Abstract:Temperature fluctuations significantly impact the wavelength of vertical-cavity surface-emitting lasers (VCSELs). Traditional PID control in complex environments with abrupt external temperature changes suffers from issues such as large overshoot, weak disturbance rejection, and poor adaptive performance. This paper simulates the three-dimensional steady-state thermal distribution characteristics of the VCSEL temperature control system, establishes a mathematical model for temperature control, and proposes a PID control method optimized using a backpropagation (BP) neural network. Within the MATLAB/Simulink environment, the proposed method is compared with classical PID control and fuzzy PID control through simulation experiments. Results demonstrate that this approach utilizes a BP neural network to achieve online tuning of PID parameters, effectively suppressing system overshoot while significantly enhancing the system's adaptability and robustness to environmental changes. Consequently, overall temperature control performance is improved, meeting the stringent requirements for laser temperature stability demanded by high-precision atomic magnetometers.