Machine Learning Energy Storage Management for Peak Demand
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Commercial and industrial sites face challenges in managing peak electricity demand due to complex and variable energy usage patterns, leading to high demand charges, and existing forecasting methods for photovoltaic cell production are inaccurate, especially under cloudy conditions, resulting in insufficient capacity to handle demand spikes.
Innovation Solution
A method that uses machine learning models, such as support vector machines and long short-term memory models, to forecast future electricity demand and photovoltaic cell production, adjusting the state of charge of energy storage devices based on forecasting errors and historical data to optimize demand management and reserve capacity, and transitions between different forecasting models to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional forecasting methods are used for photovoltaic cell production, then the system is simple to operate, but the forecasting accuracy deteriorates under cloudy conditions
Solution Approach 1:
The patent applies parameter changes by transitioning from traditional physical parameter-based forecasting to machine learning models that learn patterns from historical data. The system dynamically adjusts forecasting parameters based on weather conditions, using support vector machines and long short-term memory networks to capture non-linear relationships in photovoltaic production data, thereby improving accuracy under varying cloud cover conditions
Solution Approach 2:
The patent replaces traditional mechanical forecasting methods with intelligent machine learning systems. Instead of relying on simple physical models, the system uses support vector machines and neural networks to process complex weather and production data, substituting conventional forecasting mechanics with data-driven intelligent algorithms that adapt to changing conditions
2Quantity of substance
If energy storage capacity is reduced to lower costs, then the system becomes more economical, but the ability to handle demand spikes deteriorates
Solution Approach 1:
The patent applies preliminary action by using machine learning models to forecast future electricity demand and photovoltaic production in advance. The system predicts demand spikes before they occur and pre-charges energy storage devices accordingly, ensuring sufficient capacity is available when needed. This proactive approach allows the system to handle demand spikes reliably without requiring excessive energy storage capacity
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously monitors actual demand and production against forecasts, then adjusts future forecasting and energy storage operations accordingly. The machine learning models learn from historical errors and patterns, dynamically optimizing when to charge or discharge energy storage to maintain reliability while minimizing required capacity
3Loss of energy
If demand charge management is implemented, then electricity costs are reduced, but the complexity of energy management increases
Solution Approach 1:
The patent applies self-service by implementing an automated energy management system that uses machine learning to make autonomous decisions about when to charge or discharge energy storage devices. The system automatically forecasts demand, monitors energy storage state of charge, and executes peak shaving operations without requiring complex manual intervention or sophisticated external control systems, thereby reducing costs while keeping management complexity manageable
Data Source
AI summary
There is described a method of reserving a capacity of one or more energy storage devices. The method includes forecasting, based on past electricity demand of a site, future electricity demand of the site over a future time period. The method further includes determining a forecasting error between the forecasted future electricity demand and an actual electricity demand of the site over the future time period. The method further includes adjusting, based on the forecasting error, a target state of charge (SOC) of one or more energy storage devices. The method further includes reserving, based on the adjusted target SOC, a capacity of the one or more energy storage devices.