Self-Powered Network Energy Management via Neural Network Parameters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In mobile edge computing networks, the uncertainty in energy demand and production from renewable sources leads to inefficiencies in energy management, resulting in high costs and potential energy failures, as conventional models fail to account for dynamic resource demands and variable energy generation patterns.
Innovation Solution
An energy management system utilizing a first and second energy management unit that communicate to acquire and process observation data, learn neural network parameters, and determine an optimal energy dispatch policy, incorporating reinforcement learning to adapt to changing conditions and reduce energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If self-power generation sources (renewable energy) are added to the wireless network, then energy cost is reduced and energy sustainability is improved, but uncertainty in energy production and demand increases
Solution Approach 1:
The system performs preliminary actions by collecting data on energy production and demand patterns, training machine learning models in advance to predict future energy conditions, and preparing energy dispatch strategies before uncertainty manifests. This allows the network to proactively adjust to renewable energy variability rather than reactively responding to it.
Solution Approach 2:
The energy management system implements continuous feedback loops where data from energy generators and consumers is collected, analyzed by machine learning models, and used to adjust energy dispatch decisions in real-time. This feedback mechanism enables the system to adapt to changing energy production and demand patterns, reducing the impact of uncertainty.
2Device complexity
If conventional energy management models are used, then system complexity is kept low, but the ability to handle dynamic resource demands and variable energy generation is insufficient
Solution Approach 1:
The energy management system performs self-service by using machine learning models to automatically analyze energy data, predict future conditions, and determine optimal dispatch strategies without requiring complex manual intervention or centralized control. The system learns from historical data and autonomously adapts to new patterns, reducing the need for human expertise while handling dynamic conditions effectively.
Solution Approach 2:
The system changes parameters by transitioning from static, rule-based energy management to dynamic, data-driven decision-making. Machine learning models continuously learn from incoming data, adjusting their predictions and recommendations based on changing energy production and consumption patterns. This allows the system to adapt its behavior without increasing structural complexity.
3Productivity
If more comprehensive data collection and machine learning are implemented, then energy dispatch optimization is improved, but computational requirements and system complexity increase
Solution Approach 1:
The energy management system is segmented into distinct functional components: data collection modules at energy generators and consumers, machine learning models for prediction, and decision-making modules for energy dispatch. This segmentation allows each component to perform its specific function efficiently, reducing overall system complexity while maintaining high productivity through coordinated operation of specialized subsystems.
Data Source
AI summary
Provided are an energy management method in a self-powered network, and an energy management apparatus and an energy management system to perform the method. The energy management system includes a first energy management unit and at least one second energy management unit configured to be communicable with the first energy management unit. Here, the at least one second energy management unit may acquire observation data using collected data and may transmit the acquired observation data to the first energy management unit, the first energy management unit may acquire a global neural network parameter by performing learning based on the observation data and may transfer the neural network parameter to the second energy management unit, and the at least one second energy management unit may determine an energy dispatch by performing learning using the neural network parameter and the collected data.