Intelligent compensation method for measurement errors in optical fiber current sensor caused by temperature variation based on the Levy-Weighted-QPSO-NN algorithm

Temperature variations significantly degrade the measurement accuracy of fiber optic current sensors (FOCS) in critical power systems applications such as high-voltage transmission and renewable energy integration. To address this, we propose an intelligent error compensation method based on an improved Quantum-behaved Particle Swarm Optimization-Neural Network (Levy-Weighted-QPSO-NN) algorithm. The approach leverages easily measurable state parameters—sensing ring temperature, received optical power, half-wave voltage, SLD temperature, and SLD current—as inputs to predict temperature-induced current ratio difference. Experimental validation involved three sensing rings subjected to temperature cycling (−45 °C to 70 °C), emulating harsh substation environments. The Levy-Weighted-QPSO-NN model achieved 91.11% average prediction accuracy for ratio difference with a correlation coefficient (R²) of 0.9223, outperforming QPSO-NN (85.69%) and Weighted-QPSO-NN (88.31%). Key metrics (MAE: 0.0784; RMSE: 0.0819) confirmed superior stability and accuracy. Robustness testing demonstrated consistent performance across varying population sizes (25–70) and iterations (90–150). Using predicted ratio differences for real-time compensation reduced measurement errors from 0.82% to 0.13%, meeting IEC 61869–6/8 and GB/T standards for Class 0.2S accuracy. This method eliminates reliance on complex hardware modifications, offering a generic, algorithm-driven solution for temperature-dependent FOCS errors.