基于量子遗传算法的FARIMA模型电力负荷短期预测  

Short-term prediction of power load based on FARIMA model for quantum genetic algorithm

作　　者：杨照坤 宋万清[1] 曹琨 YANG Zhaokun;SONG Wanqing;CAO Kun(School of Electronic and Electric Engineering,Shanghai University of Engineering Science, Shanghai 201600,China)

机构地区：[1]上海工程技术大学电子电气工程学院

出　　处：《传感器与微系统》2019年第10期143-145,共3页Transducer and Microsystem Technologies

基　　金：上海工程技术大学机械电子工程学科建设项目(2018XK—A—03);上海市自然科学基金资助项目(14ZR1418500)

摘　　要：根据电力负荷历史数据对未来负荷趋势进行预测。通过计算电力负荷序列的赫斯特指数(Hurst)参数值,表明振动烈度序列符合明显的长相关特性,用长相关FARIMA模型来对电力负荷序列进行预测,为了提高预测的准确性,提出了一种基于量子遗传算法优化的长相关FARIMA预测模型。利用量子遗传算法优化经分数差分后的平稳模型ARMA(p,q)阶数,根据合适的适应度值进行全局搜索,从而能够确定模型的最佳阶数(p,q),并将其应用到实际的电力负荷值预测中验证。将实验结果与传统的AIC(p,q)准则定阶的ARMA模型进行预测的结果,以及实际电力负荷值数据进行比较,结果表明:该方法对电力负荷值的预测具有更高的准确性。Future load trends is forecasted based on historical data of power load.By calculating the Hurst parameter value of the power load sequence,it is shown that the vibration intensity sequence meets the obvious long correlation characteristics.Therefore,the long correlation FARIMA model is used to predict the power load sequence.In order to improve the accuracy of prediction,a long correlation FARIMA prediction model based on quantum genetic algorithm optimization is proposed.The quantum genetic algorithm is used to optimize the ARMA(p,q) order of the stationary model after the fractional difference,and the global search is performed according to the appropriate fitness value,so that the optimal order(p,q) of the model can be determined and applied to actual power load value prediction for verification. Experimental results are compared with the results of the traditional AIC(p,q) criteria fixed-stage ARMA model,and the actual power load value data.The comparison results show that the method has higher accuracy of power load value predicting.

关 键 词：电力负荷预测 HURST指数 长相关 FARIMA模型 量子遗传算法 

分 类 号：TM714[电气工程—电力系统及自动化]