Search

Accurate implementation of short-term wind speed prediction can not only improve the efficiency of wind power generation, but also relieve the pressure on the power system and improve the stability of the grid. As is known to all, the existing wind speed prediction systems can improve the performance of the prediction in some sense, but at the same time they have some inherent shortcomings, just like forecasting accuracy is not high or indicators are difficult to obtain. In this paper, based on 10-min wind speed data from a wind farm, a new combination model is developed, which consists of three parts: data noise reduction techniques, five artificial single-model prediction algorithms, and multi-objective optimization algorithms. Through detailed and complete experiments and tests, the results demonstrate that the combination model has better performance than other models, solving the problem of instability of traditional forecasting models and filling the gap of low-prediction short-term wind speed forecasting. [ABSTRACT FROM AUTHOR]

Water quality forecasting has great practical significance for sustainable utilization of water resources and timely pollution prevention and control. However, owing to irregularity and volatility of water quality data, achieving accurate forecasts remains a challenging problem. Existing single forecasting models based on points forecasts fail to capture the uncertainty in water quality data or obtain high forecasting accuracy. To address those disadvantages, an uncertain combined water quality forecasting system based on time‐varying filtering empirical mode decomposition strategy and multiobjective immune selection optimization is proposed. Simulations are carried out with actual water quality data obtained during dry and wet periods. Empirical results show that the improved empirical mode decomposition strategy has better data denoising effect. Based on the multiobjective optimization algorithm, not only the Pareto optimal weights of the combined forecasting modules can be obtained but also the distribution function parameters can be effectively optimized. The proposed system can integrate the respective advantages of deep learning and neural network models to improve the forecasting accuracy, and effectively analyze the uncertainty of water quality, and provide technical support for the sustainable development of regional water environment. [ABSTRACT FROM AUTHOR]