Air conditioning control device and air conditioning control method
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing air conditioning control devices face challenges in reducing the temporal cost of data collection for machine learning, as they often require substantial data accumulation and prolonged time for data collection, especially when initial data is insufficient.
Innovation Solution
The proposed solution involves an air conditioning control device with an acquisition unit, an augmentation unit, and an update unit that generates augmented data by treating certain times as virtual start times, allowing the machine learning model to be updated using this data, thereby reducing the temporal cost of data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is used to predict start time and update the model with collected data, then prediction accuracy improves, but the time required to collect sufficient data increases
Solution Approach 1:
The system performs preliminary actions by generating augmented data through virtual start times before actual machine learning training is needed. This allows the model to be trained more quickly with synthetically expanded datasets, reducing the waiting time for sufficient data accumulation while maintaining prediction accuracy.
Solution Approach 2:
The system creates copies of existing data by generating augmented datasets through virtual start times. Instead of waiting for unique real-world data to accumulate, the system synthesizes additional training samples by copying and transforming existing operational data, thereby accelerating the data collection process without sacrificing model training quality.
2Reliability
If more data is collected to improve machine learning model performance, then model accuracy improves, but the temporal cost of data collection increases
Solution Approach 1:
The system changes parameters by introducing virtual start times that differ from actual start times. This parameter transformation allows the same physical data to be reused multiple times with different temporal labels, effectively multiplying the training value of each data point and reducing the total time needed to accumulate sufficient training examples for reliable model performance.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
An air conditioning control device (2) includes: an acquisition unit (16) that acquires air conditioning data acquired by an air conditioner, and a start time of the air conditioner predicted by inputting the air conditioning data into a machine learning model; an augmentation unit (14) that generates augmented data by referring to the air conditioning data and the start time acquired by the acquisition unit; and an update unit (15) that updates the machine learning model, by referring to the air conditioning data and the start time acquired by the acquisition unit as well as the augmented data generated by the augmentation unit.