Gas Container Refill Timing Using Adaptive Remaining-Gas Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for predicting remaining gas amounts in gas containers require high-frequency data processing, leading to increased costs and processing time, especially when using machine learning, due to the need for enhanced server specifications and extensive data processing.
Innovation Solution
A remaining gas amount management system that adjusts the frequency of predicting remaining gas amounts based on predetermined trigger information, such as replacement cycles, elapsed time, gas consumption, or informing events, allowing for reduced server load and processing time by calculating at lower frequencies before specific triggers are met.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-frequency prediction is performed to improve accuracy, then prediction accuracy is improved, but server cost and processing time increase
Solution Approach 1:
The patent applies dynamics by making the prediction frequency adjustable rather than fixed. The remaining amount calculator dynamically changes the prediction frequency based on the remaining gas amount threshold. When the remaining gas amount is above the threshold, prediction is performed at a lower frequency (e.g., once every 3 days). When the remaining gas amount falls below the threshold, prediction frequency increases (e.g., once daily). This dynamic adjustment resolves the contradiction by adapting the prediction frequency to actual needs, maintaining accuracy when necessary while reducing processing time and server load during normal operation.
Solution Approach 2:
The patent applies parameter changes by modifying the prediction frequency parameter based on the remaining gas amount condition. The system changes the time interval parameter between predictions depending on whether the remaining gas amount is above or below a predetermined threshold. This parameter adjustment allows the system to balance between prediction accuracy and processing efficiency, resolving the contradiction between high-frequency prediction accuracy and acceptable processing time.
2Measurement precision
If high-frequency prediction is performed to improve accuracy, then prediction accuracy is improved, but server specifications must be increased, leading to higher cost
Solution Approach 1:
The patent applies dynamics by making the prediction frequency adjustable rather than fixed. The remaining amount calculator dynamically changes the prediction frequency based on the remaining gas amount threshold. When the remaining gas amount is above the threshold, prediction is performed at a lower frequency (e.g., once every 3 days). When the remaining gas amount falls below the threshold, prediction frequency increases (e.g., once daily). This dynamic adjustment resolves the contradiction by adapting the prediction frequency to actual needs, maintaining accuracy when necessary while reducing processing time and server load during normal operation.
Solution Approach 2:
The patent applies parameter changes by modifying the prediction frequency parameter based on the remaining gas amount condition. The system changes the time interval parameter between predictions depending on whether the remaining gas amount is above or below a predetermined threshold. This parameter adjustment allows the system to balance between prediction accuracy and processing efficiency, resolving the contradiction between high-frequency prediction accuracy and acceptable processing time.
3Measurement precision
If machine learning is used for prediction, then prediction capability is improved, but creation of input processing information and prediction model requires enormous cost and processing time
Solution Approach 1:
The patent applies partial action by not continuously performing full machine learning prediction processing at high frequency. Instead, the system performs simplified threshold-based frequency adjustment, which is a partial form of prediction that suffices for most conditions. Full machine learning model creation and processing are only intensively performed when necessary (when remaining gas amount falls below threshold), rather than continuously. This partial approach reduces the enormous cost and processing time of model creation while maintaining adequate prediction capability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A remaining gas amount management system includes: a flow rate measurer that measures a flow rate of a gas flowing to a gas appliance from a gas container located at a consumer's place or at each of a plurality of consumer's places; a remaining amount calculator that calculates a predicted value by using past data indicating past gas consumption in a first predetermined period, the predicted value representing a remaining gas amount of the gas container after a second predetermined period has elapsed; and an informer that informs of a replacement timing of the gas container based on the predicted value or a replacement cycle of the gas container. Based on predetermined prediction frequency change trigger information, the remaining amount calculator changes a frequency at which the predicted value is calculated.