作者：梁小姣¹, 马春玲¹, 马传国¹, 辛少菲¹, 程林², 史玉良^2,3

作者单位：1. 国网山东省电力公司东营供电公司，山东 东营 257100;
2. 山东大学软件学院，山东 济南 250101;
3. 山大地纬软件股份有限公司，山东 济南 250200

关键词：用电负荷;环保数据;对比学习;GAN算法;模型预测控制

摘要：

在能源转型背景下，高效、精确的环保指标监测是推动工业绿色发展的前提条件。故该文开展企业用电数据与环保数据的关联映射研究，旨在为环保监测的数字化转型发展提供探索思路。首先，通过对比学习实现用电数据与环保数据分布规律的关联映射，从而基于用电数据提取排污监测的Encoder数据；随后，结合相似样本均值，采用GAN算法构建企业排污量监测模型，并采用Wasserstein距离构建损失函数，提高生成排污监测数据的质量；最后，基于模型预测控制方法进行模型参数二次微调，保证生成数据的稳定性与预测前驱一致性。实验测试结果表明：该文方法能够根据用电负荷数据生成符合实际生产的企业排污监测数据。

Dynamic monitoring method of pollutant discharge of enterprises based on contrastive learning of power and environmental protection indicators
LIANG Xiaojiao¹, MA Chunling¹, MA Chuanguo¹, XIN Shaofei¹, CHENG Lin², SHI Yuliang^2,3
1. State Grid Shandong Electric Power Company Dongying Power Supply Company, Dongying 257100, China;
2. School of Software, Shandong University, Jinan 250101, China;
3. Shanda Dareway Software Co., Ltd., Jinan 250200, China
Abstract: In the context of energy transition, efficient and accurate environmental monitoring is a prerequisite for green industrial development. The paper conducts a study on the correlation mapping of enterprise electricity consumption data and environmental protection data, aiming to provide exploration ideas for the digital transformation development of environmental protection monitoring. First, the correlation mapping between the electricity consumption data and the environmental protection data distribution law is realized through contrastive learning, so as to obtain the Encoder data of pollution discharge prediction based on the electricity consumption data.Then, combined with similar sample means, GAN algorithm was used to build pollutant discharge prediction model of enterprises, and by Wasserstein distance to construct the loss function to improve the quality of the generated pollutant discharge prediction data. Finally, the model parameters are fine-tuned based on MPC, which ensures the stability and trend consistency of the generated data. Experiments show that this method can generate enterprise sewage data that conforms to the actual situation according to the electricity load.
Keywords: electricity load;environmental protection data;contrastive learning;GAN algorithm;model predictive control
2022, 48(10):117-124,144  收稿日期: 2022-05-10;收到修改稿日期: 2022-07-18
基金项目: 山东省重点研发计划（重大科技创新工程）项目（2021 CXGC010103）
作者简介: 梁小姣（1980-），男，山东荣成市人，高级工程师，研究方向为电力数据应用
参考文献
[1] 金立艳, 齐永宏, 赵明辉, 等. 气密空间多环境信息智能监测系统[J]. 中国测试, 2021, 47(12): 131-135+168
[2] 严玥, 江赟, 严实. 利用RBF网络的火电厂氮氧化物浓度检测方法[J]. 电子测量与仪器学报, 2017, 31(1): 45-50
[3] 杨训政, 柯余洋, 梁肖, 等. 基于LSTM的发电机组污染物排放预测研究[J]. 电气自动化, 2016, 38(5): 22-25
[4] 安军, 陈启鑫, 代飞, 等. 面向大气污染防治的电力绿色调度策略研究与实践[J]. 电网技术, 2021, 45(2): 605-612
[5] 刘忠辉. 基于电力数据的污染气体监测系统设计[D]. 广州: 广东工业大学, 2021.
[6] GANIN Y, USTINOVA E, AJAKAN H, et al. Domain-adversarial training of neural networks[J]. Journal of Machine Learning Research, 2016(17): 1-35
[7] BABUTA A, GUPTA B, KUMAR A, et al. Power and energy measurement devices: a review, comparison, discussion, and the future of research[J]. Measurement, 2021, 172 (prepublish).
[8] REN L, PAN Z B, CAO J Z, et al. Infrared and visible image fusion based on variational auto-encoder and infrared feature compensation[J]. Infrared Physics & Technology, 2021(4): (prepublish)
[9] 廖清阳, 王军, 胡凯强, 等. 基于深度并行CNN-BiLSTM的能源互联网电负荷和热负荷联合预测模型[J]. 中国测试, 2022, 48(4): 146-153
[10] 刘建伟, 赵会丹, 罗雄麟, 等. 深度学习批归一化及其相关算法研究进展[J]. 自动化学报, 2020, 46(6): 1090-1120
[11] CHEN Y, XIA S X, ZHAO J Q, et al. Appearance and shape based image synthesis by conditional variational generative adversarial network[J]. Knowledge-Based Systems, 2020, 193: 105450
[12] GHOSH A, KULHARIA V, NAMBOODIRI V P, et al. Multi-agent diverse generative adversarial networks[J]. Computer Vision and Pattern Recognition, 2018, 3: 8513-8521
[13] ANDERS B L L, SøREN K S, HUGO L, et al. Autoencoding beyond pixels using a learned similarity metric[C]// International Conference on Machine Learning. PMLR, 2016, 1558-1566.
[14] QUINTANAR R A L, RODRIGUEZ S F, PICHARDO M J L, et al. Capture device identification from digital images using kullback-leibler divergence[J]. Multimedia Tools and Applications, 2021, 80(13): 19513-19538