变电站蓄电池支持向量机建模在线监测研究

Research on Modeling of Support Vector Regression for On-Line Monitoring of Substation Battery

针对广泛应用于变电站储能的蓄电池,提出了一种新的蓄电池内阻在线检测技术。为了在线获得蓄电池性能,采用了支持向量机建模技术的变电站蓄电池性能在线监测方法,该方法综合了交流阴抗法和直流放电法的特点。充放电池实验数据作为最小二乘回归支持向量机的初始建模数据,从而得到基于交流阻抗法的蓄电池内阻模型和剩余容量模型。使用蓄电池运行过程中的核对性放电数据和瞬时放电数据作为回归模型的在线校正数据。基于该方法的变电站蓄电池在线监测系统已在变电站使用了4年以上,故障模拟和实际运行表明该方法能有效辨识已劣化电池以及预测电池性能的变化趋势。

A new battery internal resistance on-line detection method is presented based on DC discharging internal resistance detection and AC impedance detection.To achieve online access to the battery performance,a model of the battery capacity is studied by using least squares support vector regression（LS-SVR） modeling technology based on AC impedance method and DC discharge.In the process of battery DC discharge experiment,battery remained capacity and corresponding battery internal resistance values are obtained by means of AC impedance measurement and DC discharge method.Battery performance model based on LS-SVR algorithm is established.To improve on-line precision of the model,DC discharge experiment data are used as initial training samples,and measurement data collected during the process of checkup discharge and instantaneous discharge are used for on-line correction samples.The integrated method based on LS-SVR has been running for four years in the substation.Actual test results show that the method has achieved high-precision detection of substation battery internal resistance on-line,and effectively eliminates errors caused by component tolerances of detection circuit.