A hybrid approach of ultra-short term multinode load forecasting

In the electric power system, grasping active and reactive load variation regularity of each node is a key for optimal dispatch, preventive control, security assessment, and transmission capacity evaluation, etc. This paper presents an approach of ultra-short forecasting for multi-node active and reactive load on the basis of former researches. It proposes a hierarchy and subarea idea to construct a framework of self- adapting dynamic model. With the top layer load forecasting implemented by recursive least square support vector machines (RLS-SVM) algorithm, discrete state-space equations are established to describe dynamic characteristics of multi-node load parameters (active load distribution factors and power factors), and then TS fuzzy control technique is introduced to realize feedback compensation. The application in an actual power system control center of Shandong province has been verified with satisfactory result.