Electric Loadshape Forecasting via Smart Meter Inferential Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current electricity demand forecasting techniques rely on aggregate supply-side information, which lacks fine-grained data on individual customer usage, leading to inaccuracies due to varying weather conditions across regions, resulting in significant cost differences between base load generation and spot-market purchases.
Innovation Solution
A system utilizing smart meter data to train an inferential model with a Fourier-based decomposition-and-reconstruction technique, generating synthesized signals that account for ambient weather variations, enabling accurate electricity demand forecasting by projecting individual customer usage patterns into the future.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If aggregate supply-side information is used for forecasting, then the forecasting system is simple to operate, but the forecasting accuracy deteriorates due to lack of fine-grained weather data
Solution Approach 1:
The patent segments the aggregate supply-side data into individual customer-level demand data by using smart meter measurements from multiple customers. Each customer's usage pattern is analyzed separately, allowing the system to capture fine-grained weather-dependent variations while maintaining operational simplicity through automated processing.
Solution Approach 2:
The patent introduces an inferential model as an intermediary that bridges aggregate supply data and individual customer behavior. The model uses weather data and smart meter measurements to infer individual customer usage patterns, thereby improving forecasting accuracy without requiring direct measurement of every customer's weather exposure.
2Measurement precision
If fine-grained smart meter data from individual customers is collected, then the forecasting accuracy improves by capturing weather variations, but the device complexity increases
Solution Approach 1:
The patent makes the smart meters multi-functional by using them not only for billing purposes but also for forecasting. The same smart meter infrastructure that measures individual customer usage for billing is leveraged to capture weather-dependent demand patterns, eliminating the need for separate measurement devices and reducing overall system complexity.
Solution Approach 2:
The patent merges the billing data system with the forecasting system by combining smart meter measurements, weather data, and inferential modeling into a unified framework. This integration allows the system to simultaneously perform billing and forecasting functions, reducing device complexity while maintaining high forecasting accuracy.
3Device complexity
If historical aggregate supply curves are used for forecasting, then the system requires minimal data infrastructure, but the ability to predict short-term demand fluctuations deteriorates
Solution Approach 1:
The patent transitions from static historical aggregate supply curves to dynamic customer-level smart meter data that captures real-time usage patterns. By analyzing individual customer measurements and their response to weather changes, the system can dynamically predict short-term demand fluctuations while maintaining manageable data infrastructure through automated processing.
Data Source
AI summary
During operation, the system receives a set of input signals containing electrical usage data from a set of smart meters, wherein each smart meter gathers electrical usage data from a customer of the utility system. Next, the system uses the set of input signals to train an inferential model, which learns correlations among the set of input signals, and uses the inferential model to produce a set of inferential signals, wherein an inferential signal is produced for each input signal in the set of input signals. The system then uses a Fourier-based technique to decompose each inferential signal into deterministic and stochastic components, and uses the deterministic and stochastic components to generate a set of synthesized signals, which are statistically indistinguishable from the inferential signals. Finally, the system projects the set of synthesized signals into the future to produce a forecast for the electricity demand.