can significantly reduce costs. Therefore, this topic is receiving more attention. This dissertation aims to improve the economic efficiency of the enterprise by implementing forecasting system for electricity consumption at the software and hardware levels. The result of this dissertation is day-ahead and a week-ahead forecast of electricity consumption for the Sibelectromotor Enterprise located in Tomsk, Russia. A financial analysis of the implementation of electricity consumption forecasting systems was also prepared. An automatic control system for electric drive using a fuzzy logic controller was developed during the work. The implementation of the system allows high efficiency of the electric drive within a large operational parameter range. Dissertation is structured as follows: first I describe the Wholesale Market for Electricity and Power in Russia. Then the analysis of the enterprise load diagram is performed. The methodology for electricity consumption forecast models is presented next. In the following chapter I forecast day-ahead, week-ahead electricity consumption using Autoregressive Integrated Moving Average, Artificial Neural Networks and Classification and Regression Trees methods. Then economic evaluation of the investment decision is performed. The next chapter describes the automatic control system for an electric drive using fuzzy logic controller. In conclusion, I can say that short-term forecasting of electricity consumption based on ANN and CART models are the best method because the key performance indicators and the cost of purchased electricity are lower.

There are many numbers of large electricity consumers who focuses on improving the accuracy of electricity consumption forecasts because of intensive electricity market development. Wholesale market for electricity and power require consumers to make electricity consumption forecasts at different time intervals. Increasing the accuracy of electricity consumption forecasts can significantly reduce costs. Therefore, this topic is receiving more attention. This dissertation aims to improve the economic efficiency of the enterprise by implementing forecasting system for electricity consumption at the software and hardware levels. The result of this dissertation is day-ahead and a week-ahead forecast of electricity consumption for the Sibelectromotor Enterprise located in Tomsk, Russia. A financial analysis of the implementation of electricity consumption forecasting systems was also prepared. An automatic control system for electric drive using a fuzzy logic controller was developed during the work. The implementation of the system allows high efficiency of the electric drive within a large operational parameter range. Dissertation is structured as follows: first I describe the Wholesale Market for Electricity and Power in Russia. Then the analysis of the enterprise load diagram is performed. The methodology for electricity consumption forecast models is presented next. In the following chapter I forecast day-ahead, week-ahead electricity consumption using Autoregressive Integrated Moving Average, Artificial Neural Networks and Classification and Regression Trees methods. Then economic evaluation of the investment decision is performed. The next chapter describes the automatic control system for an electric drive using fuzzy logic controller. In conclusion, I can say that short-term forecasting of electricity consumption based on ANN and CART models are the best method because the key performance indicators and the cost of purchased electricity are lower.