Research For Power System Bus Load Forecasting

Load forecasting is an essential work of power market. Administer such as programming and dispatching consist of short-term load forecasting, which is very important because it establish foundation of scientific management and guarantee function of power system. So short-term forecasting is a necessarily part of energy management and distribution management system. Power supply varies in different areas, as demands particular load management. The paper discusses bus load forecasting.The study is achieved by analysesing data of the substation in chengdu. The constitutes, characteristic of bus load and difficulties of forecasting by way of comparing characteristic of bus load with system load, support bus load forecasting resemble the ways of short-team load forecasting. On the basis of it, the detailed process of building model of time series approach is presented Auto-Regressive Iterated Moving-Average (ARM) is applied to bus load forecasting of the substatioa Results show the model practicality and point to the problems at one time.In roost cases, there is a relation between bus load and weather factors such as temperature. These factors are not considered in ARIMA, so the paper provide a three-layer back propagation neural network for dealing with bus load forecasting. After discussing the standard BP algorithm and its flaw, the paper presents a modified BP algorithm. In order to design an accurate and efficient model, many methods are used. The input variables have been determined according to the values of their correlation coefficients. The paper use three temperature quantizing ways, several styles of nodes of hidden layer, numbers of samples and two transfer functions during artificial neuralnetwork (ANN) model running. After analysing experiment data, an more reasonable network is presented The forecasting results shows that the model is efficient. The new model including weather factor is showed more reasonable and reliable than before.In order to enhance the precision of forecasting, the paper also provides a combination model which combine the ARIMA and BP neural network together. It is based on the principle of optimal credit coefficient of sequence forecasting result. The forecasting result is better than any singular model used in the paper.The characteristic of the model presented in the paper is with high run-speed and better precision . The model with innovation shows more convenient, reasonable and practicable.