Simulated experiments on diesel engine speed control based on LSTM-PID control

The electronic speed regulator, as a key component of the diesel engine, governs the engine's speed, which is crucial for its operational performance. In response to the issue of PID parameter tuning for speed control under different operating conditions, this paper proposes a PID parameter self-tuning method based on Long Short-Term Memory (LSTM) neural networks. Due to the complex time-varying and nonlinear characteristics of diesel engines, conventional PID algorithms may not perform as well as the LSTM-PID strategy. Therefore, an intelligent algorithm is needed to dynamically optimize the three variables of the PID algorithm. The adaptive learning algorithm combines the conventional PID control strategy with neural networks. Firstly, the conventional PID controller directly regulates the rack displacement of the electromagnetic actuator in a closed-loop control manner, ensuring the controllability of the control system within a feasible range. Secondly, in another closed-loop, the three control parameters of the speed loop PID can be adjusted and optimized based on the learning of the neural network, enhancing the overall control performance to its best. Finally, the effectiveness of the proposed method is tested through simulation comparisons. Through these methods, the paper aims to optimize the control performance of diesel engines by integrating PID control with neural networks, particularly LSTM, to adaptively adjust parameters under varying operating conditions.