基于电力负荷历史数据挖掘的负荷预测算法研究

针对电力负荷历史数据中异常数据点影响电力负荷预测精度的缺陷，研究基于电力负荷历史数据挖掘的负荷预测算法。选取K means聚类算法挖掘电力负荷历史数据的属性特征量，检测其中所包含的异常数据点，选取灰色系统理论中的GM（1，1）模型修正电力负荷历史数据中的异常数据，利用完成修正的电力负荷历史数据建立训练集以及预测集，将训练集样本输入支持向量机中，利用支持向量机所具有的非线性映射能力映射样本至高维空间内，获取支持向量机最优阈值，将预测集输入具有最优阈值的支持向量机中，获取精准的电力负荷预测结果。所研究算法可实现长期、短期、超短期电力负荷的预测，且预测的精准性及速度较为优越。

Research on load forecasting algorithm based on electric load historical data mining

Aiming at the defect that abnormal data points in power load history data affect the accuracy of power load forecasting，a load forecasting algorithm based on power load history data mining is studied.K-means clustering algorithm is selected to mine the attribute feature quantity of power load historical data，detect the abnormal data points contained therein，GM（1，1） model in grey system theory is selected to correct the abnormal data in power load historical data，the training set and prediction set are established by using the corrected power load historical data，and the training set samples are input into support vector machine.The nonlinear mapping ability of support vector machine is used to map samples into high-dimensional space to obtain the optimal threshold of support vector machine，and the prediction set is input into the support vector machine with the optimal threshold to obtain accurate power load forecasting results.