论文部分内容阅读
为解决短期电力负荷预测中单一预测方法的预测精度差、计算时间长等问题,提出了基于相似日搜索的PSO(粒子群优化算法)优化WNN(小波神经网络)的短期电力负荷组合预测方法.首先利用模糊聚类分析方法筛选与待预测日相似的日期数据作为组合预测模型的训练样本,使训练更具有针对性,提高了训练的精度并缩短了计算时间.再通过PSO算法优化小波神经网络,克服了以往BP(back propagation)算法易陷入局部最优,且搜索效率低下等问题.实验表明,这种组合预测模型的预测精度相对于其它方法有较大提高.
In order to solve the shortcomings of the single prediction method in short-term power load forecasting, such as the poor prediction accuracy and the long computation time, a short-term load forecasting method based on PSO (Particle Swarm Optimization) with similar day search and WNN (Wavelet Neural Network) is proposed. Firstly, the fuzzy clustering analysis method is used to screen the date data similar to the predicted date as the training samples of the combined forecasting model to make the training more pertinence, improve the training accuracy and shorten the calculation time.Secondly, the wavelet neural network , Which overcomes the problems that BP (back propagation) algorithm is easy to fall into the local optimum and the search efficiency is low, etc. The experiments show that the prediction accuracy of this combined prediction model is greatly improved compared with other methods.