New energy output prediction and demand response optimization based on LSTM-BN

The study proposed a new energy output prediction model based on long and short-term memory network (LSTM)-Bayesian network (BN) by combining the benefits of BN in uncertainty quantification with the processing power of LSTM network to address the issue of volatility and uncertainty of new energy output. Meanwhile, by introducing a price-based demand response mechanism, users were incentivized to increase electricity consumption when the new energy generation was in excess and reduce electricity consumption during the peak period, so as to realize the flexible regulation of loads and the efficient utilization of new energy. The new energy output prediction model developed in the study had the highest degree of match between the anticipated and actual values in various data sets, as demonstrated by the experimental findings, which were above 0.99. In the Google Earth Engin and GEFCom2014 datasets, the operation solution speed was quick and stabilized after 64 and 80 iterations, respectively. Additionally, the model’s predicted and actual curve values almost matched, and the actual new energy output power predication's largest prediction error was less than 1%. The implementation of a price-based demand response approach to control customers' power consumption behavior yielded a net benefit of up to 4.45 million yuan for the customers in the target area, based on the precise prediction of new energy output power. The aforementioned findings demonstrated that the LSTM-BN-based new energy output prediction model is capable of precisely projecting new energy output and efficiently matching supply and demand through a price-based demand response mechanism to increase the rate at which new energy is consumed instantly.