ML Predictive Model for Energy Load Shape Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Utility companies lack a practical method to inform energy customers about available energy efficiency programs and distinguish between likely and unlikely participants, leading to resource wastage and customer irritation due to indiscriminate communication.
Innovation Solution
A machine learning predictive model is trained using data from energy-consuming locations to predict the propensity of dwellings to participate in energy reduction programs based on load shapes and dwelling characteristics, allowing targeted communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If information about all energy efficiency programs is transmitted to all energy customers, then all customers receive comprehensive program information, but computational and network resources are wasted and customers are irritated by irrelevant information
Solution Approach 1:
The patent segments the customer base into distinct groups based on their energy consumption patterns, dwelling characteristics, and participation likelihood. By dividing customers into segments such as high-propensity participants and low-propensity participants, the system delivers targeted program information only to relevant segments, eliminating the waste of resources on blanket communications to all customers.
Solution Approach 2:
The system performs preliminary analysis of customer data including load shapes, dwelling characteristics, and historical participation patterns before sending program information. This preliminary action enables the system to pre-identify which customers are likely to participate in energy efficiency programs, allowing targeted communication strategies to be implemented before resource-intensive information transmission occurs.
2Loss of information
If information about all energy efficiency programs is transmitted to all energy customers via physical channels, then all customers receive comprehensive program information, but natural resources are wasted
Solution Approach 1:
The patent segments the customer base into distinct groups based on their energy consumption patterns, dwelling characteristics, and participation likelihood. By dividing customers into segments such as high-propensity participants and low-propensity participants, the system delivers targeted program information only to relevant segments, eliminating the waste of natural resources on blanket physical communications to all customers.
Solution Approach 2:
The system changes the parameter of communication target selection from universal (all customers) to selective (specific customer segments based on participation propensity). This parameter change in the communication strategy enables the system to reduce natural resource consumption by sending physical materials only to customers who are likely to engage with energy efficiency programs.
3Loss of information
If conventional computer systems send program information to all energy customers, then no customer is missed, but the systems cannot distinguish between likely and unlikely participants
Solution Approach 1:
The patent replaces conventional mechanical sorting methods based on generalized data with a machine learning-based predictive model. This substitution enables the system to analyze complex patterns in load shapes, dwelling characteristics, and historical participation data to accurately predict customer participation propensity, achieving precise segmentation that conventional systems cannot accomplish.
Solution Approach 2:
The system introduces an intermediary predictive model that processes customer data and generates participation propensity scores. This intermediary layer between raw data and communication decisions enables the system to distinguish between likely and unlikely participants by translating complex customer characteristics into actionable segmentation criteria.
4Ease of operation
If energy customers are grouped by generalized data such as dwelling type, then data sorting is simple, but the propensity to participate in energy efficiency programs varies widely within each subgroup
Solution Approach 1:
The patent applies multi-level segmentation, starting with broad dwelling type categories and then further dividing each category into sub-segments based on participation propensity. This hierarchical segmentation maintains the simplicity of initial grouping by dwelling type while adding precision through subsequent segmentation based on load shapes and predictive modeling.
Solution Approach 2:
The system applies different levels of segmentation granularity to different customer groups. For dwelling types with high internal variability in participation propensity, the system applies more detailed segmentation criteria. This local quality approach optimizes the balance between sorting simplicity and prediction precision for each specific customer segment.
Data Source
AI summary
Systems, methods, and other embodiments associated with a machine learning predictive model for predicting a propensity to implement energy reduction settings are described. Data records including load data for a target group of dwellings is obtained. An empirical load shape is generated for each given target dwelling based on the load data. A target feature vector is generated for each given target dwelling based on at least the empirical load shape corresponding to the given target dwelling. A trained machine learning predictive model is executed on the target feature vectors of the target group of dwellings to identify a set of target dwellings that are likely to reduce electricity consumed in accordance with electricity settings based on at least a generated predicted propensity for a target dwelling to implement the electricity settings.