Multivariable space-time correction for wind speed in numerical weather prediction (NWP) based on ConvLSTM and the prediction of probability interval

With the advent of the low-carbon era, wind power has become an indispensable energy source. Accurate day-ahead wind speed forecast is crucial for the power system to absorb wind power. Due to the influence of the spatiotemporal resolution and the error of forecasting itself, there is a certain error between the original wind speed of numerical weather prediction and the actual wind speed. Aiming to minimize this error as much as possible, this paper advocates using multivariable space-time information to jointly correct the wind speed in numerical weather prediction. Firstly, the correlation analysis experiments are carried out to demonstrate the feasibility of the idea. Then, the multivariable space-time experiment based on convolutional long short-term memory network is carried out, which greatly reduced the initial wind speed error in numerical weather prediction. At the same time, various methods are used for comparison. The experimental results show that the proposed method reduces the mean absolute error of the numerical weather prediction by 41.13% similar to 77.70% and reduces the root mean square error of the numerical weather prediction by 37.30% similar to 75.10% in 10 places, which is better than other comparison methods. Finally, to adapt to the regulatory needs of the power system, the probability interval predictions are carried out based on the corrected wind speed by the proposed method. The probability interval coverage probability reaches 0.924 similar to 0.937, while the probability interval averaged width reaches 1.869 similar to 2.198 in 10 places.