Model-free wind farm power production optimization using multi-resolution optimized relative step size random search

• Main Author: Mok, Ren Hao
• Format: Thesis
• Language: English
• Published: 2019
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://umpir.ump.edu.my/id/eprint/29309/1/Model-free%20wind%20farm%20power%20production%20optimization%20using%20multi-resolution%20optimized%20relative%20step%20size%20random%20search.wm.pdf; http://umpir.ump.edu.my/id/eprint/29309/
• Institution: Universiti Malaysia Pahang

This study investigates the performance of Multi-Resolution Optimize Relative Step Size Random Search (MR-ORSSRS) based method in maximizing the total power production of wind farms. The performance is investigated based on the Horns Rev wind farm layout which consists of 80 wind turbines under the case studies of different wind directions at 170°, 200°, 220°, 240°, 250° and 270°, five wind turbines failures and non-static wind variations. The implementation of Multi-Resolution (MR) function is used to improve the convergence speed of the Optimize Relative Step Size Random Search in the case of maximizing the total power production of a wind farm in real-time optimization. The MR function is significant in improving the convergence speed since this approach exploits the dimension of the design parameter using several optimization stages. In particular, it firstly adopts a small size of design parameter tuning followed by a bigger size of design parameter tuning in the following stages. Therefore, it is expected that less computation effort is required to obtain the optimal design parameter. Even though the Multi-Resolution Stochastic Perturbation Simultaneous Approximation (MR-SPSA) is developed to solve the real-time high-dimensional problem with faster convergence, the obtained total power production of the wind farm is still not optimum. This is because the SPSA is a memory-less structure type optimization that limits the storage of the best design parameter. Alternatively, ORSSRS based method is a memory type optimization structure. Hence, it can store the best design parameter value while producing consistent objective function. However, the ORSSRS based method alone does not have the sufficient convergence speed to optimize wind farm problem in real time. Therefore, the MR function is implemented to improve the convergence speed of the ORSSRS based method. In this study, the performance of MR-ORSSRS based method is compared with MR-SPSA based method in terms of the convergence speed, accuracy, and robustness in maximizing the total power production of Horns Rev wind farm. The results show that MR-ORSSRS based method outperforms the benchmark MR-SPSA based method in terms of the convergence speed of all the study cases. In particular, it can improve the convergence speed of incoming wind direction at 170°,200°,220°,240°,250° and 270° by 88.89%,88.89%,41.66%,88.89%,88.89% and 66.67%, respectively. However, in the case of the five wind turbine failures, the speed of the incoming wind direction is 66.67%. Moreover, the MR-ORSSRS based method produces better total power production for wind direction at 170°,200°,220°,240° and 270°, as well as wind turbines failures compared to the MR-SPSA based method. In term of the convergence speed, the MR-ORSSRS based method produces higher convergence speed for all the wind direction cases even in the wind turbines failure cases. Hence, it is proven that the proposed MR-ORSSRS based method is effective in producing better total power production with faster convergence speed even with turbines failure and time-varying wind compared to the benchmark MR-SPSA based method.