In 2022, Instrumentation Technologies joined a consortium tasked with developing a big data platform with artificial intelligence for current, short-term, and long-term forecasting of the electricity production of the photovoltaic power plant -named Leonardo.
The platform is based on an IoT ecosystem with continuous monitoring of weather sensors and a weather camera installed on the photovoltaic power plant, which will capture various meteorological data. At the same time, the system also includes data from an external meteorological provider and data from the customer’s (photovoltaic plant owner) backend systems. All these data are processed using advanced mathematical procedures, computer vision and machine learning.
Instrumentation Technologies task was to implement sky object detection and prediction (mainly clouds) components. For the implementation of them, we have set up a system that includes an application, interfaces with peripherals (weather station, camera, pyranometers), and the necessary databases. The user plays a crucial role in this system, accessing the application through the API interface—HTTP Interface, thereby contributing to the process of electricity production forecasting.
Methods used:
A) One of the used approaches in this field is the GAN (Generative Adversarial Networks) machine learned model. This technique predicts solar radiation based on a GAN model that forms an image of the sky for N minutes in advance and then obtains solar radiation for N minutes in advance based on the formed image.
B) In the second method, a CNN (Convolutional Neural Networks) model predicts solar radiation for N minutes in advance. The model accepts a history or sequence of sky images and a history of weather data. The model trained on the sequence of these data can extract connections between the predicted solar radiation and the input data, thus predicting solar radiation for N minutes in advance.
Using Instrumentation Technologies solution, the Leonardo platform can accurately forecast electric energy production, thus ensuring stable and efficient (cost and environmental) short-term and long-term management of electricity production and significantly reducing the power plants’ maintenance and operating costs.
The platform effectively controls the elements of the photovoltaic plant that are subject to maintenance, thereby minimizing costs and maximizing efficiency.
