Study On PID Neural Network Temperature Control Algorithm

With the rise of the information industry and the warm trend for the PV power generation by new energies, it has great demand to the big crystals with high quality, and this means higher requirements should be made to precisely control the temperature in which the crystal grows. Due to the characteristics that the heating system has large delay and inertia during the crystal growth, the PID controller, which has been commonly used in the industry due to its simple control and higher reliabilities, cannot achieve the satisfactory controlling effects. The design for the temperature controller with higher precision is based on the full use of the PID control advantages and the artificial neural network which has self-learning and adaptive abilities and also can come close to the nonlinear function by any precision. Therefore, this has great significances both in theory and practice.The dissertation has comprehensively analyzed the PID neural network control algorithm. It starts with the brief introduction to the advantages and disadvantages of the traditional and regular PID controller's control algorithm, and then the improved algorithm to the regular PID controller has been put forward, and also the advantages and disadvantages have been analyzed. The dissertation has also carried out the comprehensive research to the artificial neural networks. Detailed discussion has been made to the three decisive factors of the artificial neural network for processing information, and they are the activation function of artificial neurons, the connection method of neural network and the learning rules of neural network. Based on the research, PID neural network has been put forward, which is the combination of the PID control algorithm and the artificial neural network. By analyzing the PID neural network, the PID neural network structure and algorithm which are suitable for the hardware to realize have been confirmed, and also the PID neural network has been simulated by using the engineering simulation software, MATLAB. The simulation has revealed that PID neural network has better control performance compared with PID control.Finally, the dissertation has explored the hardware realization of the PID neural network by using the temperature control system of the crystal growth furnace. The algorithm of the PID neural network has been improved during the exploration, and only multiplication, addition and delay operation have been used. The improved PID neural network is more convenient for the digital signal processing, its hardware realization is much easier, and also the parallel structure of the PID neural network has been realized by using FPGA (Field Programmable Gate Array). The PID neural network control module can carry out the modular design by using the Altera company's IP core which is compatible with the 32 bit IEEE754 single precision floating-point process unit, and also it can compile and do the timing simulation experiment by using Altera company's QuartusII8.0 software. Both the theoretical analysis and the simulation results have revealed that the PID neural network temperature control module has the higher control precision and the more effective response, and it can improve the temperature control precision and the heating efficiency of the crystal growth furnace, and also it is more suitable for the real-time control. Therefore, the PID neural network temperature control module has fully achieved the design purpose.