Short-term Metallurgical Load Forecasting Based on Adaptive Ensemble Learning

Abstract Accuracy and rapidity are the primary objectives of load forecasting, also necessary conditions for ensuring power supply production schedule. However, in actual production, due to variability operating modes interference environment, difficulties such as non-stable high fluctuation prediction. In light this, we propose an adaptive hybrid prediction model based on Discrete Wavelet decomposition(DWT). It is well known that DWT can show local features mutation sequence, has excellent multi-scale analysis ability. Adopt energy entropy evaluate aggregation wavelet coefficients order obtain subsequence with highest degree preserving main frequency original signal. Random Forest (RF) Long Short-Term Memory (LSTM) were used track low-frequency profile high-frequency detail fluctuations respectively. Further, parameters heterogeneou models optimized using Particle Swarm Optimization(PSO) improve adaptability different components. Compared other metallurgical forecasting techniques, effectiveness superiority proposed verified by experiments industrial data set electrical smelting furnaces magnesia.