Carbon Emission Prediction of Power Industry Based on Particle Swarm Optimization Long and Short Term Memory Neural Network

The power industry is one of the main sources of carbon emissions, and accurately predicting its carbon emission values can effectively promote China’s carbon peak target. This paper uses coal-fired power plants as its research object and suggests a carbon emission prediction model (PSO-LSTM) that uses particle swarm optimization to tune the long- and short-term memory hyperparameters of the LSTM neural network. The PSO technique successfully identifies the global optimal solution for LSTM model hyperparameter optimization, addressing the problem that it is challenging to select the best solution for LSTM hyperparameter, and continuously trains to identify the right hyperparameter and verify it. Experimental analysis was conducted on actual case data, and compared with traditional LSTM neural network prediction models, support vector regression (SVR), logistic regression (REG), and gated recurrent unit (GRU) models. The average absolute percentage error (MAPE) was 7.86%, verifying that this method has better prediction performance. The experiment shows that this carbon emission model has good accuracy and stability.