Hierarchical Solar PV Inverter Forecasting via Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The widespread implementation of solar power systems is hindered by factors such as weather conditions, seasonal changes, and intra-hour variability, necessitating reliable mechanisms for predicting solar energy production to manage energy reserve requirements and grid operations effectively.
Innovation Solution
A hierarchical approach using machine learning algorithms, specifically Artificial Neural Networks (ANN) and Support Vector Regression (SVR), is employed to forecast solar power output, leveraging historical data and reduced weather inputs to predict energy production 15-minutes, 1-hour, and 24-hours ahead, with a focus on individual inverter-level forecasting to improve overall solar power generation predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If solar power systems are widely implemented, then renewable energy production increases, but prediction accuracy of solar energy production deteriorates due to weather conditions, seasonal changes, and intra-hour variability
Solution Approach 1:
The patent segments the solar power system into multiple inverters and applies hierarchical forecasting, where individual inverter-level predictions are made first, then aggregated to system-level predictions. This segmentation allows for more granular and accurate predictions by capturing local variations in weather conditions and performance characteristics of each inverter, thereby improving overall prediction accuracy while maintaining high renewable energy production.
Solution Approach 2:
The patent introduces a hierarchical dimension to the forecasting approach, moving from traditional system-level predictions to a multi-level structure that includes inverter-level, string-level, and system-level forecasts. This dimensional change enables the model to capture variations at different scales and improve prediction accuracy by accounting for intra-hour variability and local conditions that would be invisible in aggregate system-level models.
2Device complexity
If traditional forecasting methods are used, then system complexity is low, but forecast accuracy deteriorates due to inability to capture intra-hour variability and local conditions
Solution Approach 1:
The forecasting system is segmented into multiple hierarchical levels (inverter, string, system) with dedicated models at each level. This segmentation enables the system to capture local conditions and intra-hour variability that traditional single-level models miss, improving forecast accuracy while maintaining manageable complexity through modular model structures that can be trained and deployed independently at each level.
Solution Approach 2:
The patent changes the parameters being forecasted at different hierarchical levels, with individual inverters predicting their own output based on local conditions, then aggregating these predictions. This parameter change approach allows the system to adapt to local variations in weather, orientation, and performance characteristics, significantly improving forecast accuracy compared to traditional uniform modeling approaches.
3Measurement precision
If inverter-level forecasting is implemented, then prediction accuracy improves, but computational requirements and system complexity increase
Solution Approach 1:
The system is segmented into independent inverter-level forecasting units, each capable of making predictions based on local data. This segmentation improves accuracy by capturing inverter-specific characteristics and local conditions, while the modular structure allows for scalable implementation where complexity can be adjusted by adding or removing inverter-level models without affecting the entire system.
Solution Approach 2:
Individual inverter-level predictions are merged through aggregation to produce system-level forecasts. This merging process combines the benefits of granular inverter-level accuracy with the simplicity of system-level reporting, reducing the overall computational burden compared to running a single complex system-level model while maintaining high prediction accuracy through the aggregation of simpler inverter-level models.
Data Source
AI summary
A photovoltaic system can include multiple photovoltaic power inverters that convert sunlight to power. An amount of power for each of the inverters can be measured over a period of time. These measurements, along with other data, can be collected. The collected measurements can be used to generate artificial neural networks that predict the output of each inverter based on input parameters. Using these neural networks, the total solar power generation forecast for the photovoltaic system can be predicted.