This paper proposes an intelligent control method based on the Camel Marching Optimization Algorithm (CJOA) for the adaptive adjustment of control parameters of grid structured photovoltaic inverters under dynamic impedance changes in the power grid. By establishing a control mathematical model for the inverter and combining it with a convolutional neural network (CNN) to construct a nonlinear mapping model between grid impedance and control parameters, and using CJOA to optimize CNN hyperparameters, the dynamic response capability and stability of the system have been significantly improved. The simulation results show that this method reduces voltage recovery time by 43.2% and steady-state error by 52.6% during sudden changes in grid impedance, which is superior to traditional PID and unoptimized CNN methods. The research results provide effective solutions for the stable operation of photovoltaic inverters in complex power grid environments, and have important engineering application value.