PV power probability density prediction based on the long short-term memory network with quantile regression

Considering the chaotic nature of atmospheric systems and the random behavior of electricity users, it is difficult to precisely forecast the targets in the new energy power system. Using single point values as output predictions inevitably results in irreducible errors, leading to a high level of uncertainty in the forecasting process. I kindly request that you improve and correct any errors in my language and provide a more literary and eloquent expression of the same meaning. Thank you in advance. Diverging from traditional point forecasting models, we present a short-term solar photovoltaic (PV) power probability density forecasting model based on Long Short-Term Memory Quantile Regression (LSTMQR). First, LSTMQR is employed to forecast future PV power across various quantiles. Subsequently, a Gaussian kernel function is integrated with LSTMQR and Kernel Density Estimation (KDE) to provide short-term PV power probability forecasting. Through this approach, a probability density function of future PV power predictions can be obtained. Our power system probability forecasting is outputted in the form of probability distribution, quantiles, and prediction intervals for the predicted targets, which effectively quantify prediction uncertainty. These outputs provide power system decision-makers with comprehensive supply, demand, and transaction information.