Dynamic optimization method for cleaning cycle of condenser of nuclear power plant

Fouling is an important factor affecting the heat transfer performance of condensers in nuclear power plants. In response to the lack of experience in the current cleaning of condensers in nuclear power plants, this paper proposes a particle swarm optimization based LSTM method for predicting fouling thermal resistance and a dynamic optimization theory for condenser cleaning cycles. On the basis of LSTM and the PSO algorithm, compared with the prediction results of LSTM, the MAE, RMSE and MAPE values of PSO-LSTM decreased by 43.3%, 47.3% and 42.1%, respectively, while R2 increased by 26.4%. Considering the load loss and maintenance loss of the unit, as well as the economic loss caused by condenser scaling, the optimal cleaning time of the condenser glue ball system under different fouling rate and circulating water temperature was analyzed, and a calculation theory based on the unit output, the running loss cost of the glue ball and the minimum optimal cleaning cycle of the condenser was proposed. Results show that at rated flow, optimizing the cleaning cycle, with an annual operating time of 300 days, can save $0.45 million in operating costs per year.