Power generation prediction system and method thereof
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional power generation prediction systems using single neural networks are inefficient and time-consuming to retrain when pyranometers deteriorate or are reinstalled, leading to instability in solar and wind power generation predictions, which affects power deployment in grids.
Innovation Solution
A power generation prediction system utilizing multiple neural networks, where a first neural network processes input data to generate amount prediction data, and a second neural network calculates power generation prediction data, including maximum and minimum values, with a cost function optimizing hit probability and bound difference, allowing for fine-tuning and retraining of the second neural network when devices deteriorate or are reinstalled.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single neural network is used for power generation prediction, then the system structure is simple, but the time consumption for retraining when devices deteriorate or are reinstalled increases significantly
Solution Approach 1:
The patent divides the single neural network into multiple neural networks (first neural network for solar power generation prediction, second neural network for wind power generation prediction). This segmentation allows independent training and updating of each network, reducing the overall retraining time when device changes occur, as only the affected network needs to be retrained rather than a single comprehensive network.
2Ease of operation
If a single neural network is used for power generation prediction, then the system is easier to maintain, but the prediction precision for power deployment decreases
Solution Approach 1:
The patent segments the prediction system into specialized neural networks for different power generation types (solar and wind). Each network is optimized for its specific domain, improving prediction precision for power deployment. The segmentation maintains ease of operation by allowing independent maintenance and updating of each network without affecting the entire system.
Solution Approach 2:
The patent applies local quality by designing each neural network with specific architecture and parameters optimized for its particular power generation type. The first neural network is tailored for solar power characteristics while the second is optimized for wind power patterns, enabling each to achieve higher prediction precision for its designated function.
3Reliability
If device changes (deterioration or reinstallation) occur, then the system must be retrained, but the time consumption and inconvenience for programmers increase
Solution Approach 1:
The patent segments the training process into separate, independent training procedures for each neural network. When device changes occur, only the specific network affected by the change needs to be retrained, significantly reducing the time and effort required compared to retraining a single comprehensive network. This modular approach allows programmers to efficiently update only the necessary components.
4Device complexity
If only predicted future generation of power is obtained, then the prediction system is simpler, but effective power deployment cannot be achieved without predicted bounds
Solution Approach 1:
The patent segments the output of each neural network into multiple components: predicted future generation of power, maximum predicted power, and minimum predicted power. This segmentation provides the necessary prediction bounds for effective power deployment while maintaining relatively simple system architecture. Each neural network independently generates these segmented outputs, enabling comprehensive power deployment decisions.
Data Source
AI summary
A power generation prediction system using a first and second neural networks is provided, and the first neural network is connected to the second neural network. The first neural network receives first input data, and generates the amount prediction data according to the first input data. The first input data is used to determine amount prediction data, and the amount prediction data is used to determine power generation prediction data. The second neural network receives the amount prediction data, and calculates the power generation prediction data according to the amount prediction data. When a device in a selected area is deteriorated or reinstalled, the second neural network is fine-tuned and trained again. The power generation prediction data is a power generation prediction bound having a maximum and minimum power generation prediction values, and thus the power deployment terminal in a power grid can deploy power more precisely.