Research On Short-term Electric Load Forecasting Based On BP Neural Network

The electric power industry is an important basic industry in the relationship between the economic development and the people’s livelihood. Short term load forecasting of the power system is an important part of network management system. It is the premise and basis of planning the network structure, marketing, trading, making the scheduling and trading plan. The accuracy of the forecast is directly related to the power network security, stability and economic operation. Along with the unceasing development of the social economy, people have gradually increased demand on the quality of life, as a result, the application of cooling and heating equipment are having more and more impact. So the meteorology is increasingly becoming the concern of people in the forecast. In this paper, the problem of short-term load forecasting is studied by taking the load of Shenzhen as an example.The sum of the annual cycle, week periodicity, daily periodicity and holiday characteristics of the load is summarized. The load can be predicted reasonably based on these characteristics. On the classification of power load forecasting and impact load change factors, load forecasting steps and error analysis were studied and the Shenzhen load characteristics are analyzed.BP neural network load forecasting model based on the correlation factor of day characteristic is established. The standard BP algorithm and three improved BP(adaptive BP algorithm, the elastic gradient descent method and L-M method) are respectively used in Shenzhen City load forecasting. By comparing the analysis, the L-M method is proved to be the fastest and the highest prediction accuracy of all algorithms.The BP neural network prediction model without the consideration of the daily feature correlation factor is also established. The L-M algorithm is used to predict the load, and compared with the previous forecasting results, the accuracy of the results is poor. The forecasting model of the correlation factor of the daily characteristics was used to predict the summer load of Shenzhen in July. According to the analysis, the prediction accuracy is low. In order to solve this problem, considering the real-time weather factor, the forecast model is built, and the precision is greatly improved.