Power System Short-Term Load Forecasting Based on Default Rules Mining

Abstract

Short-term load forecasting plays an increasingly important role in the electric network dispatching organization. Here, the default rules mining algorithm is applied to power system short-term load forecasting. First, the conditional attributes such as temperature and humidity that affect load characteristics are discretized by rough set discretization algorithm based on Gini index, and the consideration is given to both conditional attributes and decision-making attributes. On this basis, through computing the confidence and support of rules the network rules set in different levels, which is accompanied with rough set operator and conforms to originally specified threshold, is generated, so the redundant rules brought about by the influence of noise can be reduced, so that the generated classification rules set can be evidently minified and the efficiency of retrieving rules can be improved while the rules are used. During the load forecasting the rules set is searched layer by layer from the top to the bottom until the rules that match with the information are found out. The rough set operator reflects the significance level of the rule, so it is used as the standard to choose rules. Case applications show that the presented method can effectively remove noise and improve the efficiency of default rules mining, therefore the accuracy of load forecasting can be improved.