基于增量约简算法确定电力负荷预测模型输入参数

针对电力系统中有众多因素影响负荷预测精度的问题,文中引入粗糙集理论中的属性约简算法来挖掘与待预测量相关性大的各属性,保证预测模型输入参数的合理性,解决了神经网络模型输入参数的确定问题.针对基于区分矩阵约简算法是NP问题的弱点,提出了基于属性优先级函数的启发式约简算法(RAPHF);针对负荷预测过程中样本数据是滚动更新的特点,在RAPHF的基础上提出了具有动态挖掘能力的粗糙集增量算法RAPHF-I.通过短期负荷预测的实例研究,证实了文中改进算法的有效性.

Input Parameters Selection in Short-term Load Forecasting Model Based on Incremental Reduction Algorithm

A reduction algorithm based on rough set theory is put forward due to various factors that influence accuracy in the power load forecasting. The reduction algorithm introduced to mine more correlative attributes in the pending forecasting components, ensures not only the rationality of input parameters of forecasting model but also the selection of input parameters of ANN model. A reduction algorithm through prior heuristic function (RAPHF) algorithm based on attributes-prior algorithm is introduced because reduction algorithm based on dipartite matrix reduction algorithm is a NP problem. On the basis of RAPHF, a rough set incremental algorithm with dynamic mining ability, namely, RAPHF-I is proposed by considering the updating samples. The efficiency and advantage of the proposed method is proved by prediction results of short-term load based on the RAPFF and RAPHF-I.