Machine learning-based early warning system for abnormal oil levels in hydroelectric power station pressure oil systems

The pressure oil system is essential in hydropower stations. However, traditional methods for monitoring oil levels, which depend on fixed threshold alarms for both pressure oil tanks and return oil tanks, pose safety hazards. This study conducts a comparative analysis of various models, including Partial Least Squares Regression (PLSR), to predict the oil level in the return oil tank. Case studies indicate that the PLS model performs excellently in predicting the oil level in the return oil tank, with a Mean Squared Error (MSE) of 0.071 and a coefficient of determination (R² ) of 0.998, significantly outperforming both linear regression and ordinary least squares regression models. Based on the PLS model, this paper develops an abnormal warning method for the pressure oil system, which can promptly detect system anomalies and ensure the safe and efficient operation of the hydropower station.