Short-Term Power Load Forecasting Based on the Quantum Genetic Algorithm Artificial Recurrent Memory Network

Accurate power load forecasting is crucial for maintaining the equilibrium between power supply and demand and for safeguarding the stability of power systems. Through a comprehensive optimization of both the parameters and structure of the traditional load forecasting model, this study developed a short-term power load prediction model (QGA-RMNN) based on a quantum genetic algorithm to optimize an artificial recurrent memory network. The model utilizes the principle of quantum computing to improve the search mechanism of the genetic algorithm. It also combines the memory characteristics of the recurrent neural network, combining the advantages of the maturity and stability of traditional algorithms, as well as the intelligence and efficiency of advanced algorithms, and optimizes the memory, input, and output units of the LSTM network by using the artificial excitation network, thus improving the prediction accuracy. Then, the hyperparameters of the RMNN are optimized using quantum genetics. After that, the proposed prediction model was rigorously validated using case studies employing load datasets from a microgrid and the Elia grid in Belgium, Europe, and was compared and analyzed against the classical LSTM, GA-RBF, GM-BP, and other algorithms. Compared to existing algorithms, the results show that this model demonstrates significant advantages in predictive performance, offering an effective solution for enhancing the accuracy and stability of load forecasting.