考虑业扩报装的相关向量机月度负荷预测方法

摘要

Considering that the traditional monthly load forecasting method doses not take the load's intrinsic factors in?to account,a monthly load forecasting method is proposed with the consideration of business expansion based on rele?vance vector machine(RVM). In the proposed method,the electricity consumption trend after business expansion is studied by using growth curve fitting and k-means clustering algorithm,which is further used to extract monthly effect ratio and calculate the business expansion increment that has a substantial impact on the monthly load. Then,a load forecasting model is established based on SVM with the actual business expansion increment and historical load data as sample inputs. Meanwhile,particle swarm optimization and compound kernel function are used to improve the adapt?ability of the proposed model. From the comparison of forecasting results among the models which consider the actual and unmodified business expansion increments respectively,as well as the one that does not consider business expan?sion increment,it is proved that the actual business expansion increment influences the monthly load obviously and it can help to improve the accuracy of forecasting effectively.%针对传统月度负荷预测方法缺乏考虑负荷内在影响因素的问题,该文提出了考虑业扩报装的相关向量机月度负荷预测方法.该方法通过生长曲线拟合和k-均值聚类研究业扩报装后的用电趋势,提取出逐月影响比例,计算得到对当月负荷具有实际影响的业扩增量;将实际业扩增量和历史负荷数据作为样本输入,建立基于相关向量机的负荷预测模型,同时利用粒子群优化参数和组合核函数提高模型适应度.考虑实际业扩增量、考虑未修正业扩增量以及不考虑业扩报装的预测结果比较表明,实际业扩增量对月度负荷有较重要的影响,可以有效提高预测的精度.