This paper reviews methods to determine the number of hidden neurons in neural networks over the past 20 years and proposes a new method for fixing hidden neurons in Elman networks for wind speed prediction in renewable energy systems. The random selection of hidden neurons can lead to overfitting or underfitting. The paper tests 101 criteria based on statistical errors to determine the optimal number of hidden neurons, showing that the proposed model improves accuracy and minimizes error. The design of the neural network is substantiated using convergence theorem. Simulations on real-time wind data show that the proposed approach can be used for wind speed prediction with minimal error. The survey highlights the challenges in determining the number of hidden neurons, with most researchers using trial-and-error methods. The proposed method is simple, efficient, and effective for fixing hidden neurons in Elman networks. The paper also discusses various existing methods for determining the number of hidden neurons, including constructive and pruning approaches. The proposed method uses a new criterion based on input parameters to determine the number of hidden neurons, resulting in better performance with minimal error. The results show that the proposed model outperforms other methods in terms of statistical errors. The paper concludes that the proposed method provides a better framework for designing Elman networks for wind speed prediction in renewable energy systems, improving stability and accuracy.This paper reviews methods to determine the number of hidden neurons in neural networks over the past 20 years and proposes a new method for fixing hidden neurons in Elman networks for wind speed prediction in renewable energy systems. The random selection of hidden neurons can lead to overfitting or underfitting. The paper tests 101 criteria based on statistical errors to determine the optimal number of hidden neurons, showing that the proposed model improves accuracy and minimizes error. The design of the neural network is substantiated using convergence theorem. Simulations on real-time wind data show that the proposed approach can be used for wind speed prediction with minimal error. The survey highlights the challenges in determining the number of hidden neurons, with most researchers using trial-and-error methods. The proposed method is simple, efficient, and effective for fixing hidden neurons in Elman networks. The paper also discusses various existing methods for determining the number of hidden neurons, including constructive and pruning approaches. The proposed method uses a new criterion based on input parameters to determine the number of hidden neurons, resulting in better performance with minimal error. The results show that the proposed model outperforms other methods in terms of statistical errors. The paper concludes that the proposed method provides a better framework for designing Elman networks for wind speed prediction in renewable energy systems, improving stability and accuracy.