Air-conditioning system, machine learning device, and machine learning method

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing air-conditioning systems face high workload in optimizing device operating capacity post-installation, requiring extensive data measurement to construct accurate models for energy efficiency and comfort.

Innovation Solution

A machine learning apparatus and method that acquires state variables from both outside and inside air conditioning devices, performs reinforcement learning to associate these variables with energy consumption, and calculates rewards to optimize the operating capacity of outside and air conditioning devices, reducing the need for extensive post-installation data measurement and workload.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If a model is constructed to determine operating capacity of air conditioning devices, then energy efficiency and comfort are improved, but post-installation workload increases due to extensive data measurement requirements

Engineering Contradiction:
Improveenergy efficiencyVSAvoidpost-installation workload
Core Design Contradiction:
Use of energy by moving objectVSEase of operation

Solution Approach 1:

The patent applies preliminary action by pre-collecting and storing operation data during the manufacturing or installation phase before actual use. The data collection device gathers temperature, humidity, and device operation data in advance, creating a ready-to-use database that eliminates the need for extensive post-installation measurements. This allows the model to be constructed and deployed immediately, improving energy efficiency without burdening users with data collection work.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If device characteristics are accurately reflected in the model, then optimization precision is improved, but data measurement and model construction complexity increases

Engineering Contradiction:
Improvemodel accuracyVSAvoidmodel construction complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies self-service by enabling the system to automatically collect, store, and process its own operation data without external intervention. The data collection device autonomously gathers temperature, humidity, and device operation data, and the model construction unit automatically processes this data to create accurate device-specific models. This automation maintains high model accuracy while eliminating the complexity of manual data collection and model building.

Inventive Principle:
Principle #25Self-service

3Loss of energy

If reinforcement learning is used to optimize operating capacity, then energy consumption is reduced, but system complexity and computational requirements increase

Engineering Contradiction:
Improveenergy consumptionVSAvoidsystem complexity
Core Design Contradiction:
Loss of energyVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-collecting comprehensive operation data and constructing training datasets before deploying reinforcement learning. The system gathers historical data on temperature, humidity, and device operations, then uses this pre-processed data to train the reinforcement learning model offline. This approach reduces energy consumption during actual operation while managing system complexity by performing computationally intensive tasks in advance rather than in real-time.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3961115B1Air-conditioning system, machine learning device, and machine learning method
Publication Date: 2023.07.12 DAIKIN INDUSTRIES LTD
  • EP3961115B1 patent drawingFigure 1
  • EP3961115B1 patent drawingFigure 2
  • EP3961115B1 patent drawingFigure 3

AI summary

An air-conditioning system that optimizes operation capacity of an outside air conditioning unit and operation capacity of an air conditioning unit is provided. An air-conditioning system includes an outside air conditioning device, an air conditioning device, and a machine learning apparatus, and includes a state variable acquiring unit configured to acquire state variables including a condition of outside air, a condition of inside air, an operation condition of the outside air conditioning device, an operation condition of the air conditioning device, and a temperature or humidity set for a target space, a learning unit configured to perform learning by associating the state variables with at least either the operating capacity of the outside air conditioning device or the operating capacity of the air conditioning device, and a reward calculating unit configured to calculate a reward that correlates with a total of energy consumption of the outside air conditioning device and energy consumption of the air conditioning device. The learning unit performs the learning by using the reward.