Short-term power prediction for renewable energy using hybrid graph convolutional network and long short-term memory approach

<p>This research accounts for the outcome of a machine learning algorithm project to predict short term solar and wind power in Ontario. A Long Short Term Memory (LSTM) model was developed to monitor short term power output predictions from 4 hours ahead to 6 hours ahead. This study demonstrates a unique approach of utilizing nearby publicly available meteorological data to predict renewable energy power output that could potentially be used for wind and solar farm scheduling to prevent curtailment of clean energy. The wind power and the solar power was predicted 6 and 4 hours ahead with a coefficient of correlation (R2) of 0.817 and 0.738 respectively. It has demonstrated the usage of a LSTM network as a reliable tool for the prediction of renewable energy that can be implemented to power output on systems that require accurate prediction of wind and solar power within the 4 to 6 hours. </p>

In order to solve the problem of short-term power prediction of photovoltaic power generation, an innovative method based on enhanced Long Short-term memory network (XLSTM) is proposed in this paper. Photovoltaic power generation is affected by weather, light and other factors, showing significant volatility and uncertainty, and traditional forecasting methods are difficult to accurately capture its complex timing change law. XLSTM deeply improves the memory mechanism of traditional LSTM, greatly enhances the ability of the model to capture long-term dependencies and improves the training efficiency. The model is trained by historical PV power generation data and meteorological data and can accurately predict the PV power generation in the short term. The experimental results show that compared with traditional LSTM and other common prediction methods, XLSTM model performs better in both prediction accuracy and calculation performance, providing a solid and reliable basis for intelligent scheduling of photovoltaic power, and helping to improve the stability and efficiency of photovoltaic power system operation.

Improved informer PV power short-term prediction model based on weather typing and AHA-VMD-MPE

To enhance the accuracy of short-term wind power prediction, this paper proposes a novel two-stage forecasting framework that integrates Sequential Variational Mode Decomposition (SVMD), Bayesian Optimization (BO), and a CNN-BiLSTM-Attention model. In the first stage, the preprocessed wind power historical data is decomposed into several modal components via SVMD. These components serve as inputs to the CNN-BiLSTM-Attention model, whose hyperparameters—including the learning rate, number of hidden units, and regularization coefficient—are automatically tuned using the BO algorithm. The output of this stage is the initial power prediction. In the second stage, the prediction error sequence from the first stage is analyzed and similarly processed (decomposed and modeled) to generate an error compensation term. The final prediction is obtained by summing the initial power prediction and the predicted error compensation. Results show that compared with the CNN-BiLSTM-Attention model, the MAE, MAPE and RMSE values of the improved CNN-BiLSTM-Attention two-stage prediction model decreased by 85.7%, 75.2% and 77.3%, respectively, demonstrating the effectiveness of the two-stage short-term wind power prediction method of the improved CNN-BiLSTM-Attention model studied in this paper. 为了提升短期风电功率预测的准确性，本文研究了一种基于逐次变分模态分解（SVMD）方法、贝叶斯优化（BO）算法和CNN-BiLSTM-Attention模型结合的两阶段短期风电功率预测方法。该方法具体原理为：在第一阶段，利用SVMD方法对经过数据预处理后的风电功率数据进行分解，将得到的模态分量作为CNN-BiLSTM-Attention短期功率预测模型的输入，然后引入BO算法对预测模型的学习率、隐含层节点和正则化参数进行调优，利用训练好的模型进行预测，得到初始风电功率预测值；在第二阶段，采用预测模型对误差序列进行误差补偿预测，得到初始误差功率预测数据，将初始预测功率和误差预测功率求和得到最终的风电功率预测结果。结果表明：相比于CNN-BiLSTM-Attention模型，改进CNN-BiLSTM-Attention模型两阶段预测模型的MAE值、MAPE值和RMSE值分别下降了85.7%、75.2%和77.3%，表明了本文研究的改进CNN-BiLSTM-Attention模型的两阶段短期风电功率预测方法的有效性。

Solar power is a clean energy source that has developed quickly with considerable attention. Solar energy is required more accurate predictions, which could be integrated into the power grid. Therefore, this project attempts to improve short-term solar power prediction's accuracy, utilizing the long short-term memory (LSTM) in a deep learning machine. The collected data is acquired from the solar system installed in Kaohsiung city, Taiwan. The historical sequential weather parameter and the collected data from the battery module are utilized as input features for the predicting model. To acquire the optimum performance, hyperparameter optimization is employed to construct the best sequential historical data of the LSTM model. The experiment results are compared with a recurrent neural network (RNN), indicating that the LSTM could predict short-term solar power better.

Traditional photovoltaic (PV) power prediction methods in microgrids are insensitive to weather characteristics, and fluctuations in weather data can lead to insufficient PV power prediction accuracy. Therefore, this paper proposes a short-term PV power prediction method for microgrids based on FCM–QPSO–LSTM, which is used to predict PV power under different seasons and weather types. First, the fuzzy [Formula: see text]-means clustering algorithm (FCM) is used to perform clustering analysis on multi-source weather data, and the initial dataset is clustered into sunny, cloudy, and rainy days. Second, to address the shortcomings of a single long-short-term memory (LSTM) model in terms of prediction accuracy and stability, the quantum particle swarm optimization (QPSO) algorithm is used to optimize the model of the LSTM network, and the QPSO–LSTM prediction model is constructed. By introducing quantum dynamic inertia weights and adaptive learning factors, the QPSO algorithm enhances the global search capability and local search capability, which greatly improves the prediction accuracy and stability of the prediction model. Finally, the average absolute percentage error and root mean square error (RMSE) are applied as evaluation indexes to assess the performance of the prediction model. The simulation results show that the QPSO–LSTM model is better than the single LSTM model and the traditional PSO–LSTM model in terms of the average absolute percentage error and RMSE indicators and is able to make accurate predictions of short-term PV power under different seasons and weather types.

Photovoltaic (PV) power installed capacity of China has been growing rapidly with marked improvements in renewable energy accommodation during recent years. it is necessary to improve the accuracy of power forecasts, therefore the underlying electrical grid can be operated in a cost efficient way. This paper aims to improving the accuracy of short-term PV power predictions. Firstly, Measured power data, satellite-based data and numerical weather prediction data are utilized. The data sets of these sources are preprocessed and fused with machine learning techniques to get the sequence feature information. In particular, support vector machine based on data fusion (SVM-DF) is proposed to run as the main regression model. SVM-DF is an extension of the support vector machine and is capable of learning regression functions in continuous space by identifying structures in the mapping of input to output data. One advantage over Artificial Neural Network (ANN) is its ability to construct non-linear dependencies between the input and output data sets. Compared with predicting the PV power through classical ANN model, the SVM-DF approach acquire a more accurate dataset. Predictions of the higher quality can be achieved by SVM-DF with access to PV measurements and weather forecasts. To study further possible improvements in prediction quality, a study on NWP weather parameters, is performed to evaluate their suitability as input features for PV power forecasting. The results confirm the importance of data fusion which make use of spatio-temporal correlations between stations. Results showed that the SVM-DF model performed better than ANN model.

Wind power output has randomness and volatility, which is easy to cause frequency and tie-line power fluctuation of interconnected power grid, and even lead to tie-line out of limit, affecting grid assessment. For this reason, this paper proposes a short-term tie-line power prediction method based on CNN-LSTM, which inputs historical tie-line power into a convolutional neural network (CNN),extracts the data features, generates the feature map, and inputs them into the long-short term memory network (LSTM) for tie-line power prediction. The proposed method is applied to predict tie-line power of a certain regional power grid. The results indicate that the prediction result of the method presented is close to the real power data and has higher prediction accuracy than the traditional prediction method.

Accurate short-term photovoltaic (PV) power forecasting is of great significance for the safe and stable operation of power system. Spatial information from neighboring PV sites contributes to improving forecasting performance. However, most of the current methods considering the spatial information of neighboring sites indiscriminately use all sites data for modeling, which will lead to information redundancy, resulting in low forecasting accuracy. Therefore, this paper proposes a short-term solar power forecasting method based on optimal graph structure considering surrounding spatio-temporal correlations. Firstly, the neighboring sites data is analyzed from the perspective of geographical and weather factors to select typical PV sites. Secondly, based on the complex network theory, a new index is proposed to evaluate the connectivity of the graph structure, which improves the predictive ability of the Graph Convolutional Network (GCN) model. Finally, considering numerical weather prediction (NWP) data, a hierarchical directed graph structure is constructed to indicate the unidirectional relationship between input samples, which is used as the input of GCN model to mine the spatio-temporal correlation around the targeted site. Through carrying out the case study, the proposed method shows excellent performance in improving accuracy of power forecasting compared with other benchmark methods.

BiLSTM-InceptionV3-Transformer-fully-connected model for short-term wind power forecasting

Short-term multi-hour ahead country-wide wind power prediction for Germany using gated recurrent unit deep learning

A short-term wind power prediction model based on BiLSTM–CNN–WGAN-GP (LCWGAN-GP) is proposed in this paper, aiming at the problems of instability and low prediction accuracy of short-term wind power prediction. Firstly, the original wind energy data are decomposed into subsequences of natural mode functions with different frequencies by using the variational mode decomposition (VMD) algorithm. The VMD algorithm relies on a decision support system for the decomposition of the data into natural mode functions. Once the decomposition is performed, the nonlinear and dynamic behavior are extracted from each natural mode function. Next, the BiLSTM network is chosen as the generation model of the generative adversarial network (WGAN-GP) to obtain the data distribution characteristics of wind power’s output. The convolutional neural network (CNN) is chosen as the discrimination model, and the semi-supervised regression layer is utilized to design the discrimination model to predict wind power. The minimum–maximum game is formed by the BiLSTM and CNN network models to improve the quality of sample generation and further improve the prediction accuracy. Finally, the actual data of a wind farm in Jiuquan City, Gansu Province, China is taken as an example to prove that the proposed method has superior performance compared with other prediction algorithms.

Short-term wind power prediction based on multidimensional data cleaning and feature reconfiguration

Connecting wind power to the power grid has recently become more common. To better manage and use wind power, its strength must be predicted precisely, which is of great safety and economic significance. In this paper, the short-term power prediction of wind power is based on self-adaptive niche particle swarm optimization (NPSO) in a neural net. Improved PSO adopts the rules of classification and elimination of a niche using a self-adaptive nonlinear mutation operator. Compared with the traditional method of maximum gradient, NPSO can skip a local optimal solution and approach the global optimal solution more easily in practice. Compared with the basic PSO, the number of iterations is reduced when the global optimal solution is obtained. The method proposed in this paper is experimentally shown to be capable of efficient prediction and useful for short-term power prediction.

With the continuous increase of solar penetration rate, it has brought challenges to the smooth operation of the power grid. Therefore, to make photovoltaic power generation not affect the smooth operation of the grid, accurate photovoltaic power prediction is required. And short-term forecasting is essential for the deployment of daily power generation plans. In this paper, A short-term photovoltaic power generation forecast method based on K-means++, grey relational analysis (GRA) and support vector regression (SVR) (Hybrid Kmeans-GRA-SVR, HKGSVR) was proposed. The historical power data was clustered through the multi-index K-means++ algorithm. And the similar days and the nearest neighbor similar day of the prediction day were selected by the GRA algorithm. Then, similar days and nearest neighbor similar days were used to train SVR to obtain an accurate photovoltaic power prediction model. Under ideal weather, the average values of MAE, RMSE, and R2 were 0.8101 kW, 0.9608 kW, and 99.66%, respectively. The average computation time was 1.7487 s, which was significantly better than the SVR model. Thus, the demonstrated numerical results verify the effectiveness of the proposed model for short-term PV power prediction.