Artificial Intelligence-Based Power Module Control Method and Device

By employing an AI-based power module control method that combines deep learning and reinforcement learning, the power module control strategy is optimized, solving the adaptability problem of traditional control methods under complex operating conditions and achieving adaptive and efficient control of the power module.

CN121763781BActive Publication Date: 2026-05-26CHENGDU MILLIMETER WAVE TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Patents(China)
Current Assignee / Owner
CHENGDU MILLIMETER WAVE TECH CO LTD
Filing Date
2026-03-02
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

Traditional power module control methods are difficult to adapt to nonlinear changes under complex operating conditions, resulting in output voltage overshoot, excessively long adjustment time, and system instability.

Method used

An AI-based power module control method is adopted. Deep learning samples are constructed by collecting sample operating parameters of the power module, a deep learning model is trained, and a reinforcement learning agent is used for real-time control. The model parameters are optimized through an improved search algorithm to achieve adaptive control.

Benefits of technology

It improves the stability and response speed of the power module, enabling it to better cope with load disturbances and input fluctuations, and achieve efficient and precise power control.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121763781B_ABST
    Figure CN121763781B_ABST
Patent Text Reader

Abstract

This application discloses a power module control method and device based on artificial intelligence, relating to the field of power control technology. The method includes: collecting sample operating parameters and corresponding operating states of the power module to construct deep learning samples; training a power module operation monitoring model using an improved deep learning algorithm; collecting real-time operating parameters and using the monitoring model to identify the target operating state; mapping the real-time parameters and target state into a state space input to a reinforcement learning agent, and outputting control actions; controlling the power module's operation based on the actions and updating the reward function value to construct historical experience; when the online optimization cycle arrives, extracting historical experience to perform online optimization of the reinforcement learning agent; otherwise, proceeding to the next control cycle. This application accurately identifies the power supply state through deep learning, combines it with reinforcement learning for dynamic control, and utilizes a specific improved search algorithm to optimize model parameters, achieving adaptive and high-efficiency control of the power module and improving the power supply's stability and response speed.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Laser power parameter optimization method based on deep reinforcement learning

    CN120145855A

  • Low-voltage distribution network multi-objective collaborative optimization method and system based on reinforcement learning

    CN121546641A