New Artificial Intelligence Methods for PV Annual Degradation Rate Estimation
| Autor: | Braisaz, B., Sanaa, A., Rhazi, O.L., Bila, M., Tounkara, I., Becker, A. |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2022 |
| Předmět: | |
| DOI: | 10.4229/wcpec-82022-4do.1.3 |
| Popis: | 8th World Conference on Photovoltaic Energy Conversion; 1064-1069 Artificial Intelligence methods are overwhelming every R&D area where data are used, including PV module performances. Here we propose to test new Artificial Intelligence (AI) methods in order to estimate the degradation rate of PV Modules. Among these AI methods, we implemented Long Short Term Memory (LSTM) neural network. In the learning step, known degradation rates were simulated in a digital power plant and by using meteorological data set of an EDF Renewables outdoor test facility (almost 4 years of data, PV module production & meteorological conditions). Then, these generated data were used to teach AI systems. In the test step, the AI is expected to assess ou find the degradation rate by analyzing a 3 years-long dataset. In addition, we developed a method for estimating electrical parameters (Isc, nds, Rs) and another method for saving the state of plant in order to replay it later with different meteorological data. The results are promising with the last method and show the performance and efficiency of AI when estimating degradation rate. These methods have many interests, but one of the most important is to combine PV model and the exact field meteorological dataset to simulate various degradation amplitudes and shapes in order to feed an AI. By this way and knowing the meteorological conditions, AI can learn and propose the expected PV behavior (every shape is normalized) and interpolate the degradation rate corresponding to the field values of PV module production. A final comparison is done with data from EDF R’s test zone in US by comparing the estimated AI degradation rate and the STC (Standard Test Conditions) flashes of modules carried out recently. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|