Equipment edge controller with reinforcement learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building management systems (BMS) face challenges in efficiently controlling building equipment at the edge due to limited computing power and memory in edge controllers, leading to suboptimal control of parameters such as damper positions and fan speeds, which affects the achievement of setpoints for temperature and energy efficiency.
Innovation Solution
Implementing a reinforcement learning model on edge controllers that uses temperature setpoints and weather forecasts to control damper positions, with automatic updates based on indoor air temperature differences and occupancy, and integrating feature generation from streaming data to optimize mixed air temperature.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If advanced control approaches are used to improve temperature regulation and energy efficiency, then control precision is improved, but device complexity increases due to limited computing power in edge controllers
Solution Approach 1:
The control system is segmented into a centralized BMS that performs computationally intensive advanced control calculations and edge controllers that execute simplified control routines. This segmentation allows advanced control algorithms to improve temperature regulation precision while the edge controllers maintain simplicity with limited computing power.
Solution Approach 2:
A communication interface acts as an intermediary between the centralized BMS and edge controllers. The BMS sends control parameters and setpoints to edge controllers, which then execute local control actions. This intermediary approach enables sophisticated control strategies without overloading the limited edge controller hardware.
2Loss of energy
If more sophisticated control algorithms are implemented at the edge, then energy efficiency is improved, but computing power requirements increase beyond edge controller capabilities
Solution Approach 1:
Computational tasks are segmented between the centralized BMS and edge controllers. The BMS performs energy-intensive optimization calculations using advanced algorithms, while edge controllers execute lightweight control routines. This segmentation achieves energy efficiency improvements without requiring edge controllers to have high computing power.
Solution Approach 2:
The edge controller receives pre-calculated control parameters and setpoints from the centralized BMS, effectively copying the results of sophisticated energy optimization algorithms. This allows the edge controller to implement energy-efficient control without needing to perform the complex calculations itself.
3Manufacturing precision
If local control sophistication is increased to achieve BMS setpoints, then temperature regulation is improved, but memory requirements exceed edge controller capacity
Solution Approach 1:
Complex control logic and large data structures are extracted from the edge controller and relocated to the centralized BMS. The edge controller retains only essential control functions and minimal local memory for immediate operational parameters. This extraction enables accurate setpoint achievement while respecting edge controller memory limitations.
Solution Approach 2:
The edge controller stores copied versions of essential control parameters and setpoints received from the BMS, rather than maintaining complete control algorithms locally. This copying approach provides sufficient information for accurate temperature regulation without requiring large memory capacity at the edge controller.
Data Source
AI summary
A method for controlling a unit of building equipment includes training a reinforcement learning model to replicate outputs of a model predictive control algorithm running in a simulation. The model predictive control algorithm determines simulated control values which optimize an objective associated with running a simulated version of the unit of building equipment in the simulation. The method further includes generating a control value for an internal parameter of the unit of building equipment using the reinforcement learning model running on an edge controller. The method further includes controlling, by the edge controller, the unit of building equipment in accordance with the control value generated using the reinforcement learning model.