Inferring PV Fleet Specs via Digital Twin Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Accurate forecasting of photovoltaic power generation in fleets is hindered by inaccuracies in system configuration data and the practicality of collecting high-speed, real-time data from dispersed systems, leading to unpredictable performance and high costs.
Innovation Solution
A method to infer operational specifications of photovoltaic systems by simulating power output based on historical data and irradiance values, using a normalized solar power simulation model to identify optimal configurations, which can be applied to any fleet configuration and time resolution, even beyond the input data rate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-speed real-time data collection is implemented from dispersed photovoltaic systems, then forecasting accuracy is improved, but data collection costs and system complexity increase significantly
Solution Approach 1:
The patent creates a virtual copy of the photovoltaic system through a digital twin that replicates system behavior, configuration, and performance characteristics. This digital model allows forecasting and analysis without requiring complex real-time data collection infrastructure, as the digital twin can be updated with periodic or historical data while maintaining high forecasting accuracy through its detailed virtual representation of the physical system.
Solution Approach 2:
The patent performs preliminary configuration analysis and system characterization during the digital twin creation phase, before real-time operations begin. System specifications, component details, and operational parameters are pre-loaded into the digital model, enabling accurate forecasting from the start without requiring continuous complex measurements during operation.
2Measurement precision
If complete system configuration data is collected from all photovoltaic plants, then power output forecast accuracy is improved, but data collection burden and costs increase
Solution Approach 1:
Instead of collecting extensive configuration data from each physical system, the patent creates a digital copy that can be populated with configuration information from available sources. The digital twin model allows for virtual testing and validation of configuration data, reducing the need for exhaustive field measurements while maintaining forecast accuracy.
Solution Approach 2:
The patent introduces a normalization model as an intermediary that standardizes configuration data from various sources and fills in missing information through modeling. This intermediary layer translates incomplete or varied configuration data into a standardized format that the digital twin can use for accurate forecasting, reducing the data collection burden while maintaining accuracy.
3Power
If photovoltaic systems are dispersed over large geographic areas, then fleet capacity and energy distribution are improved, but performance predictability decreases due to varying conditions
Solution Approach 1:
The patent segments the dispersed photovoltaic fleet into individual digital twin models for each plant or system. Each digital twin captures the specific configuration, location, and operational characteristics of its corresponding physical system. This segmentation allows for accurate individual forecasting that accounts for local conditions while enabling aggregate fleet-level analysis and prediction.
Solution Approach 2:
The patent applies local quality by customizing each digital twin model with location-specific parameters, environmental conditions, and system configurations. Each digital twin reflects the unique characteristics of its geographic location, including solar resource availability, temperature conditions, and local system design, thereby maintaining high predictability despite geographic dispersion.
4Measurement precision
If detailed operational specifications are inferred for each plant, then forecasting accuracy is improved, but computational processing time increases
Solution Approach 1:
The patent performs detailed configuration analysis and parameter optimization during the digital twin creation phase, before forecasting operations begin. System specifications are inferred and validated in advance, and the model is pre-configured with optimal parameters. This preliminary processing reduces computational requirements during actual forecasting operations, maintaining high accuracy while reducing real-time processing time.
Solution Approach 2:
The patent employs parameter changes by using a normalization model that transforms detailed configuration parameters into standardized, dimensionless parameters that capture essential system behavior. This parameter transformation reduces the complexity of computational models while preserving the relationships between configuration details and performance outcomes, thereby maintaining forecasting accuracy with reduced computational burden.
Data Source
AI summary
Operational specifications of a photovoltaic plant configuration can be inferred through evaluation of historical measured system production data and measured solar resource data. Based upon the location of the photovoltaic plant, a time-series power generation data set is simulated based on a normalized and preferably substantially linearly-scalable solar power simulation model. The simulation is run for a range of hypothetical photovoltaic system configurations. The simulation can be done probabilistically. A power rating is derived for each system configuration by comparison of the measured versus simulated production data, which is applied to scale up the simulated time-series data. The simulated energy production is statistically compared to actual historical data, and the system configuration reflecting the lowest overall error is identified as the inferred (and optimal) system configuration. Inferred configurations of photovoltaic plants in a photovoltaic fleet can be aggregated into a configuration of the fleet.