LLM Prompt Template Selection for Energy System Data Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Obtaining accurate data on the operation and efficiency of complex energy systems within premises is challenging due to their complexity and lack of access to relevant data, even for users and energy suppliers.
Innovation Solution
A method using a large language model (LLM) to determine operation and efficiency data by receiving user prompts, selecting appropriate templates with data fields, completing these templates with sensor data, and providing them to the LLM to generate accurate responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If users directly query complex energy system data without structured templates, then the interaction flexibility is improved, but the data accuracy and reliability deteriorate due to incorrect formatting and incomplete information
Solution Approach 1:
The system performs preliminary actions by pre-defining multiple structured prompt templates that contain all necessary data fields and formatting requirements before user interaction. When a user submits a query, the system automatically selects and populates the appropriate template with relevant sensor data, ensuring data accuracy is maintained while allowing users to interact in natural language without worrying about complex data structures.
2Measurement precision
If multiple data fields are collected from sensors to improve response accuracy, then the measurement precision is improved, but the device complexity increases due to additional sensors and data processing requirements
Solution Approach 1:
The system implements a universal template structure that can accommodate multiple types of sensor data (energy consumption, generation, contextual information) through a standardized format. This multi-functional template design allows the same data collection framework to handle various query types and sensor inputs without requiring separate complex processing paths for each data type, thereby maintaining measurement precision while managing system complexity.
3Manufacturing precision
If structured templates with specific data fields are used to improve LLM response accuracy, then the manufacturing precision of data processing is improved, but the ease of operation deteriorates due to rigid formatting requirements
Solution Approach 1:
The system introduces an intermediary component that automatically translates user-friendly natural language queries into the required structured template format. This intermediary layer handles the complexity of template selection, data field mapping, and sensor data integration, allowing users to interact simply with their questions while the system manages the rigid formatting requirements behind the scenes, thus maintaining both ease of operation and data processing accuracy.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Methods and apparatus for determining operation and/or efficiency data of an energy system of a premises, the energy system comprising one or more devices for consuming and/or generating electrical energy, and at least one sensor for generating sensor data indicating electrical energy consumed by and/or generated by the one or more devices and/or context data associated with the premises. The methods and apparatus receive a natural language query from a user, the query corresponding to operation and/or efficiency of the energy system; determine, by a template selection unit and based on the received query, a template of a prompt for submission to a large language model, wherein determining the template comprises selecting, by a trained classification model, at least one template from a plurality of templates, and wherein the determined template comprises a language structure, a data structure and one or more data types, at least one of the one or more data types being derivable from the sensor data and/or the context data; obtain, by a template completion unit, data corresponding to the one or more data types of the determined template; input, by the template completion unit, the obtained data to the determined template based on the data structure to generate a completed prompt for submission to the large language model; submit the completed prompt to the large language model; and generate, by the large language model and based on the completed prompt, data defining the operation and/or efficiency of the energy system.