基于改进GEP的农网短期负荷预测

Short-term load forecasting of countryside distribution network based on improved gene expression programming

针对传统基因表达式程序设计(GEP)算法初始种群的产生是随机的,变异率不能在进化的过程中做适应性的调整,对基因的好坏没有标识,在解决相类似的问题时不能利用以往的运行结果的缺点进行了改进,提出了过度繁殖、环境因子选择,自适应变异率,信息素标识基因的好坏,借鉴以往程序运行得到的数学模型的方法,并将改进后的基因表达式程序设计(IGEP)算法应用于农村配电网短期负荷预测中。预测过程是先对负荷样本进行数据预处理,消除伪数据,然后运用改进后的基因表达式程序设计的灵活表达能力,把不同日同一时刻的负荷序列作为样本,分别对平日(周一到周五)和周末的未来时刻的负荷进行分时短期预测。算例表明改进后的基因表达式程序设计算法具有较高的效率,比遗传程序设计(GP)和基因表达式程序设计(GEP)算法具有更好的预测效果。

Gene expression programming （GEP） was improved to overcome the shortcomings that the initial population was generated randomly. There were no standards to measure the gene, mutation rate could not be adjusted by itself and evolution result got before could not be utilized. The method that excessive multiplication, environmental factor selecting, using pheromones to measure gene, self-adaptive mutation rate and adopting mathematical model got before was proposed. The improved gene expression programming （IGEP） was applied to countryside distribution network short-term load forecasting. Firstly, the load series of the same time but different days were chosen as the training samples. Secondly, the load samples were filtered and processed generally. And finally, the short-term load was forecasted by weekday and weekend after eliminating the pseudo-data. After comparison with the results forecasted by means of genetic programming （GP） and GEP, it proves that the method of IGEP in countryside distribution network short-term load forecasting is better.