Output Power Prediction of Thermal Power Plants Based on Optimized Random Forest Algorithm

Based on the operation data of a power plant with a rated power of 660MW unit in one week in January 2024, this paper proposes a prediction model of the random forest algorithm (AM-RF) with the introduction of the attention mechanism. The input variables of the model are filtered by calculating the Pearson correlation coefficient and the attention weighting coefficient of each operating parameter, and the model is calibrated using the test data after the model training is completed, and the prediction results are finally obtained. By analyzing the prediction results and comparing with other prediction models, it is concluded that the MAE, RMSE, R2 and computation time of the AM-RF model are better than those of other single models, which proves that the AM-RF soft measurement model has a better prediction effect for the forecasting of the unit output power, and the random forest algorithm with the addition of the attention mechanism can effectively improve the forecasting accuracy of the unit output power.