EMD-based gray combined forecasting model - Application to long-term forecasting of wind power generation

Wind power prediction based on deep learning models: The case of Adama wind farm

Wind is a renewable energy source that is used to generate electricity. Wind power is one of the suitable solutions for global warming since it is free from pollution, doesn't cause greenhouse effects, and it is a natural source of energy. However, Wind power generation highly depends on weather conditions. It is very difficult to easily predict the amount of power generated from wind at a particular instant in time. Adama wind power farm is one of the wind farms in Ethiopia. There is no accurate and reliable forecasting model for the Adama wind farm that enables the forecasting of the power generated from the farm. The main objective of this research is to develop a wind power forecasting model for the Adama wind farm using deep learning techniques. Forecasting of wind power generation capacity involves appropriate modeling techniques that use past wind power generation data. The experiments have been conducted using Long Short-Term Memory (LSTM), Bidirectional Long Short-Term Memory (Bi-LSTM) and Gated Recurrent Unit (GRU). To achieve the highest forecasting accuracy, four years of data (from 2016 to 2019), with 5-min intervals, have been collected with a total of 163,802 rows. For hyperparameter optimization grid search and random search techniques have been utilized. The performances of the proposed deep learning models were investigated error metrics, including Mean Absolute Errors (MAE) and the Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE) and R squared (R2). Bi-LSTM outperforms the other two algorithms, scoring 0.644, 0.388, 0.769 and 0.978 MAE, MAPE, RMSE and R2 values respectively. Such wind power forecasting helps energy planners and regional power providers to compute power production and energy generated from other sources.

Short-term wind farm cluster power point-interval prediction based on graph spatio-temporal features and S-Stacking combined reconstruction

Wind energy is becoming increasingly competitive, Accurate and reliable multi-engine wind power forecasts can reduce power system operating costs and improve wind power consumption capacity. Existing research on wind power forecasting has neglected the importance of interval forecasting using clusters of wind farms to capture spatial characteristics and the objective selection of forecasting sub-learners, leading to increased uncertainty and risk in system operation. This paper proposes a new "decomposition-aggregation-multi-model parallel prediction" method. The data set is pre-processed by a decomposition-aggregation strategy and spatial feature extraction, and then a Stacking model with multiple parallel sub-learners selected by bootstrap method is used for point and interval forecasting. Experiments and discussions are conducted based on 15-min resolution wind power data from a cluster dataset of a wind farm in northwest China. The experimental results indicate that the method achieves higher accuracy and reliability in both point prediction and interval prediction than other comparative models, with a root mean square error value of 7.47 and an average F value of 1.572, which can provide a reliable reference for power generation planning from wind farm clusters.

Novel MPPT algorithm based on honey bees foraging characteristics for solar power generation systems

This study proposes a novel honey bee dancing (HBD) maximum power point tracking (MPPT) algorithm inspired by the foraging behavior of honey bees. When a bee finds nectar, it returns to the honeycomb and dances to inform others about the location of the nectar. Other bees then fly towards the location and gather the nectar. The proposed HBD algorithm uses five bees searching for the nectar who communicate with each other about the location and the quantity of nectar by dancing. Finally, the five bees found the location of the most nectar which is represented by the maximum power point. The proposed HBD algorithm applies to uniform irradiance condition (UIC) and partial shaded condition (PSC). It is then compared with the PV panel output power and load relationship (OPLR) algorithm and perturb and observe (P&O) algorithm. Experimental verification has been performed under both the UIC and PSC. At the UIC's irradiance level of 200 W/m2, the PV module's output power for the proposed HBD algorithm, OPLR algorithm, and P&O algorithm are 120 W, 118 W, and 94.5 W, respectively, with efficiencies of 99 %, 97 %, and 78 %. Additionally, under the PSC with an irradiance level of 600 W/m2, the PV module's output power for the proposed HBD algorithm, OPLR algorithm, and P&O algorithm are 218 W, 189.2 W, and 74.8 W, respectively, with efficiencies of 99 %, 86 %, and 34 %, and convergence times of 4.7 ms, 6.5 ms, and 6 ms, respectively. It is evident from the results that the solar MPPT performance of the proposed HBD algorithm is better than the OPLR algorithm and P&O algorithm. This method ingeniously combines the foraging behavior of bees with solar power generation to produce the maximum natural power. This approach does not require the development of photovoltaic (PV) panel specification data, complex calculations, and additional temperature meters and heliographs. It is highly efficient and has significant economic benefits.