In Reunion, where the objective is a 100

Microgrids, which promote the production and consumption of renewable energy on site, are a relevant solution to reduce carbon emissions and the price of energy for end users. However, converting an existing building stock into a microgrid powered mainly by renewable energy requires finding a technical and economic optimum while taking into account strong constraints. This work proposes a methodology to achieve this objective on an existing university campus located in La Reunion, a French island in the Indian Ocean. The campus already has three photovoltaic (PV) systems and high-quality measurement data of weather, loads and energy production. The goal of the work is to find an optimal rooftop PV capacity that maximizes campus self-sufficiency while keeping energy price affordable for users. The results do not highlight a unique combination of roofs as a solution to the optimization problem. However, the analysis of possible combinations gives clear rules for defining the total photovoltaic capacity to be installed and selecting the most suitable roofs.

In this work, we develop simple linear models that allow users to predict solar irradiance forecast errors based solely on solar variability at a specific location on Earth. These straightforward yet actionable models enable solar forecasters to quickly estimate forecast errors for a given site, providing a clear indication of how well their forecasting models are likely to perform. The error in deterministic solar irradiance forecasts is measured by the Root Mean Square Error (RMSE), while solar variability is quantified by the standard deviation of an hourly time series of changes in the dimensionless clear sky index. Sixty sites distributed around the globe are used to build two types of RMSE prediction models. The first type is for intra-day forecasts (1-hour to 6-hour forecast horizons), while the second is for day-ahead forecasts (24-hour horizon). The derivation of the intra-day forecast error prediction model leverages on a non-linear time series approach whereas the one for day-ahead forecast error relies on forecasts issued by the European Centre for Medium-Range Weather Forecasts (ECMWF). For each type of model, we calculate also the 2.5% and 97.5% percentiles of the distribution in order to estimate the 95% uncertainty interval associated with the prediction. This uncertainty interval defines the bounds of the RMSE within which 95% of future RMSE values are expected to fall. These error bounds can provide solar forecasters with valuable insights into the performance of their solar forecasting methods in relation to the forecast challenges posed by site-specific variability. Verification against published results in the literature, specifically for seven sites of the SURFRAD network, demonstrates that these models can satisfactorily predict intra-day and day-ahead forecast RMSEs using only site-specific solar variability data.

In recent years, the prominence of probabilistic forecasting has risen among numerous research fields (finance, meteorology, banking, etc.). Best practices on using such forecasts are, however, neither well explained nor well understood. The question of the benefits derived from these forecasts is of primary interest, especially for the industrial sector. A sound methodology already exists to evaluate the value of probabilistic forecasts of binary events. In this paper, we introduce a comprehensive methodology for assessing the value of probabilistic forecasts of continuous variables, which is valid for a specific class of problems where the cost functions are piecewise linear. The proposed methodology is based on a set of visual diagnostic tools. In particular, we propose a new diagram called EVC (“Effective economic Value of a forecast of Continuous variable”) which provides the effective value of a forecast. Using simple case studies, we show that the value of probabilistic forecasts of continuous variables is strongly dependent on a key variable that we call the risk ratio. It leads to a quantitative metric of a value called the OEV (“Overall Effective Value”). The preliminary results suggest that typical OEVs demonstrate the benefits of probabilistic forecasting over a deterministic approach.

Despite the growing awareness in academia and industry of the importance of solar probabilistic forecasting for further enhancing the integration of variable photovoltaic power generation into electrical power grids, there is still no benchmark study comparing a wide range of solar probabilistic methods across various local climates. Having identified this research gap, experts involved in the activities of IEA PVPS T161 agreed to establish a benchmarking exercise to evaluate the quality of intra-hour and intra-day probabilistic irradiance forecasts. The tested forecasting methodologies are based on different input data including ground measurements, satellite-based forecasts and Numerical Weather Predictions (NWP), and different statistical methods are employed to generate probabilistic forecasts from these. The exercise highlights different forecast quality depending on the method used, and more importantly, on the input data fed into the models. In particular, the benchmarking procedure reveals that the association of a point forecast that blends ground, satellite and NWP data with a statistical technique generates high-quality probabilistic forecasts. Therefore, in a subsequent step, an additional investigation was conducted to assess the added value of such a blended point forecast on forecast quality. Three new statistical methods were implemented using the blended point forecast as input. To ensure a fair evaluation of the different methods, we calculate a skill score that measures the performance of the proposed model relative to that of a trivial baseline model. The closer the skill score is to 100%, the more efficient the method is. Overall, skill scores of methods that use the blended point forecast ranges from 42% to 46% for the intra-hour scenario and 27% to 32% for the intra-day scenario. Conversely, methods that do not use the blended point forecast exhibit skill scores ranging from 33% to 43% for intra-hour forecasts and 8% to 16% for intra-day forecasts. These results suggest that using (a) blended point forecasts that optimally combine different sources of input data and (b) a post-processing with a statistical method to produce the quantile forecasts is an effective and consistent way to generate high-quality intra-hour or intra-day probabilistic forecasts.

With the fast increase of solar energy plants, a high-quality short-term forecast is required to smoothly integrate their production in the electricity grids. Usually, forecasting systems predict the future solar energy as a continuous variable. But for particular applications, such as concentrated solar plants with tracking devices, the operator needs to anticipate the achievement of a solar irradiance threshold to start or stop their system. In this case, binary forecasts are more relevant. Moreover, while most forecasting systems are deterministic, the probabilistic approach provides additional information about their inherent uncertainty that is essential for decision-making. The objective of this work is to propose a methodology to generate probabilistic solar forecasts as a binary event for very short-term horizons between 1 and 30 min. Among the various techniques developed to predict the solar potential for the next few minutes, sky imagery is one of the most promising. Therefore, we propose in this work to combine a state-of-the-art model based on a sky camera and a discrete choice model to predict the probability of an irradiance threshold suitable for plant operators. Two well-known parametric discrete choice models, logit and probit models, and a machine learning technique, random forest, were tested to post-process the deterministic forecast derived from sky images. All three models significantly improve the quality of the original deterministic forecast. However, random forest gives the best results and especially provides reliable probability predictions.

This paper focuses on the modelling of occupant behaviour in the case of a non-residential mixed-mode building on the tropical island of La Réunion. For such areas and types of buildings, occupants can operate passive solutions to achieve comfort while energy-consuming ones can offer alternatives during the hottest months. Yet, compared to other climatic zones, specific knowledge on occupant comfort and behaviour is limited, making the work of engineers difficult during the design phase. In this work, occupants' operations on hygrothermal comfort controls, such as windows and fans, were first measured and analysed. Secondly, these behaviours were modelled using two deterministic methods based on machine learning techniques and a probabilistic graphical model. A model was also implemented to estimate the number of people, using the power demand of the electrical outlets. The estimation ability of the behavioural models was evaluated and led to F1 scores greater than 0.7. A two-classifier model was proposed to estimate the level of ceiling fan use. This combined model slightly improves the F1 scores by more that 2%, which demonstrates the necessity of taking into account the links between the different controls.

In the realm of solar forecasting, it is common to use a clear sky model output to deseasonalise the solar irradiance time series needed to build the forecasting models. However, most of these clear sky models require the setting of atmospheric parameters for which accurate values may not be available for the site under study. This can hamper the accuracy of the prediction models. Normalisation of the irradiance data with a clear sky model leads to the construction of forecasting models with the so-called clear sky index. Another way to normalize the irradiance data is to rely on the extraterrestrial irradiance, which is the irradiance at the top of the atmosphere. Extraterrestrial irradiance is defined by a simple equation that is related to the geometric course of the sun. Normalisation with the extraterrestrial irradiance leads to the building of models with the clearness index. In the solar forecasting domain, most models are built using time series based on the clear sky index. However, there is no empirical evidence thus far that the clear sky index approach outperforms the clearness index approach. Therefore the goal of this preliminary study is to evaluate and compare the two approaches. The numerical experimental setup for evaluating the two approaches is based on three forecasting methods, namely, a simple persistence model, a linear AutoRegressive (AR) model, and a non-linear neural network (NN) model, all of which are applied at six sites with different sky conditions. It is shown that normalization of the solar irradiance with the help of a clear sky model produces better forecasts irrespective of the type of model used. However, it is demonstrated that a nonlinear forecasting technique such as a neural network built with clearness time series can beat simple linear models constructed with the clear sky index.

This work proposes a methodology based on the probabilistic dynamic programming (PDP) to integrate operational probabilistic forecasts of a photovoltaic (PV) plant into the optimization of the day-ahead schedule of an energy storage system (ESS). The proposed approach is tested on a microgrid based on a real educational building, a PV farm and Li-ion batteries. The objective is to minimize the operating cost of the microgrid. The operational day-ahead forecasts are derived from the Ensemble Prediction System (EPS) provided by a well-known Numerical Weather Prediction (NWP) model. Contrary to the classical use of deterministic forecasts, we demonstrate that the integration of the probabilistic forecasts in the optimization process leads to a more efficient microgrid management and to a reduction of up to 38% of the operating costs. Besides, it is shown that the non linearity resulting from the power dependency of the efficiency of the inverters must be taken into account in order to yield relevant optimization results.

The hybridization of renewable energy resources is a known topic in sustainable technology. Many projects are going on based on the topic. The use of Photovoltaic, wind energy, and other renewable resources can be helpful to optimize the load in the utility grid. Countries like Europe and other western countries have electricity storage, whether the developing countries are still struggling to make sure the stable utility grid connection to the distribution network system. In this research, we would like to discuss the different energy production processes sustainably. As we know, the energy sources are volatile and cannot always assure stable production to keep the requirements or demand properly. We want to use the combination of the sources in a way so that we can make the balance between the demand and the supply system. This research will be an overview in terms of technical and financial sites. Also, by using the different combinations of the Internet of things and data analysis method, we will see the correlation between the different sources and their production. Based on the production data, we can determine the financial feasibility and the outcome of the system. The main problem of renewable energy sources is uncertainty. In terms of wind energy, the velocity is also not stable according to the location. We want to show a predictive model by using the intelligent formula by which we can maintain the hybrid system. The production data from different sources will tell us their contribution to the system. This contribution will help us monitor the system and control which sources have more contribution on the demand side. The predictive model will have consisted of renewable sources such as photovoltaic, wind, utility grid, and inverter systems. In the research, the tool such as Artificial Intelligence can be implemented by sustainable management. The arrangement information is prepared to extricate data and based on resources. The renewable sources data are variant according to their location and it has an impact in terms of energy production. Data acquisition and analysis could help the current technologies such as smart grid, microgrid, and their control systems. This exploration aims to introduce a predictive foundation for the management of enormous volumes of data through large Information instruments (sensors) to help the coordination of environmentally friendly power. The main difference between the conventional electricity system and the renewable energy system is the variability of sources, with conventional sources such as utility grids and diesel generators and renewable sources consisting of photovoltaic (PV), wind, etc.

In this paper, we present the results obtained by modelling the users' behaviours in a mixed mode office building in a tropical climate, more exactly in La Réunion. Few specific research studies on comfort in tropical climates have been published, and there is little feedback on the users' behaviour in these buildings. In order to improve users' assumptions in the design phase, users' actions on ceiling fans and windows have been measured and analysed. These data have then been modelled by machine learning methods, according to hygrothermal comfort and occupancy. The F1 scores eventually obtained for predicting fan use by random forests, decision trees and Bayesian networks are 99%, 98% and 95% respectively. For windows use, the F1 scores obtained are 92%, 91% and 70%, which demonstrates the ability of the models tested to predict the users' behaviours.

In this paper, the performances of two approaches for solar probabilistic are evaluated using a set of metrics previously tested by the meteorology verification community. A particular focus is put on several scores and the decomposition of a specific probabilistic metric: the continuous rank probability score (CRPS) as they give extensive information to compare the forecasting performance of both methodologies. The two solar probabilistic forecasting methodologies are used to produce intra-day solar forecasts with time horizons ranging from 1 h to 6 h. The first methodology is based on two steps. In the first step, we generated a point forecast for each horizon and in a second step, we use quantile regression methods to estimate the prediction intervals. The second methodology directly estimates the prediction intervals of the forecasted clear sky index distribution using past data as inputs. With this second methodology we also propose to add solar geometric angles as inputs. Overall, nine probabilistic forecasting performances are compared at six measurements stations with different climatic conditions. This paper shows a detailed picture of the overall performance of the models and consequently may help in selecting the best methodology.

Based on the reported literature and commonly used metrics in the realm of solar forecasting, a new methodology is developed for estimating a metric called forecastability (F). It reveals the extent to which solar radiation time series can be forecasted and provides the crucial context for judging the inherent difficulty associated with a particular forecast situation. Unlike the score given by the standard smart persistence model, the F metric which is bounded between 0% and 100% is easier to interpret, hence making comparisons between forecasting studies more consistent. This approach uses the Monte Carlo method and estimates F from the standard error metric RMSE and the persistence predictor. Based on the time series of solar radiation measured at six very different locations (with optimized clear sky model) from a meteorological point of view, it is shown that F varies between 25.5% and 68.2% and that it exists a link between forecastability and errors obtained by machine learning prediction methods. The proposed methodology is validated for 3 parameters that may affect the F estimation (time horizon, temporal granularity, and solar radiation components) and for 50 time series relative to McClear web service and to the central archive of Baseline Surface Radiation Network.

During the last decade, numerous solar forecasting tools have been developed to predict the energy generation of photovoltaic (PV) farms. The quality of solar forecasts is assessed by comparing predictions with measured solar data. However, this methodology does not consider the added value of the forecasts for their applications. As a consequence, what value could be given to the improvement of forecasts considering this evaluation framework? To answer this question, this work compares the value of different operational solar forecasts for a specific application. The aim is to look for relationships between the economic value and the error metrics defined to evaluate the forecast quality. A new generation of large-scale PV plants integrates ESS. The aim is to add flexibility to the injection of the production into the grid and thus to maximize the profit by taking advantage of the possibilities offered by the electricity market, such as energy arbitrage. To optimize the operation of these specific ESS, forecasting of the solar production is of paramount importance. The study case considered in this work is a large-scale PV farm of several megawatts associated with Li-ion batteries in the Australian energy market context. For this specific case study, the results show that the metrics used to evaluate the forecast quality based on the mean absolute error (MAE) have an almost linear relationship with the economic gain brought by applying the forecast. More precisely, an improvement of 1% point in MAE results approximately in an increase of 2% points in economical gain.

Probabilistic solar forecasting is becoming a major topic in the solar research community as it provides more information about the uncertainty of the forecast compared to deterministic forecasting. However, to facilitate the adoption of probabilistic forecasts within solar forecasting communities (industry and academic), the definition and the use of standardized best practices are a prerequisite. Among others, there is a need for benchmark models that are able to properly assess the performance of new probabilistic forecasting methods. In this work, we propose a new climatology benchmark model called "CSD-CLIM" (for Clear-Sky Dependent Climatology). This new reference model is evaluated against two other climatology benchmark models namely the naive climatology and a well-referenced model in the literature, the CH-PeEn (for Complete History Persistence Ensemble). The verification of compliance with a set of properties that a climatology benchmark model must follow demonstrates that the new CSD-CLIM model outperforms the naive climatology and that it can be a viable alternative to the CH-PeEn model. It is shown that the better performance of CSD-CLIM is due to a specific binning of the historical irradiance data based on the clear-sky irradiance values.