Multi-agent cooperation micro-grid energy management platform integrated with electricity market mechanism
By introducing a multi-agent collaborative architecture consisting of a market interaction layer, a resource scheduling layer, and a physical execution layer, the problem of microgrid energy management platforms being compatible with complex electricity market rules and resource collaborative scheduling has been solved, achieving efficient and economical energy management and market response capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UNIV OF SHANGHAI FOR SCI & TECH
- Filing Date
- 2025-12-11
- Publication Date
- 2026-05-12
AI Technical Summary
Existing microgrid energy management platforms are difficult to be compatible with complex and diverse overseas electricity market rules, lack responsiveness and bidding strategies, and in distributed environments, various energy resource entities lack autonomous decision-making capabilities and collaborative optimization, resulting in uncoordinated dispatch and unreasonable revenue distribution.
A multi-agent collaborative architecture consisting of a market interaction layer, a resource scheduling layer, and a physical execution layer is adopted. By combining electricity price prediction and carbon emission cost accounting in the market interaction layer, multi-agent reinforcement learning in the resource scheduling layer, and high-frequency data communication in the physical execution layer, collaborative control of photovoltaic, energy storage, and load equipment is achieved.
It improves the system's economy, security, and market responsiveness, enables efficient scheduling and revenue optimization in multi-market environments, reduces response latency, and enhances the system's robustness and adaptability.
Smart Images
Figure CN122022115A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent energy management technology, and in particular to a multi-agent collaborative microgrid energy management platform that integrates electricity market mechanisms. Background Technology
[0002] Currently, microgrid systems have become an important technological form for promoting energy structure transformation and regional energy independence, and are widely deployed in industrial and commercial parks, transportation hubs, and remote islands. As overseas markets increasingly integrate clean energy, microgrids are gradually taking on more market participation functions, such as load regulation, price response, and frequency support. However, overseas electricity market mechanisms generally exhibit complex and diverse characteristics, including multi-level and wide-timescale operational logics such as real-time pricing, day-ahead trading, capacity markets, and ancillary service markets.
[0003] Existing microgrid energy management platforms are mostly designed for fixed-price mechanisms, making them incompatible with frequently changing market rules. In particular, they generally lack responsiveness and bidding strategy support when facing dynamic bidding mechanisms in overseas markets. Furthermore, the widespread integration of distributed renewable energy and electric vehicle loads has significantly increased source-load volatility. Traditional centralized dispatch strategies are constrained by response latency and computing power bottlenecks, making them unsuitable for the practical needs of minute-level dynamic decision-making.
[0004] Existing technologies primarily focus on local energy balance and cost minimization, with limited consideration for electricity market rule modeling and bidding mechanism integration. Furthermore, their ability to model factors such as carbon trading costs and capacity bidding constraints is limited. In distributed environments, various energy resource entities (such as photovoltaic inverters, energy storage devices, and charging piles) generally lack autonomous decision-making capabilities and collaborative optimization mechanisms, leading to persistent problems such as uncoordinated dispatching and unreasonable distribution of system benefits. Summary of the Invention
[0005] The purpose of this invention is to propose a multi-agent collaborative microgrid energy management platform that improves the overall economic efficiency, safety, and market responsiveness of the system by introducing market mechanism modeling, reinforcement learning optimization strategies, and carbon emission cost accounting logic to achieve coordinated control of photovoltaic, energy storage, and load equipment.
[0006] To achieve the above objectives, this invention proposes a multi-agent collaborative microgrid energy management platform that integrates electricity market mechanisms. The energy management platform includes a market interaction layer, a resource scheduling layer, and a physical execution layer. The market interaction layer enables dynamic electricity price forecasting, multi-market bidding optimization, and carbon emission cost accounting; the resource scheduling layer adopts a multi-agent reinforcement learning framework to achieve resource-level optimization control; and the physical execution layer uses the IEC 61850 protocol to achieve high-frequency data communication and control response between devices.
[0007] Furthermore, the market interaction layer uses an LSTM-CNN hybrid neural network to make short-term predictions of electricity prices, integrating calendar information and meteorological factors.
[0008] Furthermore, the market interaction layer, based on stochastic programming theory, constructs an electricity price probability distribution model and a multi-stage decision tree structure, comprehensively considering the multi-timescale constraints and electricity price uncertainty of the day-ahead market and capacity market, to realize a joint bidding optimization strategy for energy storage equipment; The formula for calculating the connection pricing optimization strategy is as follows: ; In the formula, The current day electricity price; For the bid power; Conditional risk value; Furthermore, the market interaction layer dynamically adjusts operating costs based on real-time carbon trading market prices and outputs carbon constraint scheduling factors.
[0009] Furthermore, the resource scheduling layer includes a photovoltaic agent module, which adjusts the system's response to photovoltaic fluctuations by integrating irradiance prediction with power confidence intervals.
[0010] Furthermore, the resource scheduling layer includes an energy storage agent module. This module constructs a dual-objective optimization strategy to balance economic arbitrage and cycle life, and supports automatic adaptation to peak-valley electricity price differences. Objective function of energy storage agent: ; In the formula: , where is the return weighting coefficient, i.e., economic priority; Revenue is the arbitrage profit; This is the lifespan weighting coefficient, i.e., durability priority; This refers to the capacity decay. This refers to the rated capacity of the energy storage. This refers to capacity retention rate, or lifespan health.
[0011] Furthermore, the resource scheduling layer includes a charging pile agent module, which adjusts the charging power based on peak and off-peak periods and market signals to guide end-user behavior and improve load flexibility.
[0012] Furthermore, the resource scheduling layer includes a virtual power plant coordinator: enabling revenue coordination and response scheduling across devices and markets.
[0013] Furthermore, the physical execution layer enables the energy storage system's SOC (State of Charge) to be dynamically adjusted based on a second-level cycle, and the charging pile power setting to support millisecond-level response, adapting to rapid load regulation scenarios, as well as supporting the access of heterogeneous devices and plug-and-play configuration.
[0014] Compared with the prior art, the advantages of the present invention are: 1. Multi-market mechanism integration: Transform mechanisms such as capacity bidding, real-time price response, and imbalance penalty into programmable linear constraints to improve compatibility with market mechanisms.
[0015] 2. Risk perception and optimization capabilities: Introduce the CVaR (Conditional Value at Risk) model to quantify the risk of electricity price fluctuations and optimize energy storage and load-side behavior to avoid revenue fluctuations.
[0016] 3. Multi-agent distributed response: Based on the multi-agent system (MAS) design framework, the response latency is controlled within 500ms, ensuring scheduling efficiency in high-frequency trading environments.
[0017] 4. The platform has strong adaptability to deployment: The system can complete the modeling of typical market scenarios and verify the robustness and economic improvement under conditions such as intraday load fluctuations, photovoltaic forecast deviations, and abnormal electricity prices. Attached Figure Description
[0018] Figure 1 This is a schematic diagram of the three-layer architecture of the microgrid energy management platform in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the engineering implementation in an embodiment of the present invention. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be further described below.
[0020] like Figure 1 As shown, this invention proposes a multi-agent collaborative microgrid energy management platform that integrates electricity market mechanisms. The energy management platform includes a market interaction layer, a resource scheduling layer, and a physical execution layer. I. Market Interaction Layer Deploy the Market Agent component to handle data interface interaction with overseas electricity markets and generate market strategies. This includes the following modules: Dynamic electricity price forecasting module: Employs a hybrid LSTM-CNN neural network for short-term electricity price forecasting, integrating calendar information and meteorological factors. The main purpose of dynamic electricity price forecasting is to calculate revenue, and it also includes the following key functions: 1. Providing a price benchmark for multi-market bidding optimization, supporting arbitrage decisions between day-ahead and real-time markets; 2. Helping energy storage devices formulate charging and discharging strategies by predicting price fluctuation trends, maximizing arbitrage opportunities; 3. Providing electricity price input for carbon emission cost accounting, influencing the calculation of carbon constraint scheduling factors; 4. Guiding charging pile agents to adjust power, achieving flexible guidance of user charging behavior.
[0021] Multi-market bidding optimizer: Based on stochastic programming theory, it constructs an electricity price probability distribution model and a multi-stage decision tree structure. The specific construction process is as follows: 1. Establish a time series model of electricity prices using historical electricity price data, meteorological factors, and calendar information; 2. Generate the joint probability distribution of day-ahead electricity price and real-time electricity price through Monte Carlo simulation; 3. Construct a multi-stage decision tree to simulate decision paths in different market stages (day-ahead market, real-time market, capacity market); 4. Introduce Conditional Value at Risk (CVaR) to quantify the profit risk brought about by electricity price fluctuations and optimize bidding strategies.
[0022] Considering the multi-timescale constraints of the day-ahead and capacity markets, as well as the uncertainty of electricity prices, it is necessary to maximize the overall returns of energy storage devices across multiple markets while controlling risks. Therefore, a joint pricing optimization strategy for energy storage devices is constructed: ; The current day electricity price; For the bid power; This is the conditional risk value.
[0023] The output of this strategy is sent to the physical execution layer as a scheduling instruction to realize the actual control of the energy storage device. At the same time, in the resource scheduling layer, the energy storage agent formulates the bidding power and charging and discharging plan according to the strategy. Carbon emission cost accounting module: Dynamically adjusts operating costs based on real-time carbon trading market prices (such as the EUA index) and outputs carbon constraint scheduling factors.
[0024] The dynamic correction logic of the carbon emission cost accounting module is as follows: 1. Obtain real-time carbon trading market prices (such as the EUA index) and the carbon emission intensity of each device within the microgrid; 2. Calculate the carbon emission cost dynamically using the following formula: ; in, For real-time carbon prices, For the energy consumption of device i, Its carbon emission factor.
[0025] 3. Output carbon constraint scheduling factor Used to adjust the scheduling priority of each device: ; in, The baseline carbon cost is used for normalization.
[0026] II. Resource Scheduling Layer Resource-level optimization control is achieved using the MARL (Multi-Agent Reinforcement Learning) framework. This includes: Photovoltaic Agent: By integrating irradiance prediction and power confidence intervals, it adjusts the system's response to photovoltaic fluctuations; the detailed adjustment mechanism is as follows: 1. Irradiance Prediction Fusion Module • Use NWP (Numerical Weather Prediction) data as input source • Integrating satellite cloud image correction algorithms improves prediction accuracy by 30%. • Time resolution reaches 15-minute update level 2. Calculation of power confidence interval • Construct a probability density function based on historical error distribution: ; • Dynamically adjust confidence bandwidth 3. Response strategy execution When the prediction deviation exceeds the threshold: • Trigger the power compensation command of the energy storage agent • Send a power correction request to the VPP coordinator • Automatically adjust inverter limit value Energy Storage Agent: Constructing a dual-objective optimization strategy, such as Figure 2 As shown, it balances economic arbitrage and cycle life, and supports automatic adaptation of peak-valley electricity price difference strategies; Objective function of energy storage agent: ; , where is the return weighting coefficient, i.e., economic priority; Revenue is the arbitrage profit; This is the lifespan weighting coefficient, i.e., durability priority; This refers to the capacity decay. This refers to the rated capacity of the energy storage. Capacity retention rate, i.e., lifespan health. Charging station agent: Adjusts charging power based on peak and off-peak load periods and market signals to guide end-user behavior and improve load flexibility. The specific implementation method is as follows: Receive electricity price signals and carbon dispatch factors released by the market interaction layer; The charging power is dynamically adjusted according to the peak and off-peak periods of the load, for example, increasing the power during periods of low electricity prices and decreasing the power during peak periods; Utilize fuzzy logic or reinforcement learning algorithms to optimize charging plans based on historical user behavior data; By interacting with the VPP coordinator, it participates in system-level demand response and improves overall load flexibility.
[0027] Virtual Power Plant Coordinator (VPP-Coordinator): Enables revenue coordination and response scheduling across devices and markets. The core function of the VPP-Coordinator is to achieve cross-device revenue coordination through a distributed optimization algorithm. Its model can be described as follows: ; in, The objective function is for local devices (such as energy storage, photovoltaics, and charging piles).
[0028] In this embodiment, the photovoltaic agent, energy storage agent, and charging pile agent collaborate through a virtual power plant coordinator, achieving the following effects: Economic efficiency improvement: Photovoltaic agents provide renewable energy forecasts, energy storage agents realize arbitrage, and charging pile agents smooth the load curve, all of which reduce the cost of electricity purchase and carbon costs. Security assurance: Energy storage agents provide frequency support through second-level SOC adjustment, charging pile agents mitigate power surges through millisecond-level response, and photovoltaic agents reduce the impact of prediction deviations through confidence interval management; Market responsiveness: The three parties work together to participate in bidding in multiple markets, thereby enhancing the overall profitability and risk resistance of the VPP.
[0029] III. Physical Execution Layer Based on the IEC 61850 protocol, high-frequency data communication and control response between devices are achieved. Key features include: The SOC (State of Charge) of the energy storage system is dynamically adjusted based on a second-level cycle; The charging pile power setting supports millisecond-level response, adapting to rapid load adjustment scenarios; Supports heterogeneous device access and plug-and-play configuration.
[0030] To further illustrate the technical effects of this invention, taking the German microgrid as an example, the technical effects of this application are discussed by comparing its management platform with traditional technical solutions, as follows:
[0031] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.
Claims
1. A multi-agent collaborative microgrid energy management platform integrating electricity market mechanisms, characterized in that, The energy management platform includes a market interaction layer, a resource scheduling layer, and a physical execution layer; The market interaction layer enables dynamic electricity price forecasting, multi-market bidding optimization, and carbon emission cost accounting; the resource scheduling layer uses a multi-agent reinforcement learning framework to achieve hierarchical resource optimization control; and the physical execution layer uses the IEC 61850 protocol to achieve high-frequency data communication and control response between devices.
2. The multi-agent collaborative microgrid energy management platform integrating electricity market mechanisms according to claim 1, characterized in that, The market interaction layer uses an LSTM-CNN hybrid neural network to make short-term predictions of electricity prices, integrating calendar information and meteorological factors.
3. The multi-agent collaborative microgrid energy management platform integrating electricity market mechanisms according to claim 1, characterized in that, The market interaction layer is based on stochastic programming theory, constructs an electricity price probability distribution model and a multi-stage decision tree structure, and comprehensively considers the multi-timescale constraints and electricity price uncertainty of the day-ahead market and capacity market to realize a joint bidding optimization strategy for energy storage equipment. The calculation formula for the connection pricing optimization strategy is as follows: ； In the formula, The current day electricity price; For the bid power; Conditional risk value; The multi-agent collaborative microgrid energy management platform integrating the power market mechanism according to claim 1 is characterized in that the market interaction layer dynamically adjusts the operating cost based on the real-time carbon trading market price and outputs a carbon-constrained scheduling factor.
4. The multi-agent collaborative microgrid energy management platform integrating electricity market mechanisms according to claim 1, characterized in that, The resource scheduling layer includes a photovoltaic agent module, which adjusts the system's response to photovoltaic fluctuations by integrating irradiance prediction and power confidence interval.
5. The multi-agent collaborative microgrid energy management platform integrating electricity market mechanisms according to claim 1, characterized in that, The resource scheduling layer includes an energy storage agent module, which constructs a dual-objective optimization strategy to balance economic arbitrage and cycle life, and supports automatic adaptation to peak-valley electricity price differences. The objective function of the energy storage agent is: ； In the formula: , where is the return weighting coefficient, i.e., economic priority; Revenue is the arbitrage profit; This is the lifespan weighting coefficient, i.e., durability priority; This refers to the capacity decay. This refers to the rated capacity of the energy storage. This refers to capacity retention rate, or lifespan health.
6. The multi-agent collaborative microgrid energy management platform integrating electricity market mechanisms according to claim 1, characterized in that, The resource scheduling layer includes a charging pile agent module, which adjusts the charging power based on peak and off-peak periods and market signals to guide end-user behavior and improve load flexibility.
7. The multi-agent collaborative microgrid energy management platform integrating electricity market mechanisms according to claim 1, characterized in that, The resource scheduling layer includes a virtual power plant coordinator, which enables revenue coordination and response scheduling across devices and markets.
8. The multi-agent collaborative microgrid energy management platform integrating electricity market mechanisms according to claim 1, characterized in that, The physical execution layer enables the energy storage system's SOC (State of Charge) to be dynamically adjusted based on a second-level cycle, and the charging pile power setting to support millisecond-level response, adapting to rapid load control scenarios, and supporting the access of heterogeneous devices and plug-and-play configuration.