Research On Power System Short-term Load Forecasting Approaches

Short-term electric load forecasting is vital for power generation and operation. It has been becoming an important research field of the power system automation. However, the power system is a system with fixed time-variable parameters and lots of dynamic features. With more complicate of the system, the traditional forecasting methods can no more satisfy the demand of the power department, so the high accuracy and high intelligence forecasting system must be researched.At, first this paper discusses the direction of short-term load forecasting and analyzes and compares the merits and shortcomings of some forecasting methods. After discussing the constituents and characteristics of the electric load, ARIMA (Auto-Regressive Integrated Moving-Average)model is presented for load forecasting.There is a load model with time-varying coefficient. Moreover, Kalman filter is used to estimate the load model parameters. An intelligent choice of the priori estimate of the state initial value and its covariance error initial value enhances the convergence characteristics of Kalman filter. Few samples of the past load can be used to get a least-squares as an initial values.When the model is developed on the basis of an erroneous model and information of the process and measurement noise are either unknown or are known only approximately in practical situations, the filter can "learn the wrong state too well". So the adaptive estimation of noise covariances is conducted by time-varyingnoise estimator and adaptive fading Kalman filter(AFKF) is proposed to solve the divergence problem of Kalman filter.The relative error analysis of the short-term load forecasting results of actual power network by the proposed method shows that the proposed method is effective and predominant.