基于条件生成对抗网络曲线生成的短期负荷概率预测

Short-term Load Probabilistic Forecasting Based on Conditional Generative Adversarial Network Curve Generation

为指导电力系统稳定安全运行,提出一种基于条件生成对抗网络(CGAN)曲线生成的短期负荷概率预测方法。首先,以日期、温度、历史负荷等特征为输入,构建基于双向长短期记忆神经网络的日负荷关键值自适应集成预测模型。其次,采用最大信息系数方法为负荷特征赋权,基于加权K最近邻算法、加权重采样构建相似曲线数据集。然后,以负荷关键值和相似曲线数据集分别作为条件和训练集,构建基于CGAN的负荷曲线生成模型,提出数值偏差量与曲线形态偏差量修正损失函数。最后,考虑模型、噪声不确定性,构造由噪声到模型输出概率分布的映射关系,进行短期负荷概率预测。以中国华东某地区电网负荷数据为例,验证了所提方法相对于传统方法具有更高的预测精度。

In order to guide the stable and safe operation of power systems,a short-term load probabilistic forecasting approach based on conditional generative adversarial network(CGAN)curve generation is proposed.Firstly,an adaptive integrated forecasting model of key values of daily load based on bi-directional long short-term memory neural network is constructed by taking features of date,temperature and historical load as input.Secondly,the maximal information coefficient method is used to weight the load features,and the similar curve data set is constructed based on weighted K-nearest neighbor algorithm and weighted resampling.Then,taking the load key values and the similar curve data set as the conditions and training set,respectively,a load curve generation model based on CGAN is constructed.The numerical deviation and curve form deviation are proposed to correct the loss function.Finally,considering the uncertainty of the model and noise,the mapping from noise to the probability distribution of the model output is constructed,and the short-term load probabilistic forecasting is carried out.Taking the load data of a power grid in a certain region in East China as a case,it is verified that the proposed method has higher forecasting accuracy than the traditional methods.