基于聚类分析的短期负荷智能预测方法研究

短期电力负荷预测作为电网企业的基本工作,其精度的提高对于电网企业运营管理和调度管理具有较大的意义,然而由于电力负荷受到诸多非线性因素的影响,因此得到高精度的电力负荷预测结果是比较困难的.本文首先利用数据挖掘中的k-means聚类技术对训练集的气象数据进行聚类分析,分析提取相似日,在提取相似日的相关历史数据后,建立支持向量机模型进行短期电力负荷预测.经算例结果证明,由该方法得出的预测结果平均相对误差为0.88%,和同结构支持向量机预测的平均相对误差(1.66%)以及ARMA预测的平均相对误差(3.81%)相比,预测精度得到明显的提高,证明了该方法的有效性.

A New Forecasting Approach for Short-term Load Intelligence Based on Cluster Method

Load forecasting is one of the basic issues of the electric power industry. However, because load has a certain social attributes, the improvement of the accuracy of load forecasting result is a difficult issue. This paper first used k-means cluster method to find similar data from historical date and weather data, and then used support vector machine (SVM) for forecasting. Seen from the result, the proposed method's MAPE is 0.88%, but BP-ANN and ARMA are 1.66% and 3.81% respectively. It is proved that this method has a high accuracy.