Research For Short-term Load Forecasting Based On District Power Grid

Short-term load forecasting as a basic arithmetic is essential to the operation of Power System. The result of short-term Load forecasting is the foundation that the Power System arranges generation plan or estimates standby capacity base on. The results accurately or not have significance to the safe and economic operation of the grid. Therefore how to improve the accuracy of short-term load forecasting has been a importance problem that people committed to research.Currently, there have many literatures about short-term load forecasting domestic and foreign. But because many complex factors affect the load and load uncertainty, making short-term load forecasting has not been a very satisfactory solution. At first, this paper expounds on the concept and meaning of the short-term load forecasting. Then this paper introduces the current status of research on the short-term load forecasting briefly, and analyses the advantages and disadvantages of the existing types of load forecasting methods.Secondly, the neural network theory is discussed briefly in this paper. At the same time, the paper analyses the two forward-type networks detailedly and also ascertains their differences.In this paper the electricity load characteristics about district power grid are analyzed and discussed deeply. For a specific power grid, some predictable factors that affect electricity load's change are researched. This paper also explores detailedly the specific relations between the various factors and load characteristics.On the basis of an in-depth analysis about the load characteristics and neural network, a Cascade Neural Network (CNN) load forecasting models based on BP Network and RBF Network is put forward in this paper. In this method the influences of measurable factors and past load on load forecasting are separately considered by two sub-networks. In the process of the model realizing, all kinds of input variables and the correlative factors are dealed with in a reasonable and quantify treatment. Finally this model is applied to normal day load forecasting. The final results of the applicable example show that the forecasting results' precision of the Cascade Neural...