Presence Prediction Using Power Data and Learning Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for predicting the presence or absence of a person in a building based on power consumption data are inaccurate, as they do not account for cases where power usage is high when the person is absent or low when the person is present.
Innovation Solution
A method and apparatus that acquire electric power data and learning data to predict the presence or absence of a person by analyzing the correspondence between past power usage patterns and actual presence information, generating highly accurate presence and absence information for delivery companies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If the probability of presence is calculated based on the past amount of power usage being close to the maximum value, then the calculation is simple, but the prediction accuracy is low because it does not account for cases where power usage is high when absent or low when present
Solution Approach 1:
The patent introduces learning data as an intermediary element that mediates between power usage data and presence/absence determination. The learning data contains historical correspondence relationships between power usage patterns and actual presence states, enabling the system to accurately predict presence even when simple threshold-based calculations would fail. This intermediary layer resolves the contradiction by providing contextual information without complicating the core calculation mechanism.
Solution Approach 2:
The system implements feedback by using learning data that captures historical relationships between power usage and actual presence states. This feedback mechanism allows the prediction system to learn from past patterns and improve accuracy over time, correcting the limitations of simple maximum-value thresholding while maintaining computational efficiency.
2Quantity of substance
If conventional methods use only power usage magnitude to predict presence, then the data requirement is minimal, but the prediction fails in cases where power usage patterns do not correlate with actual presence
Solution Approach 1:
The patent applies preliminary action by pre-collecting and storing learning data that documents the correspondence relationships between power usage patterns and actual presence states before making predictions. This preparatory data collection enables the system to handle complex scenarios where simple power usage magnitude is insufficient, improving reliability without significantly increasing the data needed for each individual prediction.
Data Source
AI summary
A presence and absence prediction method includes: acquiring electric power data of a predetermined building at a first point in time; acquiring learning data obtained by learning, for each predetermined time period, a correspondence relationship between electric power data of the building at a point in time preceding the first point in time and information indicating whether the person was actually present in the building; predicting, on the basis of the electric power data of the building at the first point in time and the learning data, whether the person is present in the building; and generating presence and absence information that indicates a result of the prediction and outputting the presence and absence information to a predetermined terminal.