New energy consumption bottleneck dynamic dredging method based on energy storage value field topology reconstruction

Abstract

The invention discloses a new energy consumption bottleneck dynamic dredging method based on energy storage value field topology reconstruction. The method comprises the steps that an energy storage value field capable of being dynamically quantized is constructed based on the capacity state, the dynamic response characteristic and the geographic position of an energy storage unit, and initialization is completed through multi-dimensional weight evaluation space-time adjustment potential; meanwhile, a multi-agent system covering new energy, load, energy storage and power grid key elements is built, each agent is provided with a field large language model and a cognitive thinking chain, and strategy reasoning and collaborative decision in a complex environment are supported. According to the scheme, after multi-source operation data is collected in real time, relevant features are analyzed to recognize the absorption bottleneck and trigger value field reconstruction, an optimal strategy containing new energy output adjustment, energy storage scheduling and load response is generated through multi-agent multi-round negotiation iteration, and after execution, real-time monitoring is conducted, and the value field is finely adjusted based on feedback. And real-time identification, accurate positioning and efficient dredging of the new energy consumption bottleneck are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy power generation technology and data intelligent analysis, in particular to a new energy consumption bottleneck dynamic dredging method based on energy storage value field topology reconstruction. BACKGROUND

[0002] With the continuous rise of the penetration rate of intermittent new energy such as wind power and photovoltaic in the power system, the strong volatility and uncontrollability of its output pose a core challenge to real-time balance of the power grid. In new energy-rich areas, the mutual coupling of factors such as mismatch between source and load in time and space, limited transmission capacity of the power grid, and insufficient local reactive power support frequently causes consumption bottlenecks. The bottlenecks are specifically manifested in periodic power over-limit of transmission lines, voltage deviation exceeding the standard of key nodes, and forced abandonment of wind and light to maintain system stability, which not only causes serious waste of clean energy, but also restricts the healthy development of the new energy industry and poses a direct threat to the safe and stable operation of the power grid.

[0003] To address the above challenges, existing technologies mostly use energy storage scheduling strategies based on fixed rules or rely on optimization models based on day-ahead and intra-day plans. For example, some methods control the energy storage system according to preset charging and discharging thresholds or simple power fluctuation suppression targets, and their strategies are significantly rigid and difficult to adapt to real-time dramatic changes in new energy output and load demand. Some other methods attempt to build optimization models with the goal of minimizing total operating cost or minimizing power abandonment rate, but their models usually rely on prediction data and have complex solving processes and long calculation time, which cannot meet the real-time requirements of online decision-making. The fundamental limitation of these methods is that the energy storage system is regarded as an isolated and single-function regulation unit, and its multi-dimensional value in different time and space scenarios is not finely modeled and dynamically evaluated, resulting in that the comprehensive regulation potential of the energy storage is not fully tapped and the response capability of the system to sudden bottlenecks is insufficient.

[0004] Existing consumption bottleneck dredging schemes have obvious shortcomings in decision-making mechanism. Traditional centralized optimization methods often face dimension disaster problems and have low computational efficiency when dealing with high-dimensional and nonlinear power grid operation problems. Decentralized decision-making mode often lacks effective collaborative reasoning and information exchange between units, which easily forms information islands and leads to local optimization or even conflict of decisions, making it difficult to generate a globally near-optimal collaborative dredging strategy. SUMMARY

[0005] To overcome the above problems, the present application provides a new energy consumption bottleneck dynamic dredging method based on energy storage value field topology reconstruction, which realizes real-time identification, accurate positioning and efficient dredging of new energy consumption bottlenecks by constructing a dynamically quantifiable energy storage value field and combining a multi-agent collaborative decision-making mechanism.

[0006] The technical solution of the present application is as follows:

[0007] The method comprises the following steps:

[0008] Based on the capacity state, dynamic response characteristics and geographical location information of each energy storage unit in the region, a dynamically quantifiable energy storage value field is constructed. The adjustment potential of the energy storage value field in the time and space scales is evaluated through multi-dimensional weight to form an initialized energy storage value field.

[0009] A multi-agent system including new energy, load, energy storage and key elements of the power grid is constructed. Each agent is equipped with a large language model in the field and a dynamically generated cognitive thinking chain to support the agent in complex operating environment for strategy reasoning and collaborative decision-making.

[0010] Real-time acquisition of multi-source operation data of new energy station, power grid node and load center, analysis of output fluctuation characteristics, line overload characteristics and voltage out-of-limit characteristics, identification of whether there is a bottleneck for consumption; once the bottleneck for consumption is identified, the energy storage value field reconstruction mechanism is automatically triggered.

[0011] The reconstructed energy storage value field is used as the basis for strategy evaluation. Through multi-round critical negotiation and strategy iteration between each agent in the multi-agent system, the optimal unblocking strategy covering new energy output adjustment, energy storage dispatching and load response is formed.

[0012] The optimal unblocking strategy generated is executed, and the key indicators of the power grid and the state of each execution unit are monitored in real time. Based on the monitoring feedback data, the adaptability of the energy storage value field is evaluated. If it is found that the value field does not match the operation demand, the fine-tuning mechanism is started.

[0013] As a further choice of the present application, the space-time value function for constructing the energy storage value field is:

[0014] ;

[0015] Wherein: represents the intensity of the energy storage value field at the spatial position , time , is the total number of energy storage units in the region, is the global weight coefficient of the th energy storage unit; is the spatial influence function of the th energy storage unit, wherein is the geographical coordinates of the energy storage unit , and is the influence radius parameter; is the value core function of the th energy storage unit, and Energy storage unit capacity state , dynamic response characteristics , and geographical location attributes ; defined as: , is the capacity weight, is the response characteristic weight, is the spatial attenuation coefficient, is the location to the electrical distance.

[0016] As a further choice of the present application, the multi-agent system comprises a new energy station agent, a load aggregation agent, an energy storage unit agent, and a power grid node agent;

[0017] The state space of the new energy station agent includes predicted / actual output, prediction error, and grid-connected point voltage, and the action space is output adjustment suggestion or power limit value application.

[0018] The state space of the load aggregation agent includes predicted / actual load, interruptible load proportion, and price elasticity, and the action space is load reduction amount or transfer plan.

[0019] The state space of the energy storage unit agent includes state of charge and charge / discharge power limit, and the action space is charge / discharge power instruction.

[0020] The state space of the power grid node agent includes node voltage, phase angle, and connected line power flow, and the action space is reactive power compensation device switching or transformer tap adjustment.

[0021] As a further choice of the present application, the domain large language model equipped for each agent is fine-tuned in the power system domain knowledge, used for converting high-dimensional state data observed by the agent into natural language or structured text description, and generating a thinking chain of step-by-step reasoning, and outputting preliminary action suggestions or strategy options based on the thinking chain.

[0022] As a further choice of the present application, the identification of the accommodation bottleneck comprises:

[0023] The line power flow margin is calculated, and the calculation formula is as follows:

[0024] ;

[0025] wherein: is the power flow margin of the line at time , is the thermal stability limit or static safe transmission capacity of the line , is the line At time , the actual transmission power, taking the absolute value indicates considering bidirectional flow; when the flow margin is lower than the set threshold , mark the line as a potential overload risk;

[0026] The node voltage deviation calculation formula is as follows:

[0027] ;

[0028] Wherein: is the node Voltage deviation at time , is the actual voltage measurement value of node at time , is the rated voltage level of node ; When the deviation exceeds the threshold , mark the node as a voltage out-of-limit risk;

[0029] If the fluctuation leads to further deterioration of the risk line flow or the risk node voltage, that is: ; And the duration , it is determined that the system appears to be a consumption bottleneck.

[0030] As a further choice of the present application, the trigger energy storage value field reconstruction mechanism comprises:

[0031] Based on the reconstruction optimization model, dynamically adjust the global weight and the value core function parameters of each energy storage unit, the reconstruction optimization model is represented as:

[0032] ;

[0033] Wherein: is the weight set after reconstruction, is the adjustable parameter set of the value core function , is the boundary value of safe operation of the system, is the system operating state predicted based on the reconstructed value field and multi-agent strategy , , is the system state prediction model, is the regularization coefficient;

[0034] After solving the optimal and , calculate and publish the reconstructed energy storage value field .

[0035] As a further option in this application, the multiple rounds of critical negotiation and strategy iteration include:

[0036] In the first round, each agent proposes an initial action plan based on local information and the reconstructed value field.

[0037] In the second round, the agents evaluate the feasibility and global impact of each other's proposals, and the power grid node agents conduct a system security critique.

[0038] In the third round, the agents modify their strategies based on the feedback and resubmit them, iterating until all agents' strategies form a set that satisfies the system's security constraints and yields the expected value return. Near-optimal set of consensus strategies ;

[0039] Define intelligent agents In strategy The expected value return is :

[0040] ;

[0041] in, For system status, For intelligent agents The actions taken Represents the set of policies of all other agents. As a discount factor, It is an intelligent agent At any moment Local revenue, These are value field weighting coefficients, used to balance individual gains with the overall value absorbed by the system. Indicates the execution strategy Then, the system state evolved to The energy storage value field strength corresponding to the time;

[0042] Consensus Strategy Set satisfy: .

[0043] As a further option of this application, before executing the generated optimal unblocking strategy, the following steps are also included:

[0044] The optimal unblocking strategy is input into a high-precision digital power grid simulation model for safety verification. After the verification is passed, control commands are sent to each execution unit.

[0045] As a further option of this application, the initiation fine-tuning mechanism includes:

[0046] Performance metrics, including bottleneck elimination rate and strategy execution deviation, are calculated based on monitoring feedback data.

[0047] If performance metrics do not meet expectations, a lightweight algorithm will be used to weight the key energy storage units. Alternatively, the core value function parameters may be slightly adjusted to improve the match between the energy storage value field and operational needs.

[0048] As a further option of this application, the multi-agent system exchanges state summaries, thought chain reasoning processes, and policy suggestions through a communication protocol. The negotiation mechanism adopts a multi-round iterative mode of proposal-critique-improvement to achieve the generation of a globally near-optimal dredging strategy.

[0049] The beneficial effects of this application are as follows:

[0050] This method constructs a dynamically quantifiable energy storage value field, enabling precise evaluation and dynamic reconfiguration of the multidimensional adjustment potential of energy storage systems. It overcomes the limitations of existing technologies that treat energy storage as a single, isolated unit, and fully explores and quantifies the comprehensive value of energy storage in different spatiotemporal scenarios, thereby significantly improving the utilization efficiency of energy storage resources and the ability to respond to sudden bottlenecks.

[0051] This method introduces a multi-agent system equipped with a large language model and cognitive thought chain to construct an efficient distributed collaborative decision-making mechanism. Through multiple rounds of critical negotiation and policy iteration among agents, the multi-agent system effectively integrates dispersed local information with global objectives. This avoids the computational complexity and curse of dimensionality problems faced by centralized optimization, and solves the information silos and policy conflicts that may occur in decentralized decision-making, thereby generating a globally near-optimal collaborative coordination strategy.

[0052] This method forms a complete technology chain from real-time perception, bottleneck identification, dynamic reconfiguration to strategy generation and optimization. Driven by real-time operational data, it can quickly respond to drastic changes in renewable energy output and load, and continuously fine-tune the energy storage value field through execution feedback, ensuring the real-time performance and adaptability of the strategy. It fundamentally solves the core pain points of traditional methods, such as reliance on predictive data, rigid models, and computational time consumption that cannot meet the requirements of online decision-making, and achieves proactive, adaptive, and intelligent unblocking of energy consumption bottlenecks. Attached Figure Description

[0053] Figure 1 A schematic diagram of the overall process of the dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction;

[0054] Figure 2 A detailed flowchart of the S100 steps of the dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction.

[0055] Figure 3A detailed flowchart of the S200 steps of the dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction.

[0056] Figure 4 A detailed flowchart of the S300 steps of the dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction.

[0057] Figure 5 A detailed flowchart of the S400 steps of the dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction.

[0058] Figure 6 This is a detailed flowchart of the S500 method for dynamically unblocking bottlenecks in new energy consumption based on energy storage value field topology reconstruction. Detailed Implementation

[0059] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0060] In the context of new power systems, the integration of a high proportion of renewable energy sources has led to increasingly complex system operation, making the dynamic identification and resolution of bottlenecks a key challenge. This invention is based on three pillars: spatiotemporal value field theory, multi-agent reinforcement learning, and distributed optimization theory. By constructing a dynamically quantifiable energy storage value field, a unified evaluation benchmark is provided for the spatiotemporal regulation potential of energy storage resources. Distributed reasoning and collaborative decision-making are performed using a multi-agent system equipped with a large language model to generate globally near-optimal resolution strategies. Furthermore, a closed-loop feedback mechanism continuously optimizes the value field configuration, ensuring the method's adaptability and robustness.

[0061] The derivation of the core theoretical formula is as follows:

[0062] 1. Quantitative model of energy storage value field.

[0063] The energy storage value field is essentially a scalar field defined in the spatiotemporal dimension, used to characterize the marginal contribution value of energy storage units at different locations and times to the system's absorption capacity. Its mathematical definition is based on a spatiotemporal value function:

[0064] ;

[0065] in: Indicates spatial location ,time The energy storage value field intensity, This represents the total number of energy storage units in the region. For the first The global weight coefficient of each energy storage unit reflects its strategic importance in the entire system.

[0066] It is the first The spatial influence function of an energy storage unit is typically represented by a Gaussian kernel function centered on the geographical location of the energy storage unit. ,in For energy storage units Geographic coordinates To influence the radius parameter.

[0067] It is the first The core value function of an energy storage unit is its capacity state. Dynamic response characteristics and geographical location attributes A comprehensive function. Defined as: ,in, For capacity weight, For response characteristic weights, The spatial attenuation coefficient, For position arrive Electrical distance.

[0068] This model unifies and quantifies the static attributes, dynamic states, and spatial locations of energy storage, forming a computable value field.

[0069] 2. Value function of multi-agent negotiation.

[0070] In a multi-agent system, each agent makes decisions based on local information and a value field. The goal is to maximize the overall value return of the system through negotiation, while simultaneously satisfying its own operational constraints. Defining the agent... In strategy The expected return is:

[0071] ;

[0072] in, For system status, For intelligent agents The actions taken Represents the set of policies of all other agents. As a discount factor, It is an intelligent agent At any moment Local revenue, These are value field weighting coefficients, used to balance individual gains with the overall value absorbed by the system. Indicates the execution strategy Then, the system state evolved to The energy storage value field strength corresponding to the time.

[0073] Through multiple rounds of critical negotiation, the agents are essentially seeking a Nash equilibrium that satisfies the policy of all agents. satisfy: .

[0074] 3. Triggering conditions and optimization objectives for dynamic reconstruction of the value field.

[0075] When the system identifies a consumption bottleneck, it triggers a value field reconfiguration. The optimization objective of the reconfiguration is to adjust the weights of each energy storage unit. And the core value function parameters, which enable the new value field It can better guide multi-agent systems to generate effective drainage strategies. The reconstruction problem can be formulated as:

[0076] ;

[0077] ;

[0078] in: It is the reconstructed set of weights. It is the core value function A set of adjustable parameters Based on the reconstructed value field and multi-agent strategies Predicted system operating status These are the boundary values ​​for the safe operation of the system. It is a system state prediction model, which is usually a simplified power flow calculation or neural network model. It is a regularization coefficient used to prevent weight changes from being too drastic and to ensure system stability.

[0079] This optimization problem ensures that the restructured value field can effectively eliminate the current bottleneck while maintaining continuity with the original configuration as much as possible.

[0080] The above theoretical framework provides a solid mathematical foundation for this invention, ensuring the quantification, synergy, and adaptability of the process for addressing bottlenecks in new energy consumption. The specific implementation methods of this invention will be described in detail below.

[0081] Please see Figure 1 This invention illustrates the overall steps of a dynamic bottleneck-clearing method for renewable energy consumption based on energy storage value field topology reconstruction. The method includes:

[0082] S100: Construct a quantifiable energy storage value field to assess regulation potential.

[0083] S200: Establish a multi-agent system equipped with cognitive abilities to support collaborative decision-making.

[0084] S300: Monitors operational data in real time and automatically identifies bottlenecks to trigger refactoring.

[0085] S400: Generates the optimal dredging strategy through multi-agent negotiation based on the reconstructed value field.

[0086] S500: Executes strategies and continuously optimizes the configuration of energy storage value fields based on feedback.

[0087] In the dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction, S100 collects real-time operating data of all energy storage units in the area and performs data cleaning and standardization; initializes energy storage value field parameters, including setting initial values ​​for capacity weight, response characteristic weight and spatial attenuation coefficient; based on the energy storage value field quantification model, calculates the initial spatiotemporal value distribution on the set spatiotemporal grid, and generates a three-dimensional data volume containing longitude, latitude and time, which intuitively displays the energy storage value density at different locations and times.

[0088] Please see Figure 2 It illustrates the technical content of the present invention, a dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction, S100, including:

[0089] S110: Collects real-time operating data of all energy storage units within the area.

[0090] The system collects real-time operating data of all energy storage units within the region. In one possible implementation, the real-time operating data of the energy storage units includes: capacity status data, dynamic response characteristic data, and geographical location information.

[0091] Specifically, capacity status data includes the current state of charge, health status, and available capacity. Dynamic response characteristic data includes maximum charge / discharge power, ramp rate, response time constant, and efficiency curve. Geographic location information includes the node number of the energy storage unit connected to the grid, its latitude and longitude coordinates, and its electrical distance from key renewable energy power plants and load centers.

[0092] The collected raw data is cleaned, aligned, and standardized to eliminate dimensional differences, in preparation for subsequent unified calculations.

[0093] S120: Initialize energy storage value field parameters.

[0094] Initialize the core value function for each energy storage unit The parameters. Setting capacity weights. Response characteristic weights and spatial attenuation coefficient The initial value. In one possible implementation, the initial value is empirically set based on the type of energy storage unit, its rated power, and its traditional role in the power grid.

[0095] S130: Calculate the spatiotemporal value distribution.

[0096] According to the formula The initial energy storage value field is calculated on the set spatiotemporal grid. The spatial grid covers the entire regional power grid, while the temporal grid covers several future scheduling cycles. Through spatiotemporal value distribution calculations, a three-dimensional data volume is generated, including longitude, latitude, and time. This three-dimensional data volume visually displays the energy storage value density at different locations and times.

[0097] In the dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction, S200 constructs a system containing multiple types of intelligent agents such as new energy power plants, load aggregation, energy storage units and grid nodes, and defines its state space and action space; integrates a large language model finely tuned with power system domain knowledge for each intelligent agent to generate state descriptions, thought chain reasoning and preliminary action suggestions; designs communication protocols and multi-round negotiation mechanisms between intelligent agents to achieve collaborative decision-making in order to achieve a globally near-optimal strategy.

[0098] Please see Figure 3 It illustrates the technical content of the present invention, a dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction, S200, including:

[0099] S210: Definition and Modeling of Intelligent Agent Roles.

[0100] Construct a multi-agent system that includes new energy sources, loads, energy storage, and key grid components.

[0101] In one possible implementation, the multi-agent system includes: a new energy power station agent, a load aggregation agent, an energy storage unit agent, and a power grid node agent.

[0102] Specifically, the state space of the intelligent agent for new energy power plants includes predicted / actual output, prediction error, and grid connection point voltage; the action space includes output adjustment suggestions or power limit applications. The state space of the intelligent agent for load aggregation includes predicted / actual load, interruptible load ratio, and electricity price elasticity; the action space includes load reduction or transfer plans. The state space of the intelligent agent for energy storage units includes SOC and charging / discharging power limits; the action space includes charging / discharging power commands. The state space of the intelligent agent for grid nodes includes node voltage, phase angle, and power flow of connected lines; the action space includes reactive power compensation equipment switching and transformer tap adjustment.

[0103] S220: Integrating Domain-Specific Large Language Models and Cognitive Thinking Chains.

[0104] Each agent is equipped with a large language model fine-tuned using knowledge from the power system domain. This large language model describes and summarizes the complex, high-dimensional state data observed by the agent using natural language or structured text. For the current state and the goal, the model generates a progressive reasoning chain. Finally, based on this reasoning, preliminary action suggestions or policy options are generated.

[0105] For example, the thought process of an energy storage agent is as follows: The current SOC is 80%, which is high. The value field is very strong in my area, indicating a need for discharge support. My maximum discharge power is X MW. Considering the device lifespan, I propose discharging at a power of Y MW for Z hours.

[0106] S230: Establish a multi-agent communication and negotiation mechanism.

[0107] The design incorporates a communication protocol between agents, allowing them to exchange state summaries, thought chain reasoning processes, and policy proposals. The negotiation mechanism employs a multi-round iterative approach of proposal critique and improvement. For example, a discharge proposal from an energy storage agent might be criticized by a grid node agent, which points out that its discharge could cause overload of nearby lines, thus prompting the energy storage agent to modify its discharge strategy.

[0108] In the dynamic bottleneck unblocking method for new energy consumption based on energy storage value field topology reconstruction, after receiving the reconstruction trigger signal, the S300 dynamically adjusts the energy storage value field based on the optimization model, including updating the global weight and core value function parameters of each energy storage unit; each agent generates candidate actions based on the reconstructed value field, and predicts the system state evolution and value return through thought chain reasoning, and selects the optimal strategy; through multiple rounds of critical negotiation, the strategies of each agent gradually approach the maximization of the overall system value, forming a consensus unblocking strategy covering new energy, energy storage and load scheduling.

[0109] Please see Figure 4 It illustrates the technical content of the present invention, a dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction, S300, including:

[0110] S310: Synchronous acquisition of multi-source operating data.

[0111] Parallel acquisition of multi-source operational data from renewable energy power plant monitoring systems, power grid energy management systems, wide-area measurement systems, and load management systems. This multi-source operational data includes: renewable energy output fluctuations, line overload characteristics, and voltage limit exceedance characteristics.

[0112] Specifically, the fluctuation in new energy power output is calculated by comparing the actual and predicted active power output of each power station, and then determining the volatility and prediction deviation. Line overload characteristics are the ratio of the real-time power flow of critical transmission lines to their thermal stability limits. Voltage limit exceedance characteristics are the deviations of the voltage amplitude at each node from their safe operating upper and lower limits.

[0113] S320: Runs feature extraction and bottleneck identification algorithms.

[0114] The collected data is analyzed online, including the following specific analyses:

[0115] The line power flow margin is calculated, which reflects the relative distance between the current load level and the safety limit of the transmission line. The calculation formula is as follows:

[0116] ;

[0117] in: For the line At any moment The leeway of the trend, For the line The thermal stability limit or static safe transmission capacity, For the line At any moment The actual transmission power is expressed as an absolute value, taking into account bidirectional power flow.

[0118] When the margin is lower than the set threshold At that time, the line is marked as a potential overload risk.

[0119] Calculate the node voltage deviation, which reflects the degree of deviation of the node voltage from its rated value. The calculation formula is as follows:

[0120] ;

[0121] in: For nodes At any moment Voltage deviation, For nodes At any moment The actual voltage measurement value, For nodes The rated voltage level.

[0122] When the deviation exceeds the threshold At that time, the node is marked as having a voltage over-limit risk.

[0123] A comprehensive analysis of the relationship between renewable energy output fluctuations and network constraints reveals that if fluctuations lead to further deterioration of power flow on risky lines or voltage at risky nodes, then: and duration If so, it is determined that the system has a bottleneck in its absorption capacity.

[0124] S330: Triggers the reconstruction of the energy storage value field.

[0125] Once the S320 identifies a bottleneck, the system immediately generates a refactoring trigger signal. This signal includes the bottleneck's type, location, and severity. The refactoring trigger signal is then transmitted to the value field management module, initiating the S400's refactoring process.

[0126] In the dynamic bottleneck unblocking method for new energy consumption based on energy storage value field topology reconstruction, after receiving the reconstruction trigger signal, the S400 dynamically adjusts the energy storage value field based on the optimization model, including updating the global weight and core value function parameters of each energy storage unit; each agent generates candidate actions based on the reconstructed value field, and predicts the system state evolution and value return through thought chain reasoning, and selects the optimal strategy; through multiple rounds of critical negotiation, the strategies of each agent gradually approach the maximization of the overall system value, forming a consensus unblocking strategy covering new energy, energy storage and load scheduling.

[0127] Please see Figure 5 It illustrates the technical content of the present invention, a dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction, S400, including:

[0128] S410: Dynamic Reconfiguration of Energy Storage Value Field.

[0129] Upon receiving the reconstruction trigger signal from S330, and based on the aforementioned reconstruction optimization model, the current value field is... Rapid adjustments are made. In one possible implementation, the dynamic reconfiguration step of the energy storage value field includes:

[0130] Adjust the global weights of each energy storage unit. To address the bottleneck in energy consumption, fine-tune its core value function. In , and .

[0131] Solve the minimization problem using an online optimization algorithm. .in, The power flow model is simplified by inputting the current system state and preliminary strategy, specifically as follows: .

[0132] Solving for the optimal solution and Then, the reconstructed energy storage value field is calculated and published. .

[0133] S420: Individual strategy generation and evaluation based on value function.

[0134] Each agent receives the reconstructed value field. Then, based on its individual value function Perform strategic reasoning.

[0135] Each agent Based on its operational constraints, a set of feasible candidate actions is generated. For example, candidate actions for an energy storage agent include charging at power P1, discharging at power P2, and remaining idle.

[0136] For each candidate action intelligent agent It invokes its integrated domain-specific large language model to perform thought chain reasoning, predicting the evolution path of the system state after the action is performed. And estimate its return on value:

[0137] .

[0138] The model will reason: If I perform an action My local income How will it change? Simultaneously, how will the system state evolve, and what will be the value field strength corresponding to the evolved state? How much? How much system value will this bring me in return?

[0139] Each agent chooses to determine its predicted value return. Maximizing the action as its initial strategy and the strategy and its corresponding The value is sent as a proposal to the negotiation network.

[0140] S430: Multi-round critical negotiation based on maximizing value.

[0141] Through multiple rounds of negotiation, the agents critically evaluate the impact of each other's strategies on the global value, and work together to approach the system's optimality.

[0142] In one possible implementation, the multi-round critical negotiation step includes:

[0143] Round 1 – Proposal: Each agent proposes an initial action plan based on local information and value field.

[0144] The second round – critique: Agents evaluate the feasibility and global impact of each other's proposals. The grid node agents are responsible for assessing the system's security after all proposals are merged and providing critiques.

[0145] The third round – improvement and consensus: Agents that receive criticism modify their policies and resubmit them. This process iterates until all agents' policies reach a consensus that satisfies system security constraints and yields the expected value reward. Near-optimal set of consensus strategies Consensus Strategy Set satisfy: This set of strategies constitutes the optimal unblocking strategy covering the adjustment of new energy output, energy storage scheduling, and load response.

[0146] In the dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction, the S500 performs security verification on the optimal strategy generated through negotiation, and issues execution instructions after verification through high-precision digital power grid simulation; it monitors key power grid indicators in real time and collects feedback data; it evaluates the difference between the actual effect and the expected effect of the energy storage value field, and if there is a significant deviation, it initiates a lightweight fine-tuning mechanism to slightly optimize the weight of key energy storage units or the core value function parameters to improve the overall adaptability and robustness of the system.

[0147] Please see Figure 6 It illustrates the technical content of the present invention, a dynamic unblocking method for new energy consumption bottlenecks based on energy storage value field topology reconstruction, S500, including:

[0148] S510: Policy security verification and execution.

[0149] Before final execution, the optimal strategy generated through negotiation will be... The data is input into a high-precision digital power grid simulation model for a final safety check. Once the check passes, specific control commands are issued to each execution unit.

[0150] S520: Real-time monitoring and feedback data collection.

[0151] While implementing the strategy, key indicators of the power grid are monitored in real time, including but not limited to: power flow changes on previously identified bottleneck lines; voltage recovery at risk nodes; actual operating status of each energy storage unit; and actual output and power curtailment at renewable energy power plants.

[0152] S530: Energy Storage Value Field Adaptability Assessment.

[0153] The monitoring feedback data is compared with the predicted results when reconstructing the value field in S400. Performance indicators, such as bottleneck elimination rate and strategy execution deviation, are calculated. If the assessment finds that key system indicators have not been effectively improved, or that the strategy execution effect deviates significantly from expectations, then the current energy storage value field configuration is considered... It does not match the actual operational requirements.

[0154] S540: Value field fine-tuning mechanism.

[0155] When the S530 assessment deems optimization necessary, a fine-tuning mechanism is initiated. Fine-tuning differs from the S410 reconfiguration; it typically employs a lighter-weight algorithm, adjusting only the weights of a few key energy storage units. Make minor adjustments, or simply update the time-varying parameters in the core value function.

[0156] Example 2

[0157] This invention was demonstrated in a provincial power grid region with a high proportion of wind and solar power, where a distributed energy storage system totaling 200MW / 400MWh was deployed. The specific configuration used during implementation was as follows:

[0158] Data Layer: Based on a cloud-edge collaborative architecture, the regional control center deploys a cloud-based data lake to aggregate SCADA, PMU, and energy storage BMS data; edge computing gateways are deployed at each site to perform data preprocessing and caching.

[0159] Value field calculation layer: adopts a high-performance computing cluster to update the global energy storage value field every 5 minutes, with spatiotemporal resolutions of 1km×1km and 15 minutes respectively.

[0160] Multi-Agent Platform: A distributed agent runtime environment is developed based on the Ray framework, with each agent encapsulated as an independent Actor. The domain-specific large language model is based on ChatGLM6B and fine-tuned using over 100,000 power system dispatching procedures, accident analysis reports, and expert decision-making cases.

[0161] Decision-making and execution layer: The negotiation and decision-making cycle is 15 minutes. The generated control commands are sent to the power grid dispatch automation system for execution through the forward isolation device from safety zone III to zone II.

[0162] During the three-month trial operation, the system successfully handled 17 instances of localized power consumption bottlenecks caused by wind power peak shaving and sudden photovoltaic power drops. Compared with traditional centralized optimization scheduling, this invention demonstrates the following advantages:

[0163] 1) The average time from identification to strategy generation is less than 3 minutes, which is about 70% faster than traditional methods.

[0164] 2) The wind and solar curtailment rate in the demonstration area decreased by 5.8 percentage points year-on-year.

[0165] 3) Energy storage increased its average value contribution in the process of unblocking bottlenecks by 25%.

[0166] 4) Through value field fine-tuning, the system can automatically restore optimized performance within 23 decision cycles after changes in the power grid topology (such as planned maintenance).

[0167] Specifically, in a section exceeding its limit event caused by a fault in the main line N1, this invention triggered reconfiguration within 45 seconds, guiding the downstream energy storage cluster to switch to maximum discharge mode within one minute. Simultaneously, it coordinated with the upstream wind farm cluster to implement power-limited operation, successfully controlling the power flow within safe limits and avoiding load shedding. The subsequent value field fine-tuning mechanism automatically reduced the cluster's weight based on the rapid decrease in energy storage SOC, preventing overuse and demonstrating the method's intelligence and robustness.

[0168] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0169] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for dynamically addressing bottlenecks in new energy consumption based on topological reconstruction of the energy storage value field, characterized in that, include: Based on the capacity status, dynamic response characteristics, and geographical location information of each energy storage unit in the region, a dynamically quantifiable energy storage value field is constructed. The adjustment potential of the energy storage value field on the spatiotemporal scale is evaluated by multi-dimensional weight assessment to form an initial energy storage value field; Construct a multi-agent system that includes new energy sources, loads, energy storage, and key grid components; each agent is equipped with a domain-specific large language model and dynamically generated cognitive thought chains to support agents in strategic reasoning and collaborative decision-making in complex operating environments; Real-time collection of multi-source operation data from new energy power plants, grid nodes and load centers; analysis of output fluctuation characteristics, line overload characteristics and voltage over-limit characteristics; identification of whether there is a consumption bottleneck; once a consumption bottleneck is identified, the energy storage value field reconstruction mechanism is automatically triggered. Based on the reconstructed energy storage value field as the basis for strategy evaluation, through multiple rounds of critical negotiation and strategy iteration among each agent in the multi-agent system, an optimal unblocking strategy covering new energy output adjustment, energy storage scheduling and load response is formed. Execute the generated optimal unblocking strategy and monitor key power grid indicators and the status of each execution unit in real time; The adaptability of the energy storage value field is assessed based on monitoring feedback data. If it is found that the value field does not match the operational requirements, a fine-tuning mechanism is initiated.

2. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, The spatiotemporal value function for constructing the energy storage value field: ； in: Indicates spatial location ,time The energy storage value field intensity, This represents the total number of energy storage units in the region. For the first The global weighting coefficient of each energy storage unit; It is the first Spatial influence function of each energy storage unit ,in For energy storage units Geographic coordinates To influence the radius parameter; It is the first The core value function of the first energy storage unit is the first... Energy storage unit capacity status Dynamic response characteristics and geographical location attributes A comprehensive function; Defined as: ,in, For capacity weight, For response characteristic weights, The spatial attenuation coefficient, For position arrive Electrical distance.

3. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, The multi-agent system includes a new energy power station agent, a load aggregation agent, an energy storage unit agent, and a power grid node agent; Among them, the state space of the intelligent agent of the new energy power station includes predicted / actual output, prediction error, and grid connection point voltage, while the action space includes output adjustment suggestions or power limit applications. The state space of the load aggregation agent includes the predicted / actual load, the proportion of interruptible load, and the electricity price elasticity; the action space is the load reduction amount or the transfer plan. The state space of the energy storage unit's intelligent agent includes the state of charge and the charging / discharging power limit, while the action space consists of charging / discharging power commands. The state space of a power grid node agent includes node voltage, phase angle, and power flow of connected lines, while the action space includes switching of reactive power compensation equipment or adjustment of transformer taps.

4. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, Each agent is equipped with a domain-specific large language model, which is fine-tuned with knowledge of the power system domain. This model is used to transform the high-dimensional state data observed by the agent into natural language or structured text descriptions, and to generate a thought chain of progressive reasoning. Based on the thought chain, it outputs preliminary action suggestions or strategy options.

5. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, The identified bottlenecks include: The formula for calculating the power flow margin of a line is as follows: ； in: For the line At any moment The leeway of the trend, For the line The thermal stability limit or static safe transmission capacity, For the line At any moment The actual transmission power, expressed as an absolute value considering bidirectional power flow; when the power flow margin is lower than a set threshold... At that time, the line is marked as a potential overload risk; The formula for calculating node voltage deviation is as follows: ； in: For nodes At any moment Voltage deviation, For nodes At any moment The actual voltage measurement value, For nodes The rated voltage level; when the deviation exceeds the threshold At that time, mark the node as having a voltage over-limit risk; If fluctuations cause further deterioration of power flow on risky lines or voltage at risky nodes, that is: and duration If so, it is determined that the system has a bottleneck in its absorption capacity.

6. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, The mechanism for triggering the reconstruction of the energy storage value field includes: The global weights of each energy storage unit are dynamically adjusted based on the reconfiguration optimization model. And the core value function parameters, the reconstructed optimization model is expressed as: ； in: It is the reconstructed set of weights. It is the core value function A set of adjustable parameters These are the boundary values ​​for the safe operation of the system. Based on the reconstructed value field and multi-agent strategies Predicted system operating status , It is a system state prediction model. It is the regularization coefficient; Solving for the optimal solution and Then, the reconstructed energy storage value field is calculated and published. .

7. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, The multi-round critical negotiation and strategy iteration include: In the first round, each agent proposes an initial action plan based on local information and the reconstructed value field. In the second round, the agents evaluate the feasibility and global impact of each other's proposals, and the power grid node agents conduct a system security critique. In the third round, the agents modify their strategies based on the feedback and resubmit them, iterating until all agents' strategies form a set that satisfies the system's security constraints and yields the expected value return. Near-optimal set of consensus strategies ; Define intelligent agents In strategy The expected value return is : ； in, For system status, For intelligent agents The actions taken, Represents the set of policies of all other agents. As a discount factor, It is an intelligent agent At any moment Local revenue, These are value field weighting coefficients, used to balance individual gains with the overall value absorbed by the system. Indicates the execution strategy Then, the system state evolved to The energy storage value field strength corresponding to the time; Consensus Strategy Set satisfy: .

8. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, Before executing the generated optimal unblocking strategy, the following steps are also included: The optimal unblocking strategy is input into a high-precision digital power grid simulation model for safety verification. After the verification is passed, control commands are sent to each execution unit.

9. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, The fine-tuning mechanism includes: Performance metrics, including bottleneck elimination rate and strategy execution deviation, are calculated based on monitoring feedback data. If performance metrics do not meet expectations, a lightweight algorithm will be used to weight the key energy storage units. Alternatively, the core value function parameters may be slightly adjusted to improve the match between the energy storage value field and operational needs.

10. The method for dynamically unblocking new energy consumption bottlenecks based on energy storage value field topology reconstruction according to claim 1, characterized in that, The multi-agent system exchanges state summaries, thought chain reasoning processes, and strategy suggestions through a communication protocol. The negotiation mechanism adopts a multi-round iterative mode of proposal-critique-improvement to achieve the generation of a globally near-optimal unblocking strategy.

Images