测光模式
计算机科学
变压器
补偿(心理学)
电子工程
光伏系统
控制理论(社会学)
标准差
实时计算
工程类
电压
噪音(视频)
能量(信号处理)
电力负荷
电力
实时数字模拟器
卡尔曼滤波器
高斯噪声
电阻式触摸屏
功率(物理)
近似误差
数据建模
电力传输
电气工程
偏移量(计算机科学)
噪声测量
观测误差
作者
Chong Li,Hongtao Shen,Juchuan Guo,Hongming Ma,Yi Wang,Rongkun Guo
摘要
The increasing penetration of photovoltaic-storage microgrids causes time-varying electric energy metering deviations due to intermittent power fluctuations, bidirectional power flow, and harmonic disturbances. This study proposes a PatchTST-based denoising metering compensation network (PDMC-Net) for online deviation compensation. A patch-wise transformer encoder captures temporal features, a regression head predicts the subsequent five-minute deviation, and a denoising reconstruction branch regularizes the shared representation against measurement noise and packet loss. The estimated deviation is generated from real-time measurable signals and integrated into a software-based compensation layer, while periodic reference measurements are only used for drift monitoring and model updating. Furthermore, a sliding-window adaptation strategy with forgetting and parameter anchoring tracks gradual metering drift. Experiments on hardware-in-the-loop and field photovoltaic-storage datasets demonstrate that PDMC-Net achieves an mean absolute error of 0.0163 kWh and reduces cumulative field metering error from 1.95% to 0.36% with an inference latency of 8.6 ms.
科研通智能强力驱动
Strongly Powered by AbleSci AI