Central plant control system with computation reduction based on sensitivity analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for predicting thermodynamic states and optimizing operating parameters of HVAC systems are computationally inefficient, especially in complex arrangements, leading to excessive resource usage and exhaustive computations when determining power consumption for multiple sets of operating parameters.
Innovation Solution
A controller and method that generate gradient data to identify a reduced group of control variables, excluding those insensitive to performance, allowing for efficient operation by predicting states using a non-linear optimizer and determining optimal values for the HVAC system, thereby reducing computational resources and improving performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional non-linear solver is used to predict thermodynamic states for multiple candidate sets of operating parameters, then accurate power consumption prediction is achieved, but computational resources and time are excessively consumed
Solution Approach 1:
The patent segments the computation process into two stages: first performing sensitivity analysis to identify influential control variables, then using gradient data to determine optimal values only for those segmented variables. This divides the original exhaustive computation into manageable parts, reducing overall computation time while maintaining prediction accuracy.
Solution Approach 2:
The patent extracts and identifies the subset of control variables that have significant impact on power consumption through sensitivity analysis. By taking out only these influential variables from the complete set of control variables, the system avoids computing power consumption for variables that would not meaningfully affect the result, thereby reducing computational burden.
2Reliability
If full thermodynamic states are computed by non-linear solver for complex central plant arrangements, then complete system analysis is achieved, but processor usage and memory consumption increase excessively
Solution Approach 1:
The patent extracts only the essential control variables that significantly influence system performance through sensitivity analysis. By computing gradients and identifying influential variables, the system extracts the critical subset of parameters needed for reliable optimization, avoiding unnecessary computation for non-influential variables and thus reducing processor usage.
Solution Approach 2:
The patent applies partial action by computing thermodynamic states and power consumption only for the reduced set of influential control variables identified through sensitivity analysis, rather than performing exhaustive computation for all possible control variables. This partial computation approach maintains sufficient system analysis reliability while significantly reducing processor and memory consumption.
3Productivity
If multiple sets of operating parameters are evaluated through conventional approach, then optimal operating parameters are determined, but computational resources are exhausted
Solution Approach 1:
The patent extracts the critical control variables through sensitivity analysis by computing gradients of power consumption with respect to each control variable. This extraction process identifies which variables warrant detailed evaluation, allowing the system to focus computational resources on the essential subset of parameters and achieve optimization without exhausting computational resources.
Solution Approach 2:
The patent changes the approach from evaluating all control variables uniformly to selectively evaluating only those variables whose gradients exceed a threshold. By changing the parameter selection criteria based on gradient magnitude, the system achieves efficient optimization with reduced computational resource consumption while maintaining productivity.
Data Source
AI summary
Disclosed herein are related to a system, a method, and a non-transitory computer readable storing instructions for operating an energy plant comprised of heating, ventilation and air conditioning (HVAC) devices. In one aspect, the system generates gradient data indicating a gradient of operating performance of the energy plant with respect to values of a plurality of control variables of HVAC devices. The system determines, from the plurality of control variables, a reduced group of control variables of the HVAC devices based on the gradient data. The system determines a set of values of the reduced group of control variables. The system operates the energy plant according to the determined set of values of the reduced group of control variables.