Air Conditioner Data Thinning Adapter for AI Learning Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for transmitting operation history data for AI learning in air conditioners result in high communication traffic, which can lead to reduced learning accuracy if data is simply deleted to reduce traffic.
Innovation Solution
An air conditioning system with an adapter that thins operation history data by retaining only data with significant changes, transmitting only these data points to the server, thereby reducing communication traffic while maintaining learning accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all operation history data are transmitted to the server device, then learning accuracy of AI is maintained, but communication traffic becomes large
Solution Approach 1:
The patent extracts only the essential features from operation history data by generating feature quantities that represent characteristic patterns, rather than transmitting all raw data. This allows maintaining learning accuracy while significantly reducing communication traffic.
Solution Approach 2:
The patent transforms raw operation history data into different parameter representations (feature quantities) that capture the essential characteristics needed for AI learning. This parameter transformation reduces data volume while preserving the information necessary for accurate learning.
2Quantity of substance
If operation history data are deleted to reduce communication traffic, then communication traffic is reduced, but learning accuracy may be lowered
Solution Approach 1:
Instead of deleting data, the patent extracts essential features from the data. This extraction process identifies and retains only the characteristic patterns needed for learning, eliminating redundant information while preserving learning accuracy.
Solution Approach 2:
The patent applies partial action by transmitting only the necessary portion of data (feature quantities) rather than all raw data. This partial transmission is sufficient for maintaining learning accuracy while reducing communication traffic.
3Quantity of substance
If feature quantity generation is performed on operation history data, then data amount is reduced, but data processing complexity increases
Solution Approach 1:
The patent segments the data processing task into two parts: feature extraction at the air conditioner (reducing data amount) and learning model generation at the server (processing reduced data). This segmentation distributes complexity appropriately and reduces overall data processing burden.
Solution Approach 2:
The patent introduces feature quantities as an intermediary representation between raw operation history data and the AI learning model. This intermediary form reduces data complexity while preserving essential information needed for learning.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An air conditioner includes an adapter including a control unit that controls the air conditioner using a learning model and a communication unit that communicates with a server device that generates the learning model on the basis of operation history data of the air conditioner. The adapter includes an acquisition unit, a determination unit, an erasing unit, and a transmission unit. The acquisition unit acquires the data every predetermined cycle from the air conditioner. The determination unit determines whether or not an amount of change between temporally continuous data of the data acquired by the acquisition unit is within a predetermined range. The erasing unit leaves at least one of the continuous data and deletes the other data in a case where the amount of change between the continuous data is within the predetermined range. The transmission unit transmits the data after being deleted by the erasing unit to the server device. It is possible to reduce traffic of data used for learning of artificial intelligence (AI) while maintaining learning accuracy of the AI.