中文摘要：

随着我国经济的发展和经济结构的调整,居民用电占全社会用电量的比重逐渐增大并且有继续增加的趋势,科学合理地预测居民用电水平将为电力规划与需求侧管理提供决策基础。首先,采用相关系数法进行居民用电关键影响因素的选择。其次,将选取的影响因素作为LS-SVM的输入端,城乡居民用电量作为输出端,用Bayes准则进行SVM的参数选取,通过智能模拟学习,建立了Bayes-LS-SVM居民用电预测模型。最后,以中国某省居民用电量预测为例进行学习以及测试,并将其预测结果与广义回归神经网络预测法及几种常用的居民用电预测方法进行误差对比分析,证明了该组合方法比其它几种方法更精确有效。提出了采用人工智能的方法通过家用电器以及其他影响因素来预测居民用电,克服了以往采用家用电器预测中,家用电器功率以及年利用小时数预测不准确的问题。

英文摘要：

With the development of the economy and the adjustment of economic structure, the proportion of residential electricity consumption is gradually increasing and will continue to increase, thus scientific and reasonable forecasting of residential electricity consumption can provide decision foundation for power planning and DSM. Firstly, correlation coefficient was used for variable selection of influencing factors of residential electricity forecasting. Then model of Bayes least squares support vector machine（Bayes-LS-SVM） was established through intelligent simulation study and Bayes rule was applied to choose the suitable parameter combination for LS-SVM model while the influencing factors chosen by correlation coefficient were as the input terminal and urban and rural residential electricity consumption as the output terminal through learning and testing of the residential electricity of a certain province. The prediction error based on the artificial intelligence method Bayes-LS-SVM was compared with generalized regression neural network（GRNN） and several common residential electricity prediction methods, then the empirical results revealed that the proposed model outperformed the other models. This paper brought forth new ideas in forecasting of the residential electricity based on Bayes-LS-SVM through influence factors of household appliances and others, then the problems of inaccurate prediction in power and annual utilization hours of household appliances have been overcomed in the past.