Transient stability assessment of wind turbine grid-connected systems using ISSA-MSVR

To address the issue of inaccurate transient stability analysis in power systems after wind farm integration， this paper proposes a method combining improved sparrow search algorithm （ISSA）-optimized multi-output support vector regression （MSVR）. First， an energy function is established for direct-drive wind turbines integrated into the grid. For wind turbine grid-connected systems， an interpretable stability energy function is constructed using an improved Deep Q-Network （DQN） algorithm. The unstable equilibrium points （UEPs） of the system are then determined via the boundary of stability region based controlling UEP method （BCU）， generating the training and testing datasets for the prediction model. Next， to overcome the limitations of the SSA algorithm， such as susceptibility to local optima， inverse learning， piecewise weighting， and Cauchy mutation are introduced to enhance SSA. The ISSA is then employed to optimally tune the penalty factor and kernel width in MSVR. The proposed method is validated on a modified IEEE 39-bus system. Experimental results demonstrate that the ISSA-MSVR approach achieves smaller prediction errors and reduced training time compared to other state-of-the-art AI methods， effectively predicting the UEPs in wind farm-integrated systems.