Data-Driven Heat Exchanger Control via Neural Network Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing control methods for heat exchangers in air-conditioning systems face challenges due to complex differential algebraic equations (DAEs) that are difficult to solve, requiring large amounts of data for model construction and being computationally intensive, especially for large and complex systems, which limits real-time control and scalability.
Innovation Solution
A data-driven control system using a deep state-space modeling framework with a combination of convolutional neural networks (CNNs) and gated recurrent units (GRUs) to simulate heat exchanger dynamics, reducing the need for explicit DAEs and allowing for efficient real-time control by learning relationships between input and output data without requiring detailed knowledge of system geometry or operating conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physics-based modeling with differential algebraic equations (DAEs) is used to predict heat transfer and fluid flow dynamics, then the model accuracy is improved, but the computational complexity increases significantly and the system becomes difficult to solve
Solution Approach 1:
The patent replaces the physics-based mechanical/mathematical model (DAEs) with a data-driven neural network model. The neural network learns heat transfer and fluid flow dynamics directly from operational data, substituting the complex analytical modeling approach with a computational learning approach that avoids solving difficult DAE systems while maintaining prediction accuracy.
Solution Approach 2:
The patent creates a digital twin (virtual copy) of the heat exchanger using a neural network model that replicates the physical system's behavior. This virtual model is trained on historical operational data and can predict system dynamics without requiring real-time solution of complex physics equations, thus copying the essential behavior while simplifying computation.
2Reliability
If indirect data-driven control methods are used to construct a model from operational data, then the control policy can be designed with model knowledge, but large quantities of data are required for model construction
Solution Approach 1:
The patent performs preliminary model construction offline by training the neural network on historical operational data before deployment. This preliminary training phase builds the digital twin in advance, allowing the system to operate with stable control policies without requiring large amounts of data during real-time operation. The heavy data processing is done beforehand.
Solution Approach 2:
The patent changes the parameter representation from requiring large volumes of raw operational data to using a trained neural network model with fixed weights and biases. Once trained, the model consumes minimal data during operation, transforming the data requirement from a continuous large-volume need to a one-time training requirement followed by efficient inference.
3Reliability
If model-based control design is used to ensure control stability and handle constraints, then the control performance is improved, but the model construction requires large amounts of data and computational resources
Solution Approach 1:
The patent substitutes the computationally intensive model construction process with a data-driven neural network training process that can be performed offline. The resulting model requires minimal computational resources during real-time control operation, replacing the need for continuous heavy computation with a lightweight inference process that maintains control stability.
4Quantity of substance
If direct data-driven control methods are used to construct control policies from operational data, then fewer data are required, but difficulties arise in handling state and input constraints for safe operation
Solution Approach 1:
The patent introduces a neural network model as an intermediary between the raw operational data and the control policy. This intermediate model learns from data and encapsulates system dynamics, making it easier to handle constraints during control optimization. The model serves as a bridge that translates complex data relationships into a form suitable for constrained control design.
Data Source
AI summary
A system for controlling an operation of an air-conditioning system including a heat exchanger is provided. The system comprises a processor that executes a neural network trained to simulate an operation of the heat exchanger for a test control input, to produce an output of the simulation based on historical data defining a state of the heat exchanger. The historical data includes a sequence of historical control inputs provided to the heat exchanger and a sequence of historical outputs of the operation of the heat exchanger corresponding to the sequence of historical control inputs. The processor determines a control command to the air-conditioning system based on the predicted test output of the simulation of the operation of the heat exchanger for the test control input and transmits the determined control command to an actuator of the air-conditioning system.