Application Research Of Big Data In Electricity Demand Analysis

In power system, the medium and long-term electricity demand forecast is the premise of many aspects of works. Accurate demand forecasting can arrange the unit start-stop reasonably, plan power grid construction, and improve the economic and social benefits. Traditional medium and long-term power demand forecasting methods are divided into two categories: regression forecast methods based on economic factors and methods based on time series. However, current electricity demand forecast methods considering the factors such as GDP, climate and historical load and the data is limited to the structured data. In recent years, the concept of "big data" has received great attention, and the application of big data in power system become a new research hotspot. However, the existing research mainly involves the distributed computing model and the improvement of the optimization algorithm, and few work refers to the application of more data types in power system. Keywords retrieval data is introduced into electricity demand forecasting which is a typical of big data. In theory, more rich data source can better reflect the influence factors of electricity demand and its inherent law, and then improve the prediction accuracy. The main research contents are as follows:(1) Research the application of traditional data in the medium and long-term power demand forecasting. The residential electricity forecast is made using the classical time series model and historical data as an example. Result shows that the time series prediction model performs well under normal circumstances, but the precision drop seriously in the case of extreme weather. Considering the high temperature factors, the neural network model and actual temperature data are used to correct the above results. Result shows that the prediction accuracy have certain improvement after joining temperature data, but the actual effect is still limited.(2) The possibility of introducing big data into power system demand forecast is studied. Firstly, the difference between big data and traditional data are analyzed. Considering the convenience of data acquisition and use value, retrieval data of keywords on network is the emphasis.(3) Considering the complexity of big data and its inherent characteristics, the scope of application of keywords retrieval data in electricity demand analysis is studied. In some typical businesses, the correlation between keywords and special power usage is illustrated, and then some discussions and application suggestions are put forward.(4) A revised method of power demand forecast based on network keywords retrieval data is proposed. Before that, some pretreatment methods are introduced, such as seasonal adjustment, keywords selection. Actual examples show that this method can effectively improve the demand forecast accuracy, but also verify the application value of network keywords retrieval data in the field of electricity demand forecasting.