Forecasting of Solar Power Generation for Experimental Dual-Axis Solar Tracker System based on ANN and FPGA Technology | ||
| Engineering Research Journal | ||
| Volume 184, Issue 2, March 2025, Pages 1-23 PDF (1.19 M) | ||
| Document Type: Original Article | ||
| DOI: 10.21608/erj.2024.332639.1134 | ||
| Authors | ||
| Elsayed A. M. BaBars1; Shady Yehia El-Mashaad2; Saber M. Abdraboo3; Mohamed El-sayed M. Essa* 4 | ||
| 1Department of Mechanical Engineering, Faculty of Engineering, Sinai University | ||
| 2Department of Computer Systems Engineering, Faculty of Engineering at Shoubra, Benha University, Cairo 13511, Egypt; | ||
| 3Department of Mechanical Engineering, Faculty of Engineering at Shoubra, Benha University | ||
| 4Department of Electrical Power and Machines Engineering, Institute of Aviation Engineering and Technology, Egyptian Aviation Academy, Ministry of Civil Aviation | ||
| Abstract | ||
| This paper focuses on intelligent prediction of solar power generation from experimental solar system based on artificial neural network (ANN). The proposed hardware system is designed and implemented for dual-axis solar tracking system (DASTS). In addition, the Field Programmable Gate Arrays (FPGA) is applied to the system as the brain and core of the control system. The FPGA board is a spartan edge accelerator which gives high processing speed to enable speedy and precise alignments. The prediction models based ANN are compared with three types of ANNs. The ANNs algorithms are as Levenberg-Marquardt (LM), Bayesian Regularization (BR), and Scaled Conjugate Gradient (SCG). The evaluation criteria in terms of mean square error (MSE), and coefficient of determination (R^2). The data are collected from the sensors of current, voltage and solar radiation in experimental system which about 8000 samples to build the proposed forecasting model of solar power generation. These data are divided into 70% for training, 15% for validation, and 15 % for testing. The ANN forecasting models have three inputs including open circuit voltage, short circuit current, and irradiance and one output for generated power. The results indicate a high value of R^2 around 0.99 for LM, BR and SCG. However, the results show that the BR is the best ANN model with lowest MSE about 0.37 in comparing with LM and SCG with MSE about 0.4 and 2.48 respectively. Besides, the numerical and plotting results for prediction data show the superior of BR technique over other two ANN algorithm. | ||
| Keywords | ||
| Solar Power Prediction; Dual-Axis Solar Tracking; FPGA; ANN; Experimental Validation | ||
|
Statistics Article View: 245 PDF Download: 174 |
||
