基于动态贝叶斯网络的电价区间预测  

Prediction interval forecasts of electricity price based on dynamic Bayesian networks

作　　者：王洪涛 邹斌[1] WANG Hongtao;ZOU Bin(School of Mechatronic Engineering and Automation,Shanghai University,Shanghai 200444,China;School of Information Mechanical and Electrical Engineering,Ningde Normal University,Ningde 352100,China)

机构地区：[1]上海大学机电工程与自动化学院,上海200444 [2]宁德师范学院信息与机电工程学院,福建宁德352100

出　　处：《电力系统保护与控制》2022年第5期117-127,共11页Power System Protection and Control

基　　金：福建省自然科学基金项目资助(2019J01845)。

摘　　要：在高比例可再生能源参与市场竞争的背景下,电价波动更为剧烈。为了对电价区间进行预测,提出了动态贝叶斯网络(Dynamic Bayesian Network, DBN)的电价区间预测方法。该方法以风电发电量、总发电量和总用电量的预测值以及历史电价真实值为输入数据,以贪婪搜索算法确定DBN的网络结构,以最大似然估计法(Maximum Likelihood Estimate, MLE)估计DBN网络参数,建立DBN模型。然后以风电发电量、总发电量和总用电量的预测值为推理证据,采用联合树推理得到电价预测的离散值和后验概率,实现电价的区间预测。最后将所提方法与电价真实值、对比方法进行比较,验证了所提方法的有效性。所提方法不仅能得到电价的预测区间,而且能给出对应的概率,对提高市场成员的收益、规避价格风险具有指导意义。Given the high proportion of renewable energy participating in market competition, the fluctuation of electricity price will be more severe. In order to predict the range of electricity price, a dynamic Bayesian network(DBN)interval prediction method is proposed. In this method, the predicted data of wind power generation, total power generation and total electricity consumption, and the real value of electricity price, are taken as input data. The network structure of DBN is determined by a greedy search algorithm, and the network parameters of DBN are estimated by maximum likelihood estimated(MLE). The DBN model is established. Then, with the predicted value of wind power generation, total power generation and total electricity consumption as the reasoning evidence, the discrete value and a posteriori probability of electricity price prediction are obtained using union tree reasoning, and the interval prediction of electricity price is realized. Finally, the proposed method is compared with the real value of electricity price and the comparison method to verify the effectiveness of the proposed method. The proposed method can not only get the predicted range of electricity price, but also give the corresponding probability. This has guiding significance for increasing the income of market members and avoiding price risk.

关 键 词：电价预测 区间预测 动态贝叶斯网络 联合树推理 向前向后算法 改进k-means聚类 平均差异度 

分 类 号：F426.61[经济管理—产业经济]