Machine learning device, demand control system and air-conditioner control system

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing air-conditioning systems lack efficient methods to predict and control power consumption to meet predetermined demand target values, leading to inefficiencies in energy conservation.

Innovation Solution

A machine learning device that learns set temperatures and control parameters using reinforcement learning to optimize air-conditioner operation, predicting power consumption and ensuring it does not exceed a predetermined limit, thereby improving energy conservation performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional air-conditioning control methods are used, then the system operates without sophisticated prediction capabilities, but the power consumption cannot be accurately predicted or controlled to meet demand target values

Engineering Contradiction:

Improvepower consumption prediction accuracyVSAvoidcontrol system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces a machine learning device as an intermediary component between the air-conditioning system and the control mechanism. This device includes a learning unit with neural networks that process operation records and environmental data to predict power consumption and determine optimal control parameters, thereby achieving accurate prediction without directly modifying the air-conditioning hardware

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces conventional mechanical control methods with an information-processing-based machine learning system. Instead of using traditional control algorithms and mechanical adjustments, the system uses neural networks to learn from historical data and automatically determine control parameters, substituting physical control mechanisms with intelligent software-based control

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Loss of energy

If the air-conditioner operates without learned control parameters, then the control system is simpler, but the energy conservation performance is insufficient

Engineering Contradiction:

Improveenergy conservation performanceVSAvoidautomatic control capability

Core Design Contradiction:

Loss of energyVSExtent of automation

Solution Approach 1:

The machine learning device performs self-learning by automatically processing operation records and environmental data through the learning unit. The system improves its control capabilities over time without external intervention, automatically updating the neural network parameters to optimize energy conservation performance based on accumulated data

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system implements feedback by using evaluation data including actual power consumption to update the learning state of the learning unit. The control parameters are continuously refined based on the difference between predicted and actual power consumption, creating a closed-loop system that improves energy conservation through iterative learning

Inventive Principle:

Principle #23Feedback

3Adaptability or versatility

If the system uses historical operation records for control, then the control approach is simpler, but it cannot adapt to changing environmental conditions and optimize power consumption effectively

Engineering Contradiction:

Improveadaptation to environmental conditionsVSAvoidair-conditioner power consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by stationary object

Solution Approach 1:

The patent transforms the static control approach into a dynamic learning system. The neural network in the learning unit continuously adapts to changing environmental conditions by processing real-time data and updating control parameters, allowing the system to respond dynamically to variations in outdoor temperature, humidity, and usage patterns while optimizing power consumption

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS12535239B2Machine learning device, demand control system and air-conditioner control system

Publication Date: 2026.01.27 DAIKIN INDUSTRIES LTD

AI summary

A machine learning device learns a set temperature, in a target space, in order to achieve a demand target value that is an upper-limit value of a power consumption of an air conditioner, installed in the target space, in a predetermined period. The machine learning device includes a learning unit, first and second acquisition units, and an updating unit. The first acquisition unit acquires a first variable including at least one of the power consumption of the air conditioner and an indoor state value correlated with a state in the target space. The second acquisition unit acquires evaluation data useable to evaluate a control result of the air conditioner. The updating unit updates, by using the evaluation data, a learning state of the learning unit. The learning unit performs learning in accordance with an output of the updating unit. The evaluation data includes the power consumption of the air conditioner.