Research On The Application Of Short-term Load Forecasting For Electricity Sales Company

With the further intensified reform of China power system,the mechanism of power market is becoming more and more mature.For the emerging electricity sales companies,while they are enjoying market dividends with profits exceeding trillions,they are also facing huge challenges.With the operation of electricity spot market,electricity sales companies are no longer restricted only by the monthly assessment of the total power deviation,but they need to accurately control power load forecast at every trading moment to avoid causing fluctuations in real-time electricity price and thus a huge economic loss.Although scholars at home and abroad have made great achievements in the research of short-term load forecasting,the emergence of electricity sales companies has brought new problems.Current domestic electricity sellers can only provide services for industrial and commercial users,this leads to the problem of small load volume base and large load fluctuations of electricity sales companies.Therefore,according to the big difference in load forecasts between electricity sales companies and the traditional regional power grid,the paper proposes a complete set of short-term load forecasting methods for electricity sales companies with the consideration of their characteristics.The main work is as follows:First,the paper analyzes the load characteristics of the electricity sales companies during general days,general holidays and the Spring Festival respectively.By explaining the differences in the load characteristics of each time segment,the necessity of separately modeling and predicting these three stages is demonstrated.Through the analysis of the load characteristics of each stage,the different influencing factors that need to be considered in these three prediction models are found.Secondly,given the problem that the electricity sales companies have small user base and large difference in electricity consumption,classification of customer electricity consumption patterns based on spectral clustering is proposed.And on this basis,a short-term load forecasting model of general days based on the robust extreme learning machine is established,and categorized users are predicted separately in which obtain a higher prediction accuracy.Next,for the situation that the periods of power valley of electricity sales companies appear on normal holidays,the method of same holiday ratio replacement is used in load forecasting.Finally,considering the particularity of the electricity consumption of Chinese industrial users during the Spring Festival,this paper proposes a short-term load forecasting model of Spring Festival with the consideration of the date of commencement and closure.The experimental simulations are based on the actual data from an electricity sales company in Guangdong Province,and the prediction methods proposed in this paper have been verified in terms of effectiveness and accuracy.