Short-term Load Forecasting Research Considering The Effects Of Small Hydropower And Impact Load

Electric power system short-term load forecasting is an important component of the load forecasting, load at a certain moment in the future, often with the past load level, the current operation condition, forecast of meteorological factors and date types are closely related. But the load of different regions has its own particularity.In this article the research content is considering the effects of small hydropower and impact load of short-term load forecasting, as more and more distributed power grid, not only threatens the safety of power grid stability, also caused significant to power system short-term load forecasting. So. in the region, it is urgent necessary to improve the level of short-term load forecasting. Based on the above problems, this paper made the following research work:First of all, this article will more small hydropower and impact load region as the research object, according to the region of the power system load data for analysis of two aspects, one is analyzing the region electric power system short-term load characteristics and the relationship between the region’s load with all kinds of meteorological indexes, calculated the main influence factors of short-term load forecasting in the region:The second is the original load data preprocessing, to ensure the reliability of the sample.Secondly, due to the complex load characteristics in this region, new samples appear constantly make neural network can not be learning accurately and effectively, leading to low load forecasting accuracy. Aiming at this problem, this paper uses the Bayesian regularization algorithm to improve the network generalization ability. Through the model analysis, the results show that the load forecasting model containing the Bayesian algorithm has more excellent learning ability for complex load characteristics and new sample data. But there are still two problems of larger load forecasting error in sometime and not stable relative error of96points a day load forecasting.Once again, because of more impact load and different load change frequency, it is difficult to learn such a rule by Bayesian neural network, leading to that load forecasting accuracy was not able to further improve. Continue to modify the prediction model for this problem, this paper will be applied wavelet analysis to neural network load forecasting. Using wavelet transform multi-scale decomposition, the original load data has the practical significance of the four load component. According to the different characteristics of the four component, respectively set up corresponding prediction model and obtain four load weight forecasting results, finally wavelet reconstruction are used to get the final forecasting result. Through the model analysis, the results show that wavelet can better solve Bayesian neural network with small hydropower and impact load area of short-term load forecasting problems.