Multi-objective optimization model and algorithm implementation of the distributed power generation system for renewable energy in China and Russia

This study focuses on solving multi-objective optimization problems in distributed power generation systems (DPGS) for renewable energy in China and Russia, including low economic efficiency, poor environmental benefits, and insufficient system reliability. It proposes a hybrid optimization model that integrates deep learning with an improved particle swarm optimization algorithm, namely Adaptive Linear Decreasing Inertia Weight Particle Swarm Optimization with Mutation Strategy (ALD-MPSO). By introducing a Dense Bidirectional Long Short-Term Memory with Attention Mechanism (DBI-LSTM-AM) model, which combines a Bidirectional Long Short-Term Memory (Bi-LSTM) network, Dense layers, and an Attention Mechanism (AM), the model performs time-series forecasting of energy demand. Coupled with the ALD-MPSO algorithm, the model simultaneously optimizes economic efficiency, environmental benefits, and system reliability. The study designs a renewable energy prediction and optimization model for DPGS, based on the fusion of the DBI-LSTM-AM and ALD-MPSO algorithms (DBI-LSTM-2AM-PSO). Finally, the model’s performance is evaluated. Experimental results show that the proposed model achieves superior prediction accuracy (95.53%), with an F1 score of 91.41%, and a mean squared error (MSE) of 0.049, outperforming the benchmark algorithms. Additionally, the fitness value in MOO is reduced to 0.47, with a training time of only 25.7 seconds and low computational resource consumption (Center Processing Unit usage at 10.55%). This study provides effective technical support for the intelligent management of DPGS in the renewable energy sectors of China and Russia. • Propose DBI-LSTM-2AM-PSO , a renewable energy prediction and optimization model for DPGS, based on the fusion of DBI-LSTM-AM and ALD-MPSO algorithms. • The model achieves excellent prediction accuracy, optimization effectiveness, and computational efficiency: • Achieve a prediction accuracy of 95.53% and a fitness value of 0.47 in multi-objective optimization. • Reduce training time to 25.7 seconds with low CPU usage ( 10.55% ), outperforming other models. • Sustainable contribution: optimize energy utilization, reduce carbon emissions, and support green energy cooperation between China and Russia.