A Flexible and Robust Deep Learning-Based System for Solar Irradiance Forecasting

Abstract

Most studies about the solar forecasting topic do not analyze and exploit the temporal and spatial components that are inherent to such a task. Furthermore, they mostly focus just on precision and not on other meaningful features, such as flexibility and robustness. With the current energy production trends, where many solar panels are distributed across city rooftops, there is a need to manage all this information simultaneously and to be able to add and remove sensors as needed. Likewise, robust models need to be able to cope with (inevitable) sensor failure and continue producing reliable predictions. Due to all of this, solar forecasting models need to be as decoupled as possible from the number of data sources that feed them and their geographical distribution, enabling also the reusability of the models. This article contributes with a family of Deep Learning models for solar irradiance forecasting complying with the aforementioned features, i.e. flexibility and robustness. In the first stage, several Artificial Neural Networks are trained as a basis for predicting solar irradiance on several locations at the same time. Thereupon, a family of models that work with irradiance maps thanks to Convolutional Long Short-Term Memory layers is presented, obtaining forecast skills between 7.4% and 41% (depending on the location and horizon) compared to the baseline. The latter family comes with flexibility and robustness features, which are required in large-scale Intelligent Environments, such as Smart Cities. Working with irradiance maps means that new sensors can be added (or removed) as needed, without requiring rebuilding the model. Experiments carried out show that sensor failures have a mild impact on the prediction error for several forecast horizons.

Solar Radiation Forecasting by Pearson Correlation Using LSTM Neural Network and ANFIS Method: Application in the West-Central Jordan

“…The achieved RMSE ranges from 0.04 to 0.8, which is very competitive compared to other experimental results obtained in the same region. As for the research works carried out in North America and Hawaii [15,16,19], the analysis time length ranged from 6 to 20 months, obtaining RMSE values equal to 6.11 and 0.086, respectively. This means that the short-term forecast (only six months) provides a lower outcome prediction with respect to the longer ones (up to 20 months).…”

“…However, since some significant parameters are not considered in the study, this could impact the prediction accuracy negatively. In [15], the convolutional LSTM method is used to forecast solar irradiance on several locations simultaneously; the obtained RMSE is less than 15%. However, in this study, the proposed DL technique could require additional training (transfer learning) if different locations are considered.…”

Review on Spatio-Temporal Solar Forecasting Methods Driven by In Situ Measurements or Their Combination with Satellite and Numerical Weather Prediction (NWP) Estimates

“…Khodayar et al [108] built a Convolutional Graph Autoencoder (CGAE) to predict solar irradiance for 75 solar plants within a time horizon of 30 min to 6 h. They performed the extraction of spatial characteristics through convolutional graphs with the information being processed in the encoder and decoder layers to calculate the distribution of the solar data to be predicted. Prado-Rujas et al [42] used a Convolutional Long Short-Term Memory (Conv-LSTM) model to predict solar irradiance using data collected from 17 sensors; the model used a sequence of irradiance maps (obtained through a nearestneighbor interpolation) as input. The authors worked with a prediction horizon of up to 1 h and analyzed how the location influenced the ability of the system to achieve better performance, resulting in increased robustness and flexibility, as the model was capable of handling missing data events where information from one or more sensors is not available.…”

“…It is also interesting to note that several works that propose forecasting methods assume different data representations to capture the spatio-temporal component of solar variability. There are authors that use as input a sequence of irradiance maps [42], a matrix representation of the spatial-temporal relationship between the sites [107], a graph structure where the nodes are the measurements locations and the vertices are the distance [108] or the correlation relationship between them [34]. The works considered in this review study refer to spatially distributed ground sensors (e.g., [15,103,127]), complemented with gridded satellite estimates (e.g., [97]), or NWP forecasts (e.g., [29,32,65,99]), or a combination of these (e.g., [11,24,29]).…”

A deep attention-driven model to forecast solar irradiance

“…In [33], an approach using multivariate gated recurrent units (GRUs) is applied for hourahead solar irradiance forecasting. In [34], authors propose a Convolutional LSTM layers model to forecast solar irradiance. In prado2021flexible, a flexible data-driven approach using Convolutional Long Short-Term Memory layers is introduced for spatiotemporal forecasting of solar irradiance.…”