基于梯度的黑箱方法用于提升逆变器构建电站的稳定裕度而不暴露控制器细节

Inverters in a renewable-energy-generation based power station (PS) may be produced by different manufacturers whose control schemes cannot be exposed to each other. Hence, the system-level stability prediction and stability-margin-improvement-oriented parameter tuning should be completed in a black-box manner. Based on the Nyquist theorem and the Routh–Hurwitz criterion, a gradient-based black-box (GBBB) modeling method is proposed for the inverters in the PS. The GBBB model is a black-box encrypted function with tunable control parameters as inputs. Its outputs contain impedance values, open-loop-stability factors, and parameter-participation-factors of the impedance for Nyquist-based stability judgment and margin-improvement-oriented parameter tuning of the PS. Based on the GBBB models of the inverters, the PS's stability margin is described by black-box cost functions. Then, a gradient-descent based parameter tuning method is proposed where the gradients are calculated using the outputs of the GBBB models according to the chain rule. All the inverters’ parameters can be optimized in theory iteratively to improve the PS's stability margin. Under the given workflow, only the GBBB models and the PS topology are needed, which means the control details of the inverters are unexposed, i.e., the intellectual property would not be infringed.