Wind Energy Modelling and Machine Learning Approach to Study Wind Direction Effect

Wind energy is one of the green renewable energy sources that is available everywhere. The wind-generated electrical energy is much less than the available wind potential. No windmill can even harness 50% of the available wind energy. A lot of research investigations are still needed to explore and harness the maximum energy from wind potential. This work is one of such series of research, in which we modeled wind speed using the Weibull distribution through a Python program that uses the least square method based on Python built-in functions and evaluated the shape and scale parameters of the distribution. The program also compares parameters calculated by other existing methods. The Python program based on the least square method fits the Weibull distribution well compared to the existing methods. The maximum value of scale parameters was found in June (more than 6.2), the corresponding value is also close in May, where it is more than 6.1; the other two months that follow June and May (July and September) have scale parameters near 5.7. It shows that the wind potential is maximum in June, and reasonable wind energy is available in May, July, and September. The effect of wind direction on the modelling of wind speed is also investigated. Perhaps it is the first study that involves wind direction in wind speed modelling. Two different Artificial Neural Network Architectures were studied with and without wind directions in the input. It was found that the results improve if wind direction is also taken in the list of input parameters. The Root Mean Square Error is the least (RMSE = 0.7224) for the model which includes wind direction in the input layer, the performance indicator (0.5219) is also the best for this architecture as compared to the other three.