In general, reliable PV generation prediction is required to increase complete control quality and avoid potential damage. Accurate forecasting of direct solar radiation trends in PV power production could limit the influence of uncertainties on photovoltaics, enhance organizational dependability, and maximize the utilization factor of the PV systems for something such as an energy management system (EMS) of microgrids. This paper proposes an intelligent prediction of energy production level in large PV plants through AUTO-encoder-based Neural-Network (AUTO-NN) with Restricted Boltzmann feature extraction. Here, the solar energy output may be projected using prior sun illumination and meteorological data. The feature selection and prediction modules use an AUTO encoder-based Neural Network to improve the process of energy prediction (AUTO-NN). Restricted Boltzmann Machines (RBM) can be used during a set of regulations for development-based feature extraction. The proposed model’s result is evaluated using various constraints. As a result, the proposed AUTO-NN achieved 58.72% of RMSE (Root Mean Square Error), 62.72% of n-RMSE (Normalized Root Mean Square Error), 48.04% of Max-AE (Maximum Absolute Error), 48.66% of (Mean Absolute Error), and 46.76% of (Mean Absolute Percentage Error).

https://www.mdpi.com/1999-5903/15/2/46/pdf?version=1676389463

Prediction of Energy Production Level in Large PV Plants through AUTO-Encoder Based Neural-Network (AUTO-NN) with Restricted Boltzmann Feature Extraction

Hybrid deep CNN-SVR algorithm for solar radiation prediction problems in Queensland, Australia

Accurate solar radiation forecasting is essential to operate power systems safely under high shares of photovoltaic generation. This paper compares the performance of several machine learning algorithms for solar radiation forecasting using endogenous and exogenous inputs and proposes an ensemble feature selection method to choose not only the most related input parameters but also their past observations values. The machine learning algorithms used are: Support Vector Regression (SVR), Extreme Gradient Boosting (XGBT), Categorical Boosting (CatBoost) and Voting-Average (VOA), which integrates SVR, XGBT and CatBoost. The proposed ensemble feature selection is based on Pearson coefficient, random forest, mutual information and relief. Prediction accuracy is evaluated based on several metrics using a real database from Salvador, Brazil. Different prediction time-horizons are considered: 1 h, 2 h and 3 h ahead. Numerical results demonstrate that the proposed ensemble feature selection approach improves forecasting accuracy and that VOA performs better than the other algorithms in all prediction time horizons.

/pdf/solar-radiation-forecasting-using-machine-learning-and-1if2eabs.pdf

Solar Radiation Forecasting Using Machine Learning and Ensemble Feature Selection

/pdf/solar-power-time-series-forecasting-utilising-wavelet-16o2jfhx.pdf

Solar power time series forecasting utilising wavelet coefficients

The COVID-19 pandemic with subsequent economic fluctuations during consecutive epidemics waves and preventive measures in the form of lockdowns, and Russia’s invasion of Ukraine have had a meaningful impact on the European economy, including the energy market. These events have caused an increase in the prices of many products, including fossil fuels, and also a lack of their availability. The changes inspired the authors to conduct research on the current situation in Poland in the field of renewable energy and coal. The paper describes the research on the impact of the development of RES (mainly photovoltaic sources) on the change in the energy generation profile in the case of Poland. The article analyzes the impact of RES (renewable energy sources) on the reduction of CO2 emissions. The conducted analysis was based on the historical similarity of the replacement of conventional (carbon) sources by PV (photovoltaics) by determining in how many cases (%), a specific carbon source will be displaced. Based on the prepared forecast, it was determined that by the end of the year, the installed capacity in PV will reach the level of 11,213 MWp (in wind energy, it will be 7875 MWp). Replacement (reduction of consumption) of approximately 1.5 million Mg of hard coal and 2.87 million Mg of lignite was concluded to be possible (~4.4% and 6.3% of the hard-coal and lignite volume used for energy production). Limiting this volume of hard coal and lignite will also reduce CO2 emissions in the entire NPS by 5.24 million Mg (−5%) in 2022. 

https://www.mdpi.com/2076-3417/12/21/11064/pdf?version=1667291518

Iñigo Azcárate

Papers

Forecasting intra-hour solar photovoltaic energy by assembling wavelet based time-frequency analysis with deep learning neural networks