智能电网环境下的负荷预测理论与方法研究

This project is in close connection with the status of developing smart grid in China, aiming at overcoming new challenges to load forecasting brought by the development of smart grid, realizing optimization scheduling based on energy conservation and environmental protection. New theory and method system of load forecasting under the environment of the smart grid is constructed taking into account a variety of factors on load forecasting multi-levelly and strategely. In General, the theory and method of power system load forecast based on comprehensive hourly weather factors is constructed. Comprehensive meteorological index constructing method, cumulative effect, delay effect are researched. On the supply side, considering random, intermittent nature , wind power and photovoltaic power load adaptive forecasting theory and method using metal learning algorithm based on NWP (numerical weather forecast) is researched. On the demand side, electric vehicle's law of average daily electric power consumption is researched. Probability theory and method of load forecasting in electric vehicle daily charge are established using integrated probabilistic analysis, economic analysis. Automatic extraction method based on morphological similarity criteria for typical customers is studied. Interactive load forecasting theory and method is established considering a variety of factors, particularly the different demand response effects on typical customers load. On the basis of above studies, new generation of load forecasting platform in the environment of the smart grid will be developed helping to raise the level of load forecast.

紧密结合我国大力发展智能电网的现状，为克服智能电网发展给负荷预测带来的全新挑战，实现节能环保优化调度，多层次、多策略地考虑各种因素对电网负荷预测的影响，构建智能电网环境下的负荷预测理论与方法体系。在总体上，研究综合气象指标构造方法、累积效应、延迟效应，建立基于综合实时气象要素的电网负荷预测理论与方法。在供电侧，以数值天气预报NWP为基础，以元学习组合算法为核心，建立具有随机性、间歇性特点的风力发电与光伏发电负荷自适应精细化预测理论与方法。在用电侧，综合运用概率分析、经济学分析等科学原理，研究电动汽车电能日均消耗规律，建立电动汽车日充电概率性负荷预测理论与方法；研究基于形态相似准则的典型用户自动提取方法，分析多种因素，特别是不同情况下需求响应对典型用户负荷的影响，建立互动性负荷分析预测理论与方法。在以上研究基础上，开发满足智能电网环境下的新一代负荷预测平台，提高负荷预测水平。

紧密结合我国大力发展智能电网的现状，为克服智能电网发展给负荷预测带来的全新挑战，实现节能环保优化调度，多层次、多策略地考虑各种因素对电网负荷预测的影响，构建智能电网环境下的负荷预测理论与方法体系。在总体上，提出了一种用户分类的方法，实现对居民用电行为特性的差异化分析；提出了一种基于云计算的智能电网负荷预测平台架构，深入阐述了云计算关键技术、负荷预测云架构、云平台负荷预测服务等，采用Hadoop云计算技术，对负荷预测的资源调度和计算进行虚拟云计算仿真，验证了以电力私有云为基础建立的负荷预测机制更优良。在供电侧，以经验分解法算法和局部均值分解算法为核心，建立了具有随机性、间歇性特点的风力发电与光伏发电出力自适应精细化预测理论与方法。在用电侧，应用蒙特卡洛方法模拟计算充电负荷，预测未来的充电负荷曲线，并根据曲线分析了充电负荷对电网的影响；考虑需求响应条件下实时电价与负荷之间的相互影响，采用基于数据挖掘技术的模糊循环推理系统模拟人的思维过程，通过挖掘电价变化量、负荷变化量等变量之间的关联规则，模拟电价与负荷预测之间存在博弈过程,对多变量最小二乘支持向量机的初步预测结果进行循环修改，直至负荷和电价预测结果趋于稳定。在以上研究基础上，开发了满足智能电网环境下的新一代负荷预测平台，提高了负荷预测水平。