Adaptive DC-DC Power Converter with Machine-Learning Regulator
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing DC-DC power converters face challenges in maintaining stable regulation over a wide output range and adapting to varying load profiles and component variations due to manufacturing deviations and aging, especially in applications with unpredictable power demands like modern processors and VR controllers.
Innovation Solution
A power converter system that incorporates a regulator with adaptive PID coefficients determined by a machine-learning predictor, using collected operating points to update parameters in real-time, accounting for input and output parameters, and passive component variations, allowing for improved fitting across a wide range of operating conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fixed PID coefficients are used in the regulator, then the power converter can be manufactured with standard precision components, but the converter cannot maintain stable regulation over a wide output range and cannot adapt to component variations from manufacturing and aging
Solution Approach 1:
The patent applies dynamics by making the PID coefficients variable rather than fixed. The regulator dynamically adjusts the kp-, ki- and kd-coefficients based on real-time operating conditions (input voltage, input current, output voltage, output current) to maintain stable regulation across wide output ranges and adapt to component variations from manufacturing and aging.
Solution Approach 2:
The patent implements feedback by continuously monitoring operating parameters (input voltage, input current, output voltage, output current) and using this information to adjust the PID coefficients. The regulator receives feedback about the actual operating point and modifies the control parameters accordingly to maintain optimal performance under varying conditions.
2Adaptability or versatility
If PID coefficients are adjusted in real time based on operating parameters, then the converter can adapt to varying load profiles, but the complexity of the regulator increases
Solution Approach 1:
The regulator performs self-service by automatically adjusting its own PID coefficients based on monitored operating parameters. The system uses its own operating data (input voltage, input current, output voltage, output current) to autonomously determine the appropriate control parameters without requiring external intervention or complex external control systems.
Solution Approach 2:
The patent applies parameter changes by modifying the PID coefficients (kp-, ki- and kd-coefficients) based on operating conditions. The regulator changes these control parameters according to the detected operating point, allowing adaptation to varying load profiles while maintaining a relatively simple regulator structure that only requires monitoring basic electrical parameters.
3Speed
If the power converter is designed to meet transient load profile specifications, then it can handle short transition times and large load steps, but it cannot simultaneously optimize for all operating conditions across the wide output range
Solution Approach 1:
The patent applies dynamics by making the PID coefficients variable rather than fixed. The regulator dynamically adjusts the kp-, ki- and kd-coefficients based on real-time operating conditions (input voltage, input current, output voltage, output current) to maintain stable regulation across wide output ranges and adapt to component variations from manufacturing and aging.
Solution Approach 2:
The patent applies parameter changes by modifying the PID coefficients (kp-, ki- and kd-coefficients) based on operating conditions. The regulator changes these control parameters according to the detected operating point, allowing adaptation to varying load profiles while maintaining a relatively simple regulator structure that only requires monitoring basic electrical parameters.
Data Source
AI summary
A power converter comprises a regulator, a value-supply system arranged for collecting at least one operating point of the power converter, and a predictor operative to produce updated regulator parameters (such as one or more power supply coefficients) implemented by the regulator to produce an output voltage to power a load. The updated regulator parameters are determined using a process based on the at least one collected operating point samples and predictor parameters obtained from a machine-learning process.