基于改进栈式自编码算法的中期负荷预测研究

Research on Medium Term Load Forecasting Method Based on Improved Stacked Auto-Encode

综合考虑影响电力系统中期负荷预测的各个因素并提高预测精度,提出了基于改进栈式自编码算法的中期负荷预测方法。利用ICA进行特征提取,筛选主导的影响因素并对数据进行归一化处理,结合模糊理论构建气温因素的隶属度关系,通过参数自适应微分进化算法对栈式自编码算法参数展开在线优化,进而匹配最佳基于改进栈式自编码算法的组合中期负荷预测模型,并展开案例分析。其结果表明:该改进算法通过影响因素筛选、参数选择与优化,能够有效避免参数选择的盲目性,将气温动态因素进行模糊化处理,能够进一步提高预测精度,其预测结果优于PSO-SVM算法,与实际结果更为接近,且稳定性好,为电力系统负荷预测提供了重要手段。

Considering all the factors affecting the medium term load forecasting and improving the forecasting accuracy,a medium term load prediction method based on improved stacked auto-encode is proposed.Firstly,the mathematical model of ICA is constructed to extract the characteristic quantity,screen out the factors affecting the power system load and normalize it.Then,the membership relationship of temperature factors is constructed combined with fuzzy theory.The parameters of stacked auto-encoder are optimized online by adaptive differential algorithm.Finally,the combined medium term load forecasting model based on improved stacked auto-encode is applied to case analysis.The results show that the improved algorithm can effectively avoid the blindness of parameter selection through influencing factor selection,parameter selection and optimization.The temperature value factor can be fuzzified to further improve the prediction accuracy.Its prediction results are better than the PSO-SVM algorithm,which is closer to the actual results.So it has a good stability.This paper provides an important means for power system load prediction.