Estimation of Weibull Distribution Parameters for Wind Speed Characteristics Using Neural Network Algorithm

Abstract

Harvesting the power coming from the wind provides a green and environmentally friendly approach to producing electricity. To facilitate the ongoing advancement in wind energy applications, deep knowledge about wind regime behavior is essential. Wind speed is typically characterized by a statistical distribution, and the two-parameters Weibull distribution has shown its ability to represent wind speeds worldwide. Estimation of Weibull parameters, namely scale (c) and shape (k) parameters, is vital to describe the observed wind speeds data accurately. Yet, it is still a challenging task. Several numerical estimation approaches have been used by researchers to obtain c and k. However, utilizing such methods to characterize wind speeds may lead to unsatisfactory accuracy. Therefore, this study aims to investigate the performance of the metaheuristic optimization algorithm, Neural Network Algorithm (NNA), in obtaining Weibull parameters and comparing its performance with five numerical estimation approaches. In carrying out the study, the wind characteristics of three sites in Saudi Arabia, namely Hafer Al Batin, Riyadh, and Sharurah, are analyzed. Results exhibit that NNA has high accuracy fitting results compared to the numerical estimation methods. The NNA demonstrates its efficiency in optimizing Weibull parameters at all the considered sites with correlations exceeding 98.54.

Wind Energy Assessment Using Weibull Distribution with Different Numerical Estimation Methods: A Case Study

“…This paper is using more methods for Weibull parameters estimation; However, the paper shows that MM method is the best method for wind power density estimation and also the GM is the best method for Qassim region which is supporting the findings. [42] 2023 NNA outperform all the methods, However, by comparing the numerical methods, the MLM is the most accurate method for estimating Weibull parameters compared to other numerical methods while the LSM has the worst performance. This paper uses more methods but does not apply the MLM method or NNA method.…”

A New Dual-Mass MEMS Gyroscope Fault Diagnosis Platform

“…Sixty-seven sets of signals were generated, of which forty sets were used to train the model and twenty-seven sets were used to test the accuracy of the model. Feature extraction was performed for seven different signals, and classification was performed using six algorithms-ELM [26], SVM [27], KNN [28], NB [29], NN [30], and DTA [31]-and the recognition rate of the classification results is shown in Figure 8, from which we can see that the accuracy of classification recognition using the ELM algorithm was 92.86%, which was able to identify 100% of bias, blocking, drift, period, and internal fault signals, respectively. The accuracy of classification recognition using the SVM algorithm was the same as that of ELM algorithm; the accuracy of classification recognition using the KNN algorithm was 89.29%, whereby it was able to identify 100% of bias, blocking, drift, period and internal fault signals, 50% of normal signals and 75% of multiplicative signals.…”