基于传感数据和机器学习的变压器温升虚拟仿真评估技术  

Virtual Simulation Evaluation Technology for Transformer Temperature Rise Based on Sensor Data and Machine Learning

作　　者：吴杰 吴兴旺 张荐 董荣亮 马健 WU Jie;WU Xingwang;ZHANC Jian;DONG Rongliang;Ma Jian(State Grid Anhui Electric Power Co.Ltd.,Electric Power Research Institute,Hefei 230601,China;China Electric Power ResearchInstitute Co.Ltd.,Beijing 100000,China;Wujiang Transformer Co.Ltd.,Suzhou 215200,China)

机构地区：[1]国网安徽省电力有限公司电力科学研究院,安徽合肥230601 [2]中国电力科学研究院有限公司,北京100000 [3]吴江变压器有限公司,江苏苏州215200

出　　处：《工业加热》2024年第12期55-59,67,共6页Industrial Heating

基　　金：国家电网有限公司总部管理科技项目(5500-202220136A-1-1-ZN)。

摘　　要：针对目前复杂环境下变压器温升仿真效率和精度较低等问题,提出了一种基于传感数据和机器学习的变压器温升虚拟仿真方法。建立了以负载转移系数、负载持续时间、负载率等反映负载变化的瞬态特征量。提出了改进的人工鱼群(IAFSA)优化SVR参数方法描述变压器HST与瞬态特征量之间复杂的多维非线性关系。实验阶段,以500 kV配电变压器作为试验对象,对不同负荷条件和风速下的变压器温升试验和温升仿真进行分析。结果表明,所提方法可有效拟合变压器HST,SSE、MAD、MAPE、MSPE、MSE等指标综合性能明显优于GS-SVR算法。实验论证所提算法的有效性和实用性,该模型具有广阔的应用前景。A virtual simulation method for transformer temperature rise based on sensor data and machine learning is proposed to address the low efficiency and accuracy of transformer temperature rise simulation in complex environments:Established transient characteristic variables that reflect load changes,such as load transfer coefficient,load duration,and load rate.An improved artificial fish swarm(IAFSA)optimization SVR parameter method was proposed to describe the complex multidimensional nonlinear relationship between transformer HST and transient characteristic variables.In the experimental stage,the oil immersed 500 kV distribution transformer was used as the test object to analyze the temperature rise test and simulation of the transformer under different load conditions and wind speeds.The results show that the proposed method can effectively fit the transformer HST,and the comprehensive performance of SSE,MAD,MAPE,MSPE,MSE and other indicators is significantly better than the GS-SVR algorithm.The effectiveness and practicality of the proposed algorithm are experimentally demonstrated,and the model has broad application prospects.

关 键 词：变压器 热点温度 温升仿真 支持向量回归 人工鱼群优化算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]