The forecasting of solar irradiance in photovoltaic power generation is an important tool for the integration of intermittent renewable energy sources (RES) in electrical utility grids. This study evaluates two machine learning (ML) algorithms for intraday solar irradiance forecasting: multigene genetic programming (MGGP) and the multilayer perceptron (MLP) artificial neural network (ANN). MGGP is an evolutionary algorithm white-box method and is a novel approach in the field. Persistence, MGGP and MLP were compared to forecast irradiance at six locations, within horizons from 15 to 120 min, in order to compare these methods based on a wide range of reliable results. The assessment of exogenous inputs indicates that the use of additional weather variables improves irradiance forecastability, resulting in improvements of 5.68% for mean absolute error (MAE) and 3.41% for root mean square error (RMSE). It was also verified that iterative predictions improve MGGP accuracy. The obtained results show that location, forecast horizon and error metric definition affect model accuracy dominance. Both Haurwitz and Ineichen clear sky models have been implemented, and the results denoted a low influence of these models in the prediction accuracy of multivariate ML forecasting. In a broad perspective, MGGP presented more accurate and robust results in single prediction cases, providing faster solutions, while ANN presented more accurate results for ensemble forecasting, although it presented higher complexity and requires additional computational effort.

Renewable energy sources (RES) like solar and wind, naturally present high daily and seasonal variability of power generation. This paper presents a real case study of solar photovoltaic (PV) power forecasting of a building-integrated PV plant as a tool for power systems operation. A new methodology composed of data processing strategy, Artificial Neural Networks (ANN) modelling, and error metrics definitions are proposed and validated for the intra-day PV power forecasting. Data were modelled in time steps of 15-minute averages and the forecasts obtained from 15 minutes to 120 minutes ahead. A persistence model is developed for the building-integrated PV systems (BIPVS) case. The forecast skills achieved range from 9.79% to 23.75%, which measures the ANN improvement in relation to the proposed benchmark persistence model.

The operating cost reduction of photovoltaic (PV) systems is an important way to increase its economic viability for stakeholders. In this study, an inverter sizing ratio (ISR) analysis is carried out in order to quantify its potential benefit in the context of building‐integrated PV systems and PV distributed generation in the central region of Brazil. An algorithm was created to analyse the ISR considering 12 years of measured climate data and data provided by manufacturers of system components. Among three analytical methodologies discussed in this study, the power limitation analysis seems to be the best reference for design adjustments related to a specific location or energy market. By that, a potential reduction over 10% of levelised cost of energy can be achieved in case of varying ISR from the unity value (theoretical) to its new fitted value.

Development and improvement of solar forecasting models have been extensively addressed in the past years due to the importance of solar energy as a renewable energy source. This work presents an application and improvement of intra-day solar predictive models based on genetic programming. Forecasts were evaluated in time horizons of 10 minutes up to 180 minutes ahead as future steps at two completely different locations: one in northern hemisphere and another in the southern hemisphere. The improvement strategy was validated in comparison of error metrics to the ones obtained by benchmark methods of solar forecasting. The proposed model results will be presented and validated for each considered location.

Accurate and practical forecasting models are very important as tools for optimal integration of the solar energy source in smart grids. This work presents a comparison of four models of intra-day irradiance forecasting based on genetic programming. These models are evaluated at two distinct locations, with completely different climate characteristics, with data structured in 10-minute averages to forecast irradiance up to 180 minutes ahead. The models differ in the addition of exogenous weather variables or exogenous deterministic irradiance components. With the use of genetic programming, and at these specific locations, the addition of exogenous weather variables did not result in permanent accuracy improvement, while addition of the deterministic irradiance component did.