Self-Tuning Regulator With Prediction Error Filter
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional self-tuning regulators for closed-loop control systems face challenges in estimating plant parameters due to persistent excitation issues, requiring complex computational algorithms unsuitable for low-cost ASIC implementation, especially in switch-mode power supplies.
Innovation Solution
A closed-loop control system with a self-tuning regulator using a prediction error filter and least-mean-squares filter for adaptive parameter modification, allowing for low computational complexity and cost-effective ASIC implementation, overcoming persistent excitation through pseudo-random disturbances and adaptive gain adjustment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional self-tuning regulators use complex system identification algorithms to estimate plant parameters accurately, then parameter estimation precision is improved, but device complexity increases making ASIC implementation infeasible
Solution Approach 1:
The patent transforms the complex system identification problem into a simpler parameter estimation problem by changing the mathematical parameters used. Instead of using complex differential equations and system identification algorithms, the invention uses autoregressive model parameters that can be estimated through simpler correlation-based methods, enabling accurate parameter estimation with reduced computational complexity suitable for ASIC implementation
Solution Approach 2:
The patent replaces complex mechanical/computational system identification mechanisms with a simplified mathematical approach based on autoregressive modeling. By substituting the traditional system identification machinery with correlation-based parameter estimation, the invention achieves the same goal with significantly reduced computational burden
2Reliability
If closed-loop operation is used for system identification, then loop regulation is maintained, but persistent excitation is insufficient impeding identification accuracy
Solution Approach 1:
The patent applies preliminary action by pre-whitening the input signal before correlation computation. This preliminary processing step transforms the colored noise into white noise, ensuring sufficient spectral content across all frequencies without requiring persistent excitation during closed-loop operation, thereby maintaining both regulation stability and identification accuracy
Solution Approach 2:
The patent introduces an intermediary processing step (pre-whitening filter) between the raw input signal and the correlation computation. This intermediary transformation enables accurate parameter estimation under closed-loop conditions by compensating for the insufficient persistent excitation without breaking loop regulation
3Measurement precision
If open-loop operation is used for system identification, then persistent excitation is sufficient, but loop regulation breaks down causing transients that corrupt the system
Solution Approach 1:
The patent maintains feedback by operating in closed-loop mode throughout the parameter estimation process. The pre-whitening technique enables sufficient excitation while the feedback loop remains active, preventing the stability issues and transients that would occur with open-loop operation
4Adaptability or versatility
If certainty equivalence principle is applied to design controller parameters, then self-tuning is achieved, but computational complexity increases for accurate parameter convergence
Solution Approach 1:
The patent changes the mathematical parameters from complex controller coefficients to autoregressive model parameters. This parameter transformation enables the certainty equivalence principle to be applied with much simpler computation, as the autoregressive parameters can be estimated through correlation methods rather than complex optimization algorithms
Data Source
AI summary
An adaptive control system is described. The system includes a control having a plurality of control parameters, the control parameters providing for control of an associated plant. The control parameters are tuned using a prediction error filter, the prediction error filter selecting values of the control parameters that minimize the values of a prediction error between actual and predicted values of an autoregressive process.