Deep Reinforcement Learning Based Optimization Strategy for Hydro-Governor PID Parameters to Suppress ULFO
2020
In this paper, the mechanism of ultra-low frequency oscillation (ULFO) is studied according to the damping torque theory and the influence of proportional, integral and differential parts (PID) parameter settings of hydro-governor on ULFO is investigated. After that, a deep Q network (DQN) based method is proposed for the hydro-governor PID parameter settings self-tuning, which considering the uncertainty of the hydraulic turbine operating conditions. Simulation results show that the proposed DQN-based method can provide the optimal parameters for the hydro-governor PID under all operating conditions, which achieves better performance in preventing ULFO in comparison with other methods.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
13
References
1
Citations
NaN
KQI