【摘要】 电网数据具有海量、高维的特点,现有的短期电力负荷预测模型无法提取用户的用电习惯。提出一种基于负荷聚类的全网短期负荷预测模型,首先采用自组织映射网络对全网负荷进行聚类,将不同特性的用户负荷曲线作为子网;然后引入遗传算法对Elman神经网络的参数进行寻优,得到针对不同子网负荷特性的差异化预测网络;最后基于负荷综合稳定度得到全网负荷预测结果。将该集成模型用于某市电网进行算例仿真,预测结果表明,所提方法比传统预测方法的准确率更高,同时适用于部分子网数据缺失而需要得到全网结果的情况。更多还原

【Abstract】 The power grid data has the characteristics of massive and high-dimensionality, and the existing short-term power load forecasting models cannot extract users’ electricity consumption habits. To this end, a network-wide short-term load forecasting model based on load clustering was proposed. Firstly, self-organizing mapping network was adopted to cluster the load of the whole network, and user load curves with different characteristics were used as subnets; then genetic algorithm was introduced to optimize the parameters of Elman neural network. The optimization was performed to obtain a differentiated prediction network for the load characteristics of different subnets, and finally the load prediction result of the whole network was obtained based on the comprehensive stability of the load. The integrated model was applied to a certain city’s grid for example simulation. The prediction results show that the proposed method is more accurate than the traditional prediction methods, and also suitable for the situation where some sub-network data is missing, but the whole network result needs to be obtained.更多还原