Başak Öztürk

Recently, wind speed (WS) prediction has attracted the interest of clean energy researchers. As WS is inherently uncontrollable, it can cause imbalances between consumption and wind energy production. Therefore, high-accuracy WS estimation is essential. In this study, the LSTM and Prophet models were used to predict future trends based on past WS values obtained from the Nevşehir Meteorology Directorate wind measurement station. Model performances were compared using error metrics. The Prophet model achieved an R² value of 0.61, while the LSTM model outperformed it with an R² of 0.734, indicating that the LSTM model's forecasts correspond more closely to actual WS data.

One of the most used artificial intelligence techniques for maximum power point tracking is artificial neural networks. In order to achieve successful results in maximum power point tracking, the training process of artificial neural networks is important. Metaheuristic algorithms are used extensively in the literature for neural network training. An important group of metaheuristic algorithms is swarm-intelligent-based optimization algorithms. In this study, feed-forward neural network training is carried out for maximum power point tracking by using 13 swarm-intelligent-based optimization algorithms. These algorithms are artificial bee colony, butterfly optimization, cuckoo search, chicken swarm optimization, dragonfly algorithm, firefly algorithm, grasshopper optimization algorithm, krill herd algorithm, particle swarm optimization, salp swarm algorithm, selfish herd optimizer, tunicate swarm algorithm, and tuna swarm optimization. Mean squared error is used as the error metric, and the performances of the algorithms in different network structures are evaluated. Considering the results, a success ranking score is obtained for each algorithm. The three most successful algorithms in both training and testing processes are the firefly algorithm, selfish herd optimizer, and grasshopper optimization algorithm, respectively. The training error values obtained with these algorithms are 4.5 × 10⁻⁴, 1.6 × 10⁻³, and 2.3 × 10⁻³, respectively. The test error values are 4.6 × 10⁻⁴, 1.6 × 10⁻³, and 2.4 × 10⁻³, respectively. With these algorithms, effective results have been achieved in a low number of evaluations. In addition to these three algorithms, other algorithms have also achieved mostly acceptable results. This shows that the related algorithms are generally successful ANFIS training algorithms for maximum power point tracking.

Artificial neural networks have emerged as a promising tool for estimating hydrogen production process variables for reaction condition optimization. Here we aim to predict complex nonlinear systems that use of artificial neural networks for modeling hydrogen production via water electrolysis and to evaluate the common challenges that arise. To estimate the effect of different electrolyzer systems input parameters such as electrolyte material, electrolyte type, supplied power (voltage and current), temperature, and time on hydrogen production, a predictive model was developed. The percentage contributions of the input parameters to hydrogen production and the best network architecture to minimize computation time and maximize network accuracy were shown. The results show that the hydrogen production parameters from electrolysis and the predicted safety explosive limit are 7% of the average root mean square error. Furthermore, coefficient of determination value was found 0.93. This predicted value is very close to the observed values. The neural network algorithm developed in this study could be used to make critical decisions in the electrolysis process for parameters affecting hydrogen production.

With the increasing integration of wind power systems into conventional power systems, an accurate wind speed (WS) forecasting technique is essential for the reliable and stable operation of the power grid. The availability of wind energy increases with the correct WS estimation technique. Linear statistical methods such as the Seasonal Auto-Regressive Integrated Moving Average (SARIMA) model have been popular in recent times for short-term and very short-term forecasting of WS. However, studies are continuing the effect of forecast errors resulting from different disaggregated time series on the resulting WS forecast error. In this paper, we present a state-of-the-art machine learning model, the Prophet model, which has not been used in short-term wind forecasting before in the literature. The performance reliability and accuracy of the developed Prophet model’s results were compared to those of the SARIMA model. The success of forecasting time-series data was evaluated using Mean Absolute Error (MAE) and Mean Square Error (RMSE) results. The models’ reliability, accuracy, fit, and performance were evaluated using standardized residuals, autocorrelation function (ACF), and partial autocorrelation (PACF). The 7-day performance metrics of the Prophet model are 1.9, 2.5, 4.1, 3.8, 4.0, 8.4, and 2.1 for MAE, and 2.3, 2.9, 4.7, 5.0, 4.4, 8.6, and 2.5 for RMSE, respectively. The performance measures of the SARIMA model are 3.3, 4.1, 5.1, 4.8, 5.1, 1.8, and 5.4 for the MAE, and 4.2, 5.1, 6.2, 6.2, 5.7, 2.6, and 6.2 for the RMSE, respectively. The Prophet model gave higher accuracy than SARIMA for all other days except 1 day (06.01.2018). As a result, the Prophet model outperformed the SARIMA model in predicting WS measurements 24 h ahead using quarterly training data.