Solar Power Forecasting Research Articles

Predictive Control of a Real Residential Heating System with Short-Term Solar Power Forecast

Predictive control has great potential in the home energy management domain. However, such controls need reliable predictions of the system dynamics as well as energy consumption and generation, and the actual implementation in the real system is associated with many challenges. This paper presents the implementation of predictive controls for a heat pump with thermal storage in a real single-family house with a photovoltaic rooftop system. The predictive controls make use of a novel cloud camera-based short-term solar energy prediction and an intraday prediction system that includes additional data sources. In addition, machine learning methods were used to model the dynamics of the heating system and predict loads using extensive measured data. The results of the real and simulated operation will be presented.

Short-term solar power forecasting- An approach using JAYA based recurrent network model

Short-term solar power forecasting- An approach using JAYA based recurrent network model

Univariate and multivariate short-term solar power forecasting of 25MWac Pasir Gudang utility-scale photovoltaic system using LSTM approach

The generation of solar photovoltaic (PV) systems in Malaysia has great potential due to the abundance of sunlight and high irradiation level. Malaysia’s unique location near the equator makes solar energy the most attractive option for future energy sources. However, the erratic weather leads to variability of power generation, especially during high feed-in of solar energy that will eventually affect grid system stability. Hereof, solar power forecasting is critical, especially in operating utility-scale photovoltaic systems or large-scale solar (LSS) plants. This paper presents an approach to forecasting solar power generation for 10-min to 180-min ahead based on univariate and multivariate using Long -Short-Term Memory (LSTM) technique. The model performances are evaluated based on a real dataset from the 25MWac Pasir Gudang LSS plant from July 2021 to May 2022. The result shows that LSTM with a univariate model outperformed the multivariate model for the short-term forecasting (10-min to 50-min ahead) by 2.09% of RMSE. However, multivariate outperformed univariate for the longer forecasting horizon at 180-min ahead by 34.87% of RMSE. The forecasting output for the multivariate model that only uses historical meteorological data is less reliable than the multivariate model that uses historical meteorological and AC power output data. This research finding is envisaged to provide benefits to the grid system operation, planning, maintenance, and scheduling, thereby improving the reliability of the LSS plant.

Feature Transfer and Rapid Adaptation for Few-Shot Solar Power Forecasting

A common dilemma with deep-learning-based solar power forecasting models is their heavy dependence on a large amount of training data. Few-Shot Solar Power Forecasting (FSSPF) has been investigated in this paper, which aims to obtain accurate forecasting models with limited training data. Integrating Transfer Learning and Meta-Learning, approaches of Feature Transfer and Rapid Adaptation (FTRA), have been proposed for FSSPF. Specifically, the adopted model will be divided into Transferable learner and Adaptive learner. Using massive training data from source solar plants, Transferable learner and Adaptive learner will be pre-trained through a Transfer Learning and Meta-Learning algorithm, respectively. Ultimately, the parameters of the Adaptive learner will undergo fine-tuning using the limited training data obtained directly from the target solar plant. Three open solar power forecasting datasets (GEFCom2014) were utilized to conduct 24-h-ahead FSSPF experiments. The results illustrate that the proposed FTRA is able to outperform other FSSPF approaches, under various amounts of training data as well as different deep-learning models. Notably, with only 10-day training data, the proposed FTRA can achieve an RMSR of 8.42%, which will be lower than the 0.5% achieved by the state-of-the-art approaches.

Comprehensive Analysis of Spectral Mismatch Factor for Solar Cells Based on In Situ Observation of Aerosol Optical Depth Spectra and Solar Spectral Irradiance in Korea

The spectral mismatch factor for solar cells quantifies their relative performance in converting solar irradiance between the incident and reference solar spectra into electricity. This study attempted to evaluate the spectral mismatch factor for eight types of solar cells and investigate their sensitivity to changes in the solar spectral irradiance, which is dependent on the aerosol optical properties in a clear sky. Copper indium gallium diselenide cells have the highest mean value of the spectral mismatch factor, implying that they are less sensitive to changes in the solar spectral irradiance. In contrast, perovskite and amorphous silicon cells are more sensitive to atmospheric conditions, with broader distributions of the spectral mismatch factor values. Additionally, our study found that heterojunction with intrinsic thin-layer cells has the highest substantial efficiency, considering the nameplate efficiency. The spectral mismatch factor decreased with increasing aerosol optical depth at 500 nm and was proportional to the humidity. The effects of aerosol optical properties on the spectral mismatch factor for different solar cells were clarified using clustering analysis and back-trajectory modeling results. In the present study, the aerosol optical depth spectra were found to be more important in determining the spectral mismatch factor than the aerosol optical depth at 500 nm. This study recommends further research on the relationship between the aerosol optical properties and solar spectral irradiance to better predict or estimate the spectral mismatch factor in solar power forecasting.

A Review of State-of-the-Art and Short-Term Forecasting Models for Solar PV Power Generation

Accurately predicting the power produced during solar power generation can greatly reduce the impact of the randomness and volatility of power generation on the stability of the power grid system, which is beneficial for its balanced operation and optimized dispatch and reduces operating costs. Solar PV power generation depends on the weather conditions, such as temperature, relative humidity, rainfall (precipitation), global solar radiation, wind speed, etc., and it is prone to large fluctuations under different weather conditions. Its power generation is characterized by randomness, volatility, and intermittency. Recently, the demand for further investigation into the uncertainty of short-term solar PV power generation prediction and its effective use in many applications in renewable energy sources has increased. In order to improve the predictive accuracy of the output power of solar PV power generation and develop a precise predictive model, the authors used predictive algorithms for the output power of a solar PV power generation system. Moreover, since short-term solar PV power forecasting is an important aspect of optimizing the operation and control of renewable energy systems and electricity markets, this review focuses on the predictive models of solar PV power generation, which can be verified in the daily planning and operation of a smart grid system. In addition, the predictive methods identified in the reviewed literature are classified according to the input data source, and the case studies and examples proposed are analyzed in detail. The contributions, advantages, and disadvantages of the predictive probabilistic methods are compared. Finally, future studies on short-term solar PV power forecasting are proposed.

Sky Image Analysis and Solar Power Forecasting: A Convolutional Neural Network Approach

Recently, the share of renewable sources in the energy mix of production units has been steadily increasing. The unpredictability of renewable sources leads to difficulties in planning, managing and controlling the electric energy system (EES). One of the ways to reduce the negative impact of unpredictable renewable sources is to predict the availability of these energy sources. Short-term forecasting of photovoltaic power plant production is one of the tools that enable greater integration of renewable energy sources into the EES. One way to gather information for the short-term forecast production model is to continuously photograph the hemisphere above the photovoltaic power plant. By processing the data contained within the images, parameters related to the current output power of the observed power plant are obtained. This paper presents a model that utilises a convolutional neural network to analyse images of the hemispherical sky above a power plant to predict the current output power of the power plant. Estimating current production is a crucial step in developing models for short-term solar forecasts. The model was specifically developed for photovoltaic power plants and is capable of achieving high accuracy in power prediction. The estimation of power production from photovoltaic power plants enables the use of next-frame prediction for short-term forecasting.

Deep hybrid neural net (DHN-Net) for minute-level day-ahead solar and wind power forecast in a decarbonized power system

The need to reduce global carbon emissions has led to a significant increase in clean energy globally. While renewable energy penetration into energy grids and power systems is increasing in many countries, the intermittency and stochastic nature of wind and solar energy resources is still a major challenge. These can affect the safety, stability, and reliability of the energy grid. In existing works of literature, the forecast and prediction of wind energy, solar power, wind power, and solar energy with various models have been considered independently. However, with the rise in solar power and wind power penetration, there exists a gap in literature on the development of models that can simultaneously forecast solar and wind power production. In this paper, two deep hybrid neural networks (DHN-Net) models are developed for the simultaneous forecast of wind and solar power. The novelty of this study is further strengthened as a minute-level timestep is considered for the application of the models developed. The models are trained and tested with data collected from Zone 1 of four different power system operators in the USA. The two DHN-Net models are built on the foundation of artificial, convolutional, and recurrent neural networks (ANN, CNN, and RNN). Results from this study show that the two DHN-Nets can accurately forecast solar and wind power with an R-squared (r2) value of 0.9915, RMSE of 0.01920, and MAE of 0.00736 for data collected from PJM_Zone1. The DHN-Net models recorded a better performance when compared to the benchmark results in literature.

Short-term Power Prediction Method of Photovoltaic Based on Output Clustering in Smart Grid

Accurate photovoltaic power prediction can ensure the smart grid's safe and stable operation, as well as reasonable energy scheduling. Weather influences photovoltaic output, which is irregular and unstable, and photovoltaic output is similar under similar meteorological conditions. The paper proposes a solar power forecast model based on Mean-Shift clustering, support vector machine (SVM), and residual neural network (ResNet) in this regard. Firstly, the Mean-Shift algorithm is used to cluster similar days. Then, the SVM model is constructed to learn the similarity between the data of each meteorological type, and the similar day matching is performed on the forecast day. Finally, the short-term photovoltaic output prediction based on ResNet algorithm is carried out for the corresponding weather type. The suggested method's high forecast accuracy and stability are confirmed by experimental examination of a commercial photovoltaic power station.

A taxonomy of short‐term solar power forecasting: Classifications focused on climatic conditions and input data

AbstractA review of the state‐of‐the‐art in short‐term Solar Power Forecasting (SPF) methodologies is presented in this paper. Over the last few years, developing and improving solar forecasting models has been the main focus of researchers, considering the need to efficiently increase their forecasting accuracy. Forecasting models aim to be used as an efficient tool to help with the stability and control of energy systems and electricity markets. Intending to further comprehend the factors affecting the quality of SPF models, this paper focuses on short‐term solar forecasting methodologies since they pose a crucial role in the daily operation and scheduling of power systems, since they focus on forecasting horizons typically ranging from 1 h to 1 day. The reviewed works are classified according to the climatic conditions, technical characteristics, and the forecasting errors of the different methodologies, providing readers with information over various different cases of SPF. Considering the need to improve the SPF efficiency, such classifications allow for important comparative conclusions to be drawn, depending on the location of each case and the meteorological data available. Future directions in the field of short‐term solar power forecasting are proposed considering the increasing development of SPF models’ architecture and their field of focus.

A hybrid data-driven and model-based approach for computationally efficient stochastic unit commitment and economic dispatch under wind and solar uncertainty

Stochastic unit commitment (UC) and economic dispatch (ED) are imperative in dealing with uncertainty in renewable forecast for power system operation and planning such that the overall expected production cost is minimized over the planning horizon. However, accurate calculation of the expected production cost requires assessment of a very large number of different scenarios of uncertain renewable resources, such as solar and wind, which is practically infeasible to simulate in real time. This article proposes a hybrid data-driven and physics-based model-predictive paradigm to efficiently solve for stochastic unit commitment and economic dispatch considering uncertainty in wind and solar power forecasts. The novelty of the approach lies in decoupling the production cost estimation from the unit commitment and economic dispatch optimization problems under uncertainty without compromising on the fidelity of the solutions. A data-driven machine learning model is first developed to predict the mean optimal production cost. A physics-based inverse problem is then solved to get the stochastic UC and ED profiles from the expected cost. The presented approach considers, for the first time, solar uncertainty in UC/ED determination and enables efficient and accurate propagation of wind and solar uncertainty to estimate the statistics of the production cost. The effectiveness of the developed approach is demonstrated systematically on a stylized RTS-GMLC single-node system. The overall framework predicts the expected cost 62.5% more accurately than the existing state-of-the-art, on unforeseen days during the entire year, and yields, for the first time, the associated physically consistent UC and ED profiles. The solutions are also shown to be flexible in providing adequate daily reserves to address any statistical deviations from probabilistic power forecasts. The computational time associated with the presented method is only about 10 s compared to over 24 h needed for a conventional stochastic UC/ED determination under uncertainty on an Intel Core i9 processor with 32 GB of RAM.

Short-Term Solar Power Forecasting Based on CEEMDAN and Kernel Extreme Learning Machine

The use of renewable energy sources contributes to environmental awareness and sustainable development policy. The inexhaustible and nonpolluting nature of solar energy has attracted worldwide attention. Accurate forecasting of solar power is vital for the reliability and stability of power systems. However, the effect of the intermittency nature of solar radiation makes the development of accurate prediction models challenging. This paper presents a hybrid model based on Kernel Extreme Learning Machine (Kernel-ELM) and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) for short-term solar power forecasting. The decomposition technique increases the number of stable, stationary, and regular patterns of the original signals. Each decomposed signal is fed into Kernel-ELM. To validate the performance of the hybrid model, solar power data from the BSEU Renewable Energy Laboratory, measured at 5-minute intervals, are used. To validate the proposed model, its performance is compared to some state-of-the-art forecasting models with seasonal data. The results highlight the good performance of the proposed hybrid model compared to other classical algorithms according to the metrics.

Prediction Interval Estimation and Deterministic Forecasting Model Using Ground-Based Sky Image

Due to the explosive penetration of photovoltaic (PV) systems in the power grid, accurate and well-informed solar PV power forecasting has become exceptionally crucial, while accurate irradiance prediction is the basis of PV power forecasting. Yet most research on PV power/irradiance forecasting focus on deterministic forecasting, whereas it is difficult to express the credibility of the forecasting results, especially in the scene of instantaneous fluctuation of irradiance and photovoltaic power caused by cloud movement. Therefore, in order to accurately estimate and track the rapid fluctuations of solar irradiance, this paper takes the ground-based sky image information into consideration and obtains the probability-level prediction results along with the deterministic point forecasting, proposes a comprehensive evaluation function constructed with the average coverage and the average width as the fitness function to assist in the optimization of the deterministic forecasting results, and then constructs a three-output model, naming it DenLUBE. In addition, a framework for irradiance variation pattern prediction is proposed to provide pattern information for the forecasting time period to further improve the performance of the ultra-short-term forecasting model. The results show that the proposed model can achieve high quality and strongly robust deterministic forecasting results while producing tight prediction intervals independent of weather conditions. In addition, the performance of our model is greatly enhanced compared to the traditional approach of taking the middle position of the interval as the deterministic forecasting result, as well as other benchmark models.

Exploring the PV Power Forecasting at Building Façades Using Gradient Boosting Methods

Solar power forecasting is of high interest in managing any power system based on solar energy. In the case of photovoltaic (PV) systems, and building integrated PV (BIPV) in particular, it may help to better operate the power grid and to manage the power load and storage. Power forecasting directly based on PV time series has some advantages over solar irradiance forecasting first and PV power modeling afterwards. In this paper, the power forecasting for BIPV systems in a vertical façade is studied using machine learning algorithms based on decision trees. The forecasting scheme employs the skforecast library from the Python environment, which facilitates the implementation of different schemes for both deterministic and probabilistic forecasting applications. Firstly, deterministic forecasting of hourly BIPV power was performed with XGBoost and Random Forest algorithms for different cases, showing an improvement in forecasting accuracy when some exogenous variables were used. Secondly, probabilistic forecasting was performed with XGBoost combined with the Bootstrap method. The results of this paper show the capabilities of Random Forest and gradient boosting algorithms, such as XGBoost, to work as regressors in time series forecasting of BIPV power. Mean absolute error in the deterministic forecast, using the most influencing exogenous variables, were around 40% and close below 30% for the south and east array, respectively.

Impacts of the Aerosol Representation in WRF-Solar Clear-Sky Irradiance Forecasts over CONUS

Abstract Aerosol optical depth (AOD) is a primary source of solar irradiance forecast error in clear-sky conditions. Improving the accuracy of AOD in NWP models like WRF will thus reduce error in both direct normal irradiance (DNI) and global horizontal irradiance (GHI), which should improve solar power forecast errors, at least in cloud-free conditions. In this study clear-sky GHI and DNI was analyzed from four configurations of the WRF-Solar model with different aerosol representations: 1) the default Tegen climatology, 2) imposing AOD forecasts from the GEOS-5 model, 3) imposing AOD forecasts from the Copernicus Atmosphere Monitoring Service (CAMS) model, and 4) the Thompson–Eidhammer aerosol-aware water/ice-friendly aerosol climatology. More than 8 months of these 15-min output forecasts are compared with high-quality irradiance observations at NOAA SURFRAD and Solar Radiation (SOLRAD) stations located across CONUS. In general, WRF-Solar with GEOS-5 AOD had the lowest errors in clear-sky DNI, while WRF-Solar with CAMS AOD had the highest errors, higher even than the two aerosol climatologies, which is consistent with validation of the four AOD550 datasets against AERONET stations. For clear-sky GHI, the statistics differed little between the four models, as expected because of the lesser sensitivity of GHI to aerosol loading. Hourly average clear-sky DNI and GHI were also analyzed, and they were additionally compared with CAMS model output directly. CAMS irradiance performed competitively with the best WRF-Solar configuration (with GEOS-5 AOD). The markedly different performance of CAMS versus WRF-Solar with CAMS AOD indicates that CAMS is apparently less sensitive to AOD550 than WRF-Solar is. Significance Statement Particles in the atmosphere called aerosols, which can include dust, smoke, sea salt, sulfates, black carbon, and organic carbon, absorb and scatter incoming sunlight. Improving the representation of aerosols in numerical weather prediction models reduces forecast errors in solar irradiance at ground level, particularly direct normal irradiance, during cloud-free conditions. This in turn should result in improved accuracy of solar power forecasts, especially for concentrated solar power (CSP) plants. CSP plants tend to be built in more arid, less cloudy regions that are also prone to dust loading, so accurate aerosol forecasts are particularly relevant. Comparing four representations of aerosols in the WRF-Solar model over eight months of forecasts across the United States reveals substantial differences in clear-sky irradiance forecast skill.

A Comprehensive Review on Ensemble Solar Power Forecasting Algorithms

With increasing demand for energy, the penetration of alternative sources such as renewable energy in power grids has increased. Solar energy is one of the most common and well-known sources of energy in existing networks. But because of its non-stationary and non-linear characteristics, it needs to predict solar irradiance to provide more reliable Photovoltaic (PV) plants and manage the power of supply and demand. Although there are various methods to predict the solar irradiance. This paper gives the overview of recent studies with focus on solar irradiance forecasting with ensemble methods which are divided into two main categories: competitive and cooperative ensemble forecasting. In addition, parameter diversity and data diversity are considered as competitive ensemble forecasting and also preprocessing and post-processing are as cooperative ensemble forecasting. All these ensemble forecasting methods are investigated in this study. In the end, the conclusion has been drawn and the recommendations for future studies have been discussed.

Improved Reptile Search Optimization Algorithm: Application on Regression and Classification Problems

The reptile search algorithm is a newly developed optimization technique that can efficiently solve various optimization problems. However, while solving high-dimensional nonconvex optimization problems, the reptile search algorithm retains some drawbacks, such as slow convergence speed, high computational complexity, and local minima trapping. Therefore, an improved reptile search algorithm (IRSA) based on a sine cosine algorithm and Levy flight is proposed in this work. The modified sine cosine algorithm with enhanced global search capabilities avoids local minima trapping by conducting a full-scale search of the solution space, and the Levy flight operator with a jump size control factor increases the exploitation capabilities of the search agents. The enhanced algorithm was applied to a set of 23 well-known test functions. Additionally, statistical analysis was performed by considering 30 runs for various performance measures like best, worse, average values, and standard deviation. The statistical results showed that the improved reptile search algorithm gives a fast convergence speed, low time complexity, and efficient global search. For further verification, improved reptile search algorithm results were compared with the RSA and various state-of-the-art metaheuristic techniques. In the second phase of the paper, we used the IRSA to train hyperparameters such as weight and biases for a multi-layer perceptron neural network and a smoothing parameter (σ) for a radial basis function neural network. To validate the effectiveness of training, the improved reptile search algorithm trained multi-layer perceptron neural network classifier was tested on various challenging, real-world classification problems. Furthermore, as a second application, the IRSA-trained RBFNN regression model was used for day-ahead wind and solar power forecasting. Experimental results clearly demonstrated the superior classification and prediction capabilities of the proposed hybrid model. Qualitative, quantitative, comparative, statistical, and complexity analysis revealed improved global exploration, high efficiency, high convergence speed, high prediction accuracy, and low time complexity in the proposed technique.

Machine Learning Based Solar Photovoltaic Power Forecasting: A Review and Comparison

The growing interest in renewable energy and the falling prices of solar panels place solar electricity in a favourable position for adoption. However, the high-rate adoption of intermittent renewable energy introduces challenges and the potential to create power instability between the available power generation and the load demand. Hence, accurate solar Photovoltaic (PV) power forecasting is essential to maintain system reliability and maximize renewable energy integration. The current solar PV power forecasting approaches are an essential tool to maintain system reliability and maximize renewable energy integration. This paper presents a comprehensive and comparative review of existing Machine Learning (ML) based approaches used in PV power forecasting, focusing on short-term horizons. We provide an overview of factors affecting solar PV power forecasting and an overview of existing PV power forecasting methods in the literature, with a specific focus on ML-based models. To further enhance the comparison and provide more insights into the advancement in the area, we simulate the performance of different ML methods used in solar PV power forecasting and, finally, a discussion on the results of the work.

A Survey of the Researches on Grid-Connected Solar Power Generation Systems and Power Forecasting Methods Based on Ground-Based Cloud Atlas

Photovoltaic power generating is one of the primary methods of utilizing solar energy resources, with large-scale photovoltaic grid-connected power generation being the most efficient way to fully utilize solar energy. In order to provide reference strategies for pertinent researchers as well as potential implementation, this paper tries to provide a survey investigation and technical analysis of machine learning-related approaches, statistical approaches and optimization techniques for solar power generation and forecasting. Deep learning-related methods, in particular, can theoretically handle arbitrary nonlinear transformations through proper model structural design, such as hidden layer topology optimization and objective function analysis to save information that can increase forecasting accuracy while filtering out irrelevant or less affected data for forecasting. The research’s results indicate that RBFNN-AG performed the best when applying the predetermined number of days, with an NRMSE value of 4.65%. RBFNN-AG performs better than sophisticated models like DenseNet (5.69%), SLFN-ELM (5.95%), and ANN-k-means-linear regression correction (6.11%). Additionally, scenario application and PV system investment techniques are provided to evaluate the current condition of new energy development and market trends both domestically and internationally.

A Novel Forecasting Model for Solar Power Generation by a Deep Learning Framework With Data Preprocessing and Postprocessing

Photovoltaic power has become one of the most popular forms of energy owing to the growing consideration of environmental factors; however, solar power generation has brought many challenges for power system operations. With regard to optimizing safety and reducing the costs of power system operations, an accurate and reliable solar power forecasting model would be a significant step forward. This study proposes a deep learning method to improve the performance of short-term one-hour-ahead solar power forecasting, which includes data preprocessing, feature engineering, kernel principal component analysis, a gated recurrent unit network training mode based on time-of-day classification, and postprocessing with error correction. Both historical solar power, solar irradiance, and numerical weather prediction (NWP) data, such as temperature, irradiance, rainfall, wind speed, air pressure, and humidity, were used as the input dataset in this work. As a case study, the measured power from ten PV sites in Taiwan were collected and predicted with a one-hour resolution. The normalized root mean squared error and normalized mean absolute percent error were chosen to evaluate the performance of the forecasting models. Compared with other benchmark models, including ANN, LSTM, XGBoost, and single GRU, the experimental results showed the proposed model's superior performance. Furthermore, the importance of data preprocessing and postprocessing based on error correction was demonstrated.