Data Mining For Short-term Load Forecasting

The characteristics of short-term load forecasting is obviously subject to the influence of various factors, such as weather changes, holidays, type of days, major social activities and emergencies. The difficulty of short-term load forecasting is that we not only need to consider the characteristics of the load itself as a time series, also need to consider the influence of various non-load factors.Power system load historical data is complex and huge. The characteristics of the data are limited, incompleteness and complexity of influencing factors. Therefore its processing and data mining are necessary. This paper processes the historical data by data cleaning, data integration and conversion, data reduction. In this way it not only ensures the correctness and validity of data, but also makes the data format more suitable for load forecasting.Considering the impact of the non-load factors on short-term load forecasting, the paper uses data mining methods for data preprocessing and uses the decision tree C4.5algorithm to establish a short-term load forecasting model, and then uses real instances to validation and analysis. The forecasting results can meet even exceed the requirements of utility and demonstrate high-accuracy of the proposed model.