The present work aims to analyze the elements that affect corporate green technology innovation and investigate a method suitable for predicting and evaluating corporate performance. First, the elements of green technology innovation and their relationships are analyzed and explained. Then, the Complex Adaptive System (CAS) theory is introduced. On this basis, a computer model for the driving mechanism system of corporate green technology innovation is constructed on the Recursive Porus Agent Simulation (Repast) platform. Finally, the Backpropagation Neural Network (BPNN) model is optimized by Particle Swarm Optimization (PSO), constituting the PSO-BPNN algorithm to evaluate corporate performance. The results of network training and simulation demonstrate that compared with traditional BPNN, PSO-BPNN achieve a faster convergence speed and fewer errors. Besides, the actual output value has a tiny difference from the expected value, showing the application potential of this algorithm in corporate performance prediction. Moreover, the driving factors of green technology innovation greatly affect the profitability and performance of enterprises. Given insufficient corporate profit margin, continuous technological innovation activities can ensure the normal operation of enterprises. A smaller corporate tax rate can shorten the time for the system to reach equilibrium. When the corporate tax rate is above 0.2, the system takes longer to reach equilibrium. In addition, the public opinion coefficient directly affects the time needed for the system to attain equilibrium. When the public opinion coefficient is within 50,00 ~ 6,000 interval, the time that the system takes to reach equilibrium changes significantly. Furthermore, corporate internal and external driving factors have a direct effect on corporate green technology innovation and performance. The research findings indicate that the PSO-BPNN algorithm is of vital practical value to corporate performance evaluation.