基于深度学习的断面约束概率快速预警方法  

Fast Probabilistic Pre-Warning Method for Transmission Capacity Constraints Based on Deep Learning

作　　者：周毓敏[1] 邓韦斯 吴云亮[1] 张蔷[1] 赖晓文 杨燕 孙宇军 ZHOU Yumin;DENG Weisi;WU Yunliang;ZHANG Qiang;LAI Xiaowen;YANG Yan;SUN Yujun(CSG Power Dispatching Control Center,Guangzhou 510663,China;Beijing Tsintergy Technology Co.,Ltd.,Beijing 100084,China)

机构地区：[1]中国南方电网电力调度控制中心,广州510663 [2]北京清能互联科技有限公司,北京100084

出　　处：《南方电网技术》2020年第9期45-52,共8页Southern Power System Technology

基　　金：中国南方电网有限责任公司科技项目(ZBKJXM20180070)。

摘　　要：断面安全约束对现货市场出清的计算效率和结果产生直接影响,目前现货市场交易组织前主要通过方式专家的经验对确定性的场景进行断面约束预警。现货市场模式下电网运行方式更为多变,为增强断面约束设置的合理性、全面性,考虑新能源、负荷、电厂报价等预测的不确定性,首先提出了基于蒙特卡洛的断面约束预警思路,该思路可有效评估断面约束的安全裕度/风险程度。在此基础上,考虑蒙特卡洛法需要求解大量样本存在计算负担大的问题,进一步提出了基于深度学习技术的断面预警快速求解算法,该算法通过函数映射的方式直接求解出所有蒙特卡洛法产生的样本,可实现秒级计算,为方式人员判定现货出清中的约束条件提供参考。最后,在接入新能源的IEEE 118标准测试系统中展开算例分析,验证了所提方法的有效性。The section constrains could make an impact on the calculating efficiency and results of the spot market clearing.At present,the pre-warning methods of transmission capacity constraints generally focus on deterministic scenarios through artificial experience.The operation mode of power system could be more complicated under spot market.To enhance the compatibility and comprehensiveness of section constrains in spot market clearing,this paper proposes a pre-warning idea of transmission capacity constraints based on Monte Carlo method to handle the uncertainty of the prediction error from renewable energy,load,and power plant bidding,which can effectively evaluate the safety margin/risk degree of transmission capacity constraints.On this basis,considering that Monte Carlo method needs to solve a large number of samples,which the computational burden is huge,this paper further proposes a fast algorithm for transmission capacity constraints pre-warning based on deep learning technology.The algorithm directly solves all the samples generated by Monte Carlo by function mapping.The samples can be calculated by the proposed algorithm in seconds.Finally,the effectiveness of the proposed method is verified in the IEEE 118 standard test system with renewable energy sources.

关 键 词：电力现货市场 断面预警 不确定性 深度学习 蒙特卡洛 

分 类 号：F426[经济管理—产业经济] TM744[电气工程—电力系统及自动化]