摘要
为了协调电力系统供需平衡,降低运维风险和成本,引入Shaping技术改进了深度确定性策略梯度(DDPG)算法,得到Shaping-DDPG预测模型。加入经验回放技术和目标网络技术以消除数据间的关联性;设计变压器网络评估函数来评价变压器网络的优劣状态;通过数据处理模块和卷积模块提取原始数据特征,提高变压器系统的感知能力和学习效率。研究表明:与其他算法预测效果相比,Shaping-DDPG模型的RMSE误差平均值(93 MW)最低,比DDPG模型、RNN模型和SVM模型分别降低了42 MW、93 MW和145 MW。相较于非线性变压器负荷系统,Shaping-DDPG模型具有强大的反馈记忆功能,能准确获取负荷序列潜在的变化趋势,在变压器负荷曲线呈现波动时依然能够保证良好的预测能力。该研究为降低电网公司资源浪费和运维成本、协调电网公司与变压器系统之间的供需平衡提供了思路,提高了运作效益。
To coordinate the supply and demand balance of the power system and reduce the risk and cost of operation and maintenance,Shaping technology was introduced to improve the deep deterministic policy gradient(DDPG)algorithm,and the Shaping-DDPG prediction model was obtained.The correlation between data was eliminated by adding experience playback technology and target network technology;the quality of the transformer network was evaluated by designing a transformer network evaluation function;the original data features were extracted by using the data processing module and the convolution module,and the perception ability and learning efficiency of the transformer system were improved.Compared with the prediction effects of other algorithms,the average RMSE(root mean square error)error(93 MW)of Shaping-DDPG model was the lowest,which was 42 MW,93 MW and 145 MW lower than the DDPG model,RNN model and SVM model respectively.Compared with non-linear transformer load systems,the Shapeing-DDPG model had a powerful feedback memory function to accurately obtain the potential change trend of the load sequence,and it can still ensure good predictive ability when the power load curve is fluctuating.This study provides ideas for reducing the waste of resources and operation and maintenance costs of power grid companies,coordinating the balance of supply and demand between grid companies and transformer systems,and improving operational efficiency.
作者
山宪武
宋秩行
邱书琦
叶新青
SHAN Xianwu;SONG Zhixing;QIU Shuqi;YE Xinqing(State Grid Xinjiang Electric Power Co.,LTD,Urumqi 830063,China;不详)
出处
《武汉理工大学学报（信息与管理工程版）》
CAS
2023年第3期377-382,396,共7页
Journal of Wuhan University of Technology：Information & Management Engineering
基金
国家电网公司科技研发项目(SGGSKY00FJJS2000201)
国网新疆电力有限公司电力科学研究院项目(新电发【2019】233号).
