| Work Detail |
A Dutch team of researchers has developed a solar radiation forecasting model that uses the short-term memory (LSTM) technique. Apparently, the proposed methodology achieves better results than other forecasting methods. Researchers from Utrecht University in the Netherlands and EKO Instruments Europe have developed a novel short-term solar irradiance prediction and machine learning system based on all-sky images. The model is based on the long-term memory model (LSTM), which is a type of recurrent neural network capable of learning order dependence in sequential prediction problems. The LSTM technique takes the relevant parts of a pre-trained machine learning model and applies it to a new but similar problem. “Our goal is to improve prediction accuracy by increasing the data and possibly expanding the sensor network to include real-time data from multiple locations,” Khadija Barhmi, lead author of the research, told pv magazine . The academics explained that LSTM uses memory cells, which can store information for a long period, and gates, which control the flow of information and make decisions about what is forgotten and what is kept. “LSTM networks, a subset of recurrent neural networks (RNN), are known for their exceptional ability to model temporal sequences,” they stated. “By integrating LSTM with all-sky images and local meteorological data, our research formulates a novel approach whose method allows creating future representations of the evolution of the sky.” The prediction method developed by the scientist uses images of the entire sky, capturing images of the sky from which the model can learn about the location, movement and formation of clouds. It employs the threshold algorithm, which classifies pixels as clouds or clear sky based on predefined thresholds. Cloud motion is calculated using a two-frame flux estimation algorithm. The team placed all-sky cameras at the Plataforma Solar de Almería (PSA) facilities in southern Spain, spaced about 880.2 meters apart. Each took images covering a 180-degree field of view, with a sampling rate of 15 seconds. In addition, sensors measuring ambient temperature, global horizontal irradiance (GHI), and relative humidity were placed in the same location. The LSTM model was also fed with external meteorological data collected from open sources, such as the precise sky index and sun-earth distance. “As a data set, we consider the data from August 1, 2019 to December 31, 2019. These comprise 121 sunny days, 29 partly cloudy days, and three cloudy days,” the research group said. “To evaluate the effectiveness of various parameters, we conducted a comprehensive comparison over nine days.” The validation set comprised different weather conditions, including sunny days, partly cloudy days, and cloudy days. The novel method has been measured against other machine learning methods such as Random Forest (RF) and artificial neural networks (ANN) with this validation set. It has also been compared with the persistence model, a standard reference model in solar irradiance prediction, and with the state-of-the-art SKIPPD model. All of them were tested with variations of the input data, with different combinations of sky images, sensor data and weather data. “RF and ANN models perform best when using only the in situ feature subset, indicating that they fail to capture valuable information from features extracted from images,” the scientists say. “However, LSTM outperforms them when given access to this subset. By capturing the complexity of these features, LSTM provides the best predictions among all tested models. Furthermore, the “all data” feature set provides the best results for LSTM on average, across all weather conditions.” The ramp score (RS) of the LSTM model - used to measure the forecasting ability of GHI fluctuations - was 39% in sunny conditions and 25% in partly cloudy conditions. “Our benchmarking activities involved a comparative analysis between our deep learning model and the SKIPPD method,” the group added. “The LSTM model excelled in validation, demonstrating superior capture of temporal dynamics crucial for solar prediction.” The scientists presented their model in the study “ All sky imaging-based short-term solar irradiance forecasting with Long Short-Term Memory networks. ” long term), published in Solar Energy . |
