SP-Transformer: A Medium- and Long-Term Photovoltaic Power Forecasting Model Integrating Multi-Source Spatiotemporal Features

<p>Photovoltaic (PV) power data are a valuable but as yet under-utilised resource that could be used to characterise global irradiance with unprecedented spatio-temporal resolution. The resulting knowledge of atmospheric conditions can then be fed back into weather models and will ultimately serve to improve forecasts of PV power itself. This provides a data-driven alternative to statistical methods that use post-processing to overcome inconsistencies between ground-based irradiance measurements and the corresponding predictions of regional weather models (see for instance Frank et al., 2018). This work reports first results from an algorithm developed to infer global horizontal irradiance as well as atmospheric optical properties such as aerosol or cloud optical depth from PV power measurements.</p><p>Building on previous work (Buchmann, 2018), an improved forward model of PV power as a function of atmospheric conditions was developed. As part of the BMWi-funded project MetPVNet, PV power data from twenty systems in the Allgäu region were made available, and the corresponding irradiance, temperature and wind speed were measured during two measurement campaigns in autumn 2018 and summer 2019. System calibration was performed using all available clear sky days; the corresponding irradiance was simulated using libRadtran (Emde et al., 2016). Particular attention was paid to describing the dynamic variations in PV module temperature in order to correctly take into account the heat capacity of the solar panels.</p><p>PV power data from the calibrated systems were then used together with both the DISORT and MYSTIC radiative transfer codes (Emde et al., 2016) to infer aerosol optical depth, cloud optical depth and irradiance under all sky conditions.  The results were compared to predictions from the COSMO weather model, and the accuracy of the inverted quantities was compared using both a simple and more complex forward model. The potential of the method to extract irradiance data over a larger area as well as the increase in information from combining neighbouring PV systems will be explored in future work.</p><p><strong>References</strong><br>  <br>Buchmann, T., 2018: Potenzial von Photovoltaikanlagen zur Ableitung raum-zeitlich hoch aufgelöster Globalstrahlungsdaten. Heidelberg University, http://archiv.ub.uni-heidelberg.de/volltextserver/24687/.<br>Emde, C., and Coauthors, 2016: The libRadtran software package for radiative transfer calculations (version 2.0.1). <em>Geosci. Model Dev.</em>, 9, 1647–1672, doi:10.5194/gmd-9-1647-2016. https://www.geosci-model-dev.net/9/1647/2016/.<br>Frank, C. W., S. Wahl, J. D. Keller, B. Pospichal, A. Hense, and S. Crewell, 2018: Bias correction of a novel European reanalysis data set for solar energy applications.<em> Sol. Energy</em>, 164, 12–24, doi:10.1016/j.solener.2018.02.012. https://doi.org/10.1016/j.solener.2018.02.012.</p>

Novel data-driven energy management of a hybrid photovoltaic-reverse osmosis desalination system using deep reinforcement learning

This paper proposes a multi-data driven hybrid learning method for weekly photovoltaic (PV) power scenario forecast that is coordinately driven by weather forecasts and historical PV power output data. Patterns of historical data and weather forecast information are simultaneously captured to ensure the quality of the generated scenarios. By combining bicubic interpolation and bidirectional long-short term memory (BiLSTM), a super resolution algorithm is first presented to enhance the time resolution of weather forecast data from three hours to one hour and increase the precision of weather forecasting. A weather process-based weekly PV power classification strategy is proposed to capture the coupling relationships between meteorological elements, continuous weather changes and weekly PV power. A gated recurrent unit (GRU)-convolutional neural network (CNN)-based scenario forecast method is developed to generate weekly PV power scenarios. Evaluation indices are presented to comprehensively assess the quality of the generated weekly scenarios of PV power. Finally, the PV power, weather observation and weather forecast data collected from five PV plants located in Northeast Asia are used to verify the effectiveness and correctness of the proposed method.

Day-ahead forecasting of photovoltaic output power with similar cloud space fusion based on incomplete historical data mining

Accurate photovoltaic (PV) power prediction is important for the utilization of solar energy resources. However, PV power is non-stationary due to the variable influence of meteorological factors, which poses a challenge for accurate forecasting. In this paper, a hybrid method based on signal decomposition and a deep learning model is proposed. The hybrid model integrates complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and the Informer model. The CEEMDAN algorithm is used to separate different modes from the photovoltaic power sequence, enhancing its predictability. The deep learning model, the Informer, is employed to capture the complex relationship between photovoltaic power data and its historical data as well as external meteorological factors, ultimately enabling multi-step forecasting of photovoltaic power data. In hourly PV power forecasting experiments using a public dataset, the model exhibits significant performance improvements when compared to benchmark models such as LSTM, GRU, and Transformer. Specifically, the RMSE reduces by 6.07%-34.74% and the MAE reduces by 7.07%-37.5%. The results demonstrate that the hybrid model exhibits accurate predictive performance in the task of hourly photovoltaic power forecasting.

With the rapid increase of the total installed capacity of photovoltaic (PV) power generation, more accurate PV power prediction is required to ensure the safe and stable operation of the power grid. In order to improve the prediction accuracy of PV power when only irradiance and PV power data can be obtained, while other multi-source data such as temperature, precipitation and other meteorological data, are unavailable, the paper proposes a PV power prediction model based on data mining and the multi-kernel support vector machine (SVM). Firstly, the wavelet threshold denoising method is used to denoise the data of irradiance and PV power which contains many burrs and the large signal fluctuation. Then, the parameters are extracted by irradiance and power characteristic representation, which include six irradiance characteristic parameters and two power characteristic parameters. With the characteristic parameters, the similar days are selected by the data mining technology, a clustering algorithm using SOM and K-Means. Finally, the multi-kernel SVM is used for PV power prediction, where the multi-kernel function is used to deal with the distribution characteristics of data and improve the accuracy of PV power prediction. The experimental results show that the prediction accuracy can be improved by the wavelet threshold denoising and multi-kernel SVM. The high precision PV prediction results can also be obtained with the irradiance and PV power data only, and the PV prediction accuracy of multi-kernel SVM is higher than that of the single-kernel SVM and classical back propagation (BP) neural network.

Aiming at the problem of abnormal data cleaning in photovoltaic(PV) power generation systems, this paper considers the data variation characteristics in PV power data and its strong correlation with the irradiance data. The paper proposes an abnormal data cleaning method based on variance change points and correlation analysis. First, the abnormal data in the PV power time series is classified into single-point outliers and continuous outliers. Then the median absolute deviation(MAD) is used to detect the variance change point of the PV power data and the abnormal values are determined through correlation analysis. Further the linear regression model of PV power and irradiance is used to repair the abnormal values. Finally, the abnormal data cleaning method is tested on the PV power data from a real power station. The results show that this method can effectively clean the abnormal PV power data.

With the increasing global warming and enormous pollution, it is very obvious to generate power from renewable energy sources. In fact, photo voltaic (PV) power is a main renewable energy source for sustainable power generation. But PV power generation is uncertain and intermittent in nature. Stable and reliable power supply may not be feasible with PV power. Power supply reliability is important as much as its availability. To handle the inconsistency of PV power generation, power sector is highly dependent on forecasting methods and techniques. In this paper the feed forward neural network (FFNN) of artificial neural networks(ANN) with optimization technique particle swarm optimization (PSO) and support vector regression (SVR) are used in short term PV power generation forecasting and their performance is compared with the error calculation in terms of the mean absolute error (MAE) and root mean squared error (RMSE). The PV power and meteorological data of solar irradiation and temperature of Kolkata region of India was used for forecasting. Similar days of the year sunny days, cloudy days and rainy days are clustered using k-means clustering technique. Thereafter, a feed forward neural network is implemented to forecast a day head PV power and week ahead photo voltaic power. PSO is used to optimize the weights of the neural network. The performance of the proposed forecasting method is compared to a data mining approach, support vector regression (SVR).In this paper, the significance of proper selection of input parameters in PV power forecasting is emphasized. With the application of K-means clustering, the accuracy of ANN-PSO approach is improved significantly in a day ahead PV power forecasting and a week ahead PV power forecasting with a good margin. Even the accuracy of ANN-PSO approach after K-means clustering is better than SVR model in a week ahead forecasting

Ultra-short-term photovoltaic power prediction based on similar day clustering and temporal convolutional network with bidirectional long short-term memory model: A case study using DKASC data

Distributed photovoltaic power forecasting based on personalized federated adversarial learning

A Multi-step ahead photovoltaic power forecasting model based on TimeGAN, Soft DTW-based K-medoids clustering, and a CNN-GRU hybrid neural network

Photovoltaic (PV) power is greatly uncertain due to the random meteorological parameters. Therefore, accurate PV power forecasting results are significant for the dispatching of power and improving of system stability. This paper proposes a hybrid forecasting model for one-day-ahead PV power forecasting under different cloud amount conditions. The proposed model consists of an improved artificial neural network (ANN) algorithm and a PV power conversion model. First, the ANN model is designed to forecast the plane of array (POA) irradiance and ambient temperature. Backpropagation, gradient descent, and L2 regularization methods are applied in the structure of the ANN model to achieve the best weights, improve the prediction accuracy, and alleviate the effect of overfitting. Second, the PV power conversion model employs the forecasted results of POA irradiance and ambient temperature to determine the PV power produced by a PV module. In addition to the basic temperature factor, environmental efficiency and a reflection efficiency are incorporated into the conversion model to account for real PV module losses. The performance of the proposed model is validated with real weather and PV power data from Alice Springs and Climate Data Store. Results indicate that the model improves the forecast accuracy compared to four benchmark models. Specifically, it reduces root mean square error (RMSE) and normalized RMSE (nRMSE) by up to 25% under cloudy conditions and offers a 3% shorter training time compared to extreme gradient boosting.

With the increasing proportion of solar grid-connected, the establishment of an accurate photovoltaic (PV) power prediction model is very important for safe operation and efficient dispatching of a power grid. Considering the multi-level periodicity of PV power caused by many factors, such as seasons and weather, a short-term PV power prediction model based on transfer component analysis is designed by introducing the idea of transfer learning. In order to measure the uncertainty of numerical weather prediction (NWP) and power sequence, a novel algorithm considering weather similarity and power trend similarity is proposed. First, the intrinsic trend is measured by extracting permutation entropy, variance, and mean from the historical PV power sequence. Second, weighting of NWP is accomplished based on the Pearson correlation coefficient. PV power data are divided into different clusters by K-medoids clustering. At the same time, the transfer component analysis alleviates the time-varying problem of data distribution caused by multi-level time periodicity and effectively improves the prediction accuracy of the model. Finally, simulation experiments are carried out on the PV power output dataset (PVOD). The results show that the prediction accuracy of the proposed method is better than the traditional methods, and the accuracy and applicability of the proposed method are verified.

Photovoltaic (PV) power generation, with its volatile, intermittent, and random characteristics, and large-scale PV access pose a threat to grid stability. For this reason, predicting the photovoltaic output will help keep the grid safe and stable. On the basis of the influence of cloud groups on solar radiation, a very short-term forecast of distributed PV energy will be made using satellite cloud picture information to improve the forecast accuracy of PV energy production. The paper presents a method to predict distributed PV power at very short notice based on satellite clouds and a network model with Long Short-Term Memory (LSTM). First, extract a subset of meteorological and PV power data from the forecast area as training samples., and the abnormal part of the samples is cleaned by an isolated forest algorithm. Secondly, the occlusion feature is extracted from the satellite cloud image in the same period. Finally, the measured solar irradiance, meteorological information, and obscuration features are input into the LSTM network for prediction, and the photovoltaic power prediction results in the next 4 hours are obtained. The measured PV power of Jinghai Guangfu Power Station in Hefei, Anhui province on the 5th day was the training sample for the prediction of PV power on the 6th day. The prediction results show that the prediction error is 2.73% when a satellite cloud image is added, and 16.15% when a satellite cloud image is not added, and the prediction error is reduced by 13.42%.

Accurate photovoltaic (PV) power prediction plays an increasingly crucial role to maintain the safety and reliability of power grid operation. However, the fluctuation and non-stationarity of PV power make it a challenging task to optimize the accurate results. This paper presents a novel prediction model which is the combination of Hodrick–Prescott (HP) filter, optimized variational mode decomposition (OVMD), and enhanced emotional neural network (EENN). It overcomes the adverse effects of random changes under highly volatile weather conditions. First, the trend component and fluctuation component of PV power are screened through HP filter as the pre-step to alleviate the non-linearity impact of PV power data. Then, OVMD is used to decompose the residual PV power time series into a series of relatively stationary intrinsic modes. Finally, the EENN model optimized by the grey wolf optimization (GWO) is established to predict each subseries, and the prediction results of each subseries are reconstructed to obtain the final predicted results. The numerical results based on actual PV power data show that the prediction accuracy of the proposed model is significantly improved compared with the contrast models, and the proposed model achieves the best accuracy against the OVMD-GWO-EENN, VMD-GWO-EENN, and GWO-EENN models.