Abstract: When the traditional PID algorithm controls temperature objects with complex dynamic characteristics such as large lag, nonlinearity, and time-varying, there are problems such as large overshoot, inability to self-adjust parameters, poor model adaptive ability, and low system stability. According to the structure characteristics of PID algorithm, using the single output characteristic of T-S type fuzzy neural network, this paper established a triple network model that can output three parameters of PID separately, and proposed multiple T-S type fuzzy neural network PID temperature control algorithm. In Matlab environment, simulation experiments were carried out and the results show that compared with traditional PID, BP neural network PID, and conventional fuzzy neural network PID, the algorithm in this paper has the lowest overshoot, the highest stability, the strongest model adaptability, the best anti-interference ability, and the best comprehensive performance index.