Application research of wind power prediction method based on improved seagull algorithm

As the global energy structure undergoes a transition toward cleaner sources, the increase in the proportion of wind power poses severe challenges to the stable operation of the power grid due to its random and fluctuating output. Traditional prediction models tend to get stuck in local optima and have insufficient generalization ability, making them unable to meet the high-precision requirements for wind power prediction. Therefore, this study proposes an improved seagull optimization algorithm (LSC-SOA) based on the Logistic-Sine-Cosine composite chaotic mapping, which generates a uniformly distributed initial population through the Logistic-Sine-Cosine chaotic mapping to enhance the global search ability. It is combined with the backpropagation neural network (BP) to construct the LSC-SOA-BP wind power prediction model, which is used to optimize the initial weights and thresholds of the network. Experimental results show that the improved algorithm converges faster and has lower fitness; the built model has the lowest RMSE of 0.11 on the test set, MAE of 0.08, R 2 of 0.97, and an annual RMSE of 0.115 in seasonal and variable wind speed scenarios, which is at least 30.3% lower than the comparison model. This study provides a high-precision solution for the uncertainty of wind power prediction and has practical value for ensuring the security of the power system and promoting the integration of wind power.