Building Management System Identification Model Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing HVAC control systems face challenges in accurately modeling complex, nonlinear building systems for predictive control due to the complexity of physical phenomena involved, leading to inefficiencies in heating and cooling processes.
Innovation Solution
A building management system that includes a system identification process to generate a predictive model by optimizing a prediction error function, discarding initial guesses that violate physical laws or lead to unstable systems, and using a multi-step ahead prediction error method to identify model parameters and Kalman gain parameters for improved control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a system identification process is used to generate a predictive model for HVAC control, then control accuracy is improved, but computational time and resources increase due to the complexity of optimizing the prediction error function
Solution Approach 1:
The patent applies preliminary action by generating multiple initial guesses of model parameters before the main optimization process. This allows the system to pre-screen potential solutions and eliminate poor candidates early, reducing the computational burden of the full optimization process while maintaining accuracy in identifying the optimal parameters for the predictive model.
Solution Approach 2:
The patent segments the parameter identification process into distinct phases: generating multiple initial guesses, evaluating each guess against criteria, and selectively refining promising candidates. This segmentation divides the complex optimization problem into manageable stages, reducing overall computational time while preserving the ability to achieve high control accuracy.
2Reliability
If multiple initial guesses are evaluated to ensure model stability and physical law compliance, then model reliability is improved, but the complexity of the system identification process increases
Solution Approach 1:
The patent uses preliminary action by establishing clear evaluation criteria before the optimization process begins. Multiple initial guesses are generated and screened against these pre-defined criteria (physical law compliance, stability requirements) to filter out inadequate candidates early, ensuring reliable models without requiring complex real-time evaluation during optimization.
Solution Approach 2:
The patent applies parameter changes by systematically varying initial parameter guesses and evaluating their impact on model stability and physical law compliance. This structured approach to parameter exploration allows the system to identify reliable models through controlled experimentation rather than complex adaptive mechanisms.
3Measurement precision
If the optimization process runs to local optimality for each initial guess, then parameter identification accuracy is improved, but computational resources are excessively consumed
Solution Approach 1:
The patent applies partial action by running the optimization process to local optimality only for selected initial guesses that meet predetermined criteria, rather than for all possible guesses. This selective approach achieves sufficient parameter identification accuracy for the most promising candidates while avoiding the excessive computational resource consumption that would result from exhaustive optimization of all initial guesses.
Solution Approach 2:
The patent uses preliminary screening to identify which initial guesses warrant full optimization to local optimality. By pre-evaluating initial guesses against stability and physical law criteria, the system determines in advance which candidates deserve the computational investment of complete optimization, thereby reducing overall resource consumption while maintaining accuracy for the best solutions.
Data Source
AI summary
A building management system includes building equipment operable generate training data relating to behavior of a building system and a controller configured to perform a system identification process that includes generating a prediction error function based on the training data and a system model, generating initial guesses of one or more parameters of the system model, running an optimization problem of the prediction error function for a first group of iterations, discarding, after the first group of iterations, a portion of the initial guesses based on one or more criteria and ranking a remaining portion of the initial guesses, running the optimization problem of the prediction error function for a top-ranked initial guess of the remaining portion to local optimality to identify a first set of values of the one or more parameters, and identifying the one or more parameters as having the first set of values.