摘 要 售电量是电力市场中一个重要的经济指标。在某个时间段内售电量反映了地区经济发展对电量的整体需求情况，也从侧面反映了当地电网经营企业的电量销售和经营情况。售电量预测作为电力经营生产中的一项重要工作，对于电网升级改造、企业经营、财务管控、智能电网大数据及电力市场营销策略的制定和调整具有非常重要的意义。 本文首先分类对影响售电量的内部影响因素和外部影响因素进行分析。外部影响因素主要从人口、经济社会发展、固定投资、环境及电力价格政策等五方面，内部因素主要结合全行业的分类和供电企业营销统计报表数据源划分为农林牧渔业、工业、建筑业等九个方面。结合实际情况和预测的时效性要求，月售电量的影响因素主要考虑内部影响因素；年售电量的影响因素主要考虑内、外部影响因素。 其次本文从细化影响年、月售电量影响因素、发挥自组织自学习电量预测的复杂非线性映射、提高预测精度考虑出发，寻求基于BP神经网络建立了售电量预测的新方法，设计了售电量的预测模型，并使用MATLAB编程分别实现对宁夏银川市、西吉县、彭阳县等地区的年、月售电量仿真预测，预测结果表明，基于BP神经网络的售电量预测新方法可行。同时，以银川市年、月售电量预测为例分别与常用的其他三种售电量预测方法的预测结果进行了对比分析，结果表明，基于BP神经网络的售电量预测新方法的预测精度较其他三种常用方法高，能够满足实际需要。 最后本文基于MATLAB GUI设计了基于BP神经网络的售电量预测系统。该系统分版本信息、外部导入数据、数据归一化、神经网络参数结构和网络训练参数设置、预测控制及误差绘图模块、附属菜单设置模块等六大模块。在实例验证中，利用该系统对宁夏银川市2014年的月售电量进行了预测，预测结果表明，该系统能够简单、快速的对地区售电量进行预测。

ABSTRACT In the whole society, Electricity sales is an very important economic index in the power market, which reflected the overall demand for electricity in a certain period of time, and also reflected the sales and operation of the electricity in the local power grid. As an important work in power management and production, electricity sales forecast is of great significance to the upgrading and transformation of power grid, enterprise management, financial management and management, big data in smart grid and the formulation and adjustment of the marketing strategy of electric power. First of all, this paper classifies the internal and external factors by the affection of the electricity sales. The external factors mainly included five aspects, such as population, economic and social development, fixed investment, environment and electricity price policy. The internal factors are mainly included nine aspects, such as agriculture, forestry, animal husbandry, industry, construction, and so on, which are mainly classified into the classification of the whole industry and the data sources of the marketing statistics report of the power supply enterprises. The influence of the monthly electricity sale mainly consider the internal influence factors according to the actual situation and the requirement of the prediction timely; the influence of the annual electricity sale mainly consider the internal and external factors. Secondly, a new method seeks to be established which based on BP neural network. It considered of the monthly and yearly influencing factors, the complex nonlinear mapping of self organizing and self-learning forecasting, and the improvement of the forecasting accuracy. The model was designed for the forecasting of electricity sales, which simulated the forecasting of monthly and yearly sales in Yinchuan, Xiji and Pengyang by the programming of MATLAB. The results show that the accuracy of the new method of electricity sales based on BP neural work was better than the other three methods frequently. Finally, we design a forecasting system of electricity sales based on BP neural network by the MATLAB GUI. The system is divided into six major modules: version information, external import data, data normalization, neural network parameter structure and network training parameter setting, forecasting control and error drawing module, subsidiary menu setting module and so on. An example is provided to verify the forecasting of the Yinchuan monthly sales in 2014. In the case verification, it shows that the system can forecast the sales of electricity simply and quickly in area.