A cloud computing-based intelligent substation energy management method

Abstract

The application belongs to the field of power dispatching management, and discloses an intelligent substation energy management method based on cloud computing, which comprises the following steps: constructing a "prediction-dispatching" game framework; generating a three-dimensional uncertainty cloud picture with space-time coupling characteristics by fusing regional meteorological data and adjacent substation output correlation, solving a Nash equilibrium point based on the three-dimensional uncertainty cloud picture in an edge layer, and generating a dispatching scheme; performing dynamic electrical partitioning based on a feeder voltage-power sensitivity matrix to obtain a partitioning result, implanting programmable impedance characteristics in a photovoltaic inverter and an energy storage power conversion system, and performing equipment autonomous response based on the dispatching scheme; performing linkage control across time scales based on a source-load fluctuation event chain model and the partitioning result; updating the dispatching scheme based on an execution result, adjusting equipment operating states based on the updated dispatching scheme, and realizing intelligent substation energy management based on cloud computing.

Description

Technical Field

[0001] This invention relates to the field of power dispatching and management, and more specifically, to a cloud computing-based intelligent substation energy management method. Background Technology

[0002] As the proportion of new energy sources (such as photovoltaic and wind power) in the power system continues to increase, the volatility of their output and the randomness of load demand pose severe challenges to the dispatching and operation of the power system. The output of new energy sources is affected by meteorological conditions such as sunlight and wind speed, exhibiting power fluctuations on a minute-by-minute basis. This strong volatility leads to forecasting errors (especially in short-term and ultra-short-term forecasts), directly impacting the accuracy of dispatching plans. Specifically, if the actual output of new energy sources exceeds the forecast, it may cause overvoltage problems in the grid, threatening grid stability; if the actual output is lower than the forecast, emergency use of backup resources (such as energy storage devices) is required for power compensation, significantly increasing operating costs.

[0003] Meanwhile, the randomness of load demand further exacerbates the operational difficulties of the power system. Sudden increases or decreases in industrial user loads, as well as the unpredictability of residential electricity consumption, cause load curves to frequently deviate from predicted values. This prediction deviation necessitates real-time balancing of unplanned power deficits or redundancies, placing extreme demands on the response speed of the dispatch system, typically requiring adjustments to be completed within seconds.

[0004] However, existing technologies mainly rely on complex algorithm models to improve the accuracy of new energy output prediction. But these methods often lack economic incentive mechanisms for the prediction subject and cannot fundamentally solve the prediction error problem. Especially when faced with minute-level power fluctuations, the prediction accuracy is still limited and the scheduling efficiency is low, resulting in equipment response delays.

[0005] In view of this, a cloud computing-based intelligent substation energy management method is designed. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the existing technology and to achieve the above objectives, the present invention provides the following technical solution: a cloud computing-based intelligent substation energy management method, comprising: constructing a "prediction-scheduling" game framework at the source-load prediction layer, wherein new energy operators and load aggregators submit prediction commitment intervals to the cloud platform, and implement reward and punishment rules based on actual deviations; and generating a three-dimensional uncertainty cloud map with spatiotemporal coupling characteristics by integrating regional meteorological data and the output correlation of adjacent substations, which serves as the decision basis for the scheduling and allocation layer.

[0007] At the dispatch and allocation layer, substation resources are divided into four types of autonomous decision-making entities: new energy clusters, energy storage alliances, adjustable load groups, and grid interface agents. The edge layer solves for the Nash equilibrium point based on the three-dimensional uncertainty cloud map to generate a dispatch scheme. Dynamic electrical partitioning is performed based on the feeder voltage-power sensitivity matrix to obtain the partitioning results. The dispatch scheme and partitioning results serve as inputs to the execution control layer.

[0008] At the execution control layer, programmable impedance characteristics are embedded in the photovoltaic inverter and energy storage power conversion system to enable autonomous equipment response based on the scheduling scheme; the source-load fluctuation event chain model performs cross-time scale linkage control based on the partitioning results; and the execution results are fed back to the integration layer.

[0009] At the integration layer, the cloud platform performs game arbitration and market settlement based on the execution results. The edge layer uses quantum-inspired algorithms to accelerate the search for equilibrium points and updates the scheduling scheme based on the execution results. The terminal layer dynamically reconstructs the impedance characteristics of the equipment through power electronic switch arrays and adjusts the operating status of the equipment based on the updated scheduling scheme.

[0010] Preferably, the method for generating the three-dimensional uncertainty cloud map includes:

[0011] The game participants are initialized, with new energy operators and load aggregators as the main participants in the game framework; the game rules are set, including that if the actual value is within the predicted commitment range, the participant will receive scheduling priority and economic compensation, and if the actual value exceeds the predicted commitment range, the participant will pay a deviation penalty, which is injected into the standby resource pool as an incentive mechanism for the game framework; the participants submit the initial predicted commitment range as the input to the game framework to construct the "prediction-scheduling" game framework.

[0012] Based on the game theory framework, environmental data is decomposed hierarchically to extract disturbance features at different time and spatial scales. The environmental data includes wind speed, light intensity and load demand, and the disturbance features include minute-level abrupt change features, hourly-level trend features, local abrupt change features and regional trend features, which serve as the basis for predicting the reshaping of the commitment interval.

[0013] For perturbation features of different standards, minute-level mutation features have higher priority than hour-level trend features, local mutation features have higher priority than regional trend features, and for perturbation features of the same standard, priority is ranked according to the degree of impact of the perturbation features on the prediction accuracy.

[0014] Based on the priority sequence and the strength of the perturbation features, the leniency of the initial prediction commitment interval is dynamically calibrated. That is, if a high-priority minute-level mutation feature is detected, the tolerance range of the prediction commitment interval is significantly widened and the penalty intensity is reduced. If only a medium-priority hour-level trend feature is detected, the prediction commitment interval is moderately widened to generate an adjusted prediction commitment interval, which serves as the update input for the game framework.

[0015] Based on the adjusted forecast commitment range, new energy operators and load aggregators submit forecast commitments to the cloud platform, and implement reward and penalty rules according to the actual deviations to generate forecast deviation data, which serves as the basis for correcting the generation of the three-dimensional uncertainty cloud map.

[0016] By integrating prediction bias data, regional meteorological data, and the output correlation of adjacent substations, a three-dimensional uncertainty cloud map with spatiotemporal coupling characteristics is generated. The three-dimensional uncertainty cloud map includes the distribution of time, power, and probability of occurrence.

[0017] Preferably, the method for generating a scheduling scheme by solving for the Nash equilibrium point based on a three-dimensional uncertainty cloud map includes:

[0018] Substation resources are divided into four types of autonomous decision-making entities: new energy clusters, energy storage alliances, adjustable load groups, and grid interface agents. The goal of the new energy clusters is to maximize the absorption capacity, the goal of the energy storage alliances is to minimize the lifespan loss, the goal of the adjustable load groups is to minimize the electricity cost, and the goal of the grid interface agents is to meet the upper-level dispatch instructions, serving as the basis for strategy conflict negotiation.

[0019] Conflict features are extracted from the objective functions of each autonomous decision-making body to generate a conflict feature set. The conflict features include objective priority features and constraint boundary features. The objective priority features include the priority of the absorption capacity of the new energy cluster and the priority of the life loss of energy storage. The constraint boundary features include the SOC protection boundary of energy storage and the reduction capacity boundary of adjustable load, which serve as the basis for hierarchical negotiation.

[0020] The conflict feature set is decomposed into multiple conflict levels based on the severity and scope of the conflict. The conflict levels include the initial conflict level and the severe conflict level. The initial conflict level corresponds to the conflict where the SOC is close to exceeding the limit, and the severe conflict level corresponds to the conflict where the SOC actually exceeds the limit. These levels serve as the input for constraint convergence.

[0021] Based on the three-dimensional uncertainty cloud map, a multi-stage negotiation mechanism is used to resolve policy conflicts at the conflict level. In the initial conflict level, the new energy cluster adjusts its output plan to meet the SOC protection boundary of energy storage. In the severe conflict level, the energy storage alliance relaxes the SOC protection boundary. The negotiation process adopts an iterative approach, with each iteration narrowing the feasible domain of the strategy and generating a negotiated strategy set as input for solving the Nash equilibrium.

[0022] The edge layer solves for the Nash equilibrium point based on the negotiated strategy set and the three-dimensional uncertainty cloud map, and generates a scheduling scheme. The scheduling scheme includes the active / reactive power adjustment range, charging and discharging strategy, load transfer plan and power purchase and sale plan of each entity.

[0023] Preferably, the method for performing dynamic electrical partitioning based on the feeder voltage-power sensitivity matrix and obtaining the partitioning results includes:

[0024] The substation status data is decomposed into multiple dimensions to extract change features. The substation status data includes voltage, power and power flow distribution. The change features serve as the basis for disturbance perception, including instantaneous change features, trend change features and spatial distribution features.

[0025] Based on the three-dimensional uncertainty cloud map, the change characteristics are comprehensively evaluated to generate the disturbance index of the substation status. The comprehensive evaluation is based on the intensity and duration of the change characteristics. If the instantaneous change characteristics are strong and short in duration, the disturbance index is high. If the trend change characteristics are weak and long in duration, the disturbance index is low. This serves as the input for frequency optimization.

[0026] Based on the disturbance index, the update frequency of the feeder voltage-power sensitivity matrix is ​​dynamically adjusted. If the disturbance index is higher than the first threshold, the update frequency is increased to the first frequency value. If the disturbance index is lower than the second threshold, the update frequency is decreased to the second frequency value, generating an optimized update frequency as the basis for dynamic electrical zoning.

[0027] Based on the optimized update frequency and scheduling scheme, the feeder voltage-power sensitivity matrix is ​​calculated, and dynamic electrical partitioning is generated. The dynamic electrical partitioning includes a strongly coupled region and a weakly coupled region. The strongly coupled region performs centralized optimization scheduling, while the weakly coupled region performs autonomous decision scheduling, and the partitioning results are generated.

[0028] When the partition changes, the scheduling rules are adjusted through the continuous optimization function to generate the adjusted partition result, which serves as the update input for the linkage control of the execution control layer.

[0029] Preferably, the method for embedding programmable impedance characteristics into the photovoltaic inverter and energy storage power conversion system, and for autonomous equipment response based on a scheduling scheme, includes:

[0030] The equipment's operational data is decomposed into multiple levels to extract impact features. The operational data includes the operating status of the photovoltaic inverter and the energy storage power conversion system, while the impact features include geographical impact features, power impact features, and sensitivity impact features, which serve as the basis for equipment impact assessment.

[0031] Based on the partitioning results, the impact features are prioritized and evaluated to generate a comprehensive importance index for the equipment. The evaluation criteria are the contribution of the impact features to the system stability. If the geographical impact features of the equipment show that it is located in a strong coupling area and the sensitivity impact features show that it has a large impact on voltage, then the comprehensive importance index is high and it is used as the input for priority adjustment.

[0032] Based on the comprehensive importance index, the impedance parameter adjustment priority of the equipment is dynamically sorted. Among them, the equipment with a high comprehensive importance index is adjusted first, while the equipment with a low comprehensive importance index is adjusted later. This generates an equipment adjustment sequence as the basis for the equipment's autonomous response.

[0033] Based on the scheduling scheme and the equipment adjustment sequence, programmable impedance characteristics are embedded in the photovoltaic inverter and energy storage power conversion system to execute autonomous equipment response. The autonomous response includes the simulated synchronous machine inertial response of the new energy equipment and the short-circuit capacity support provided by the energy storage equipment, generating equipment response results as inputs for the linkage control of the execution control layer.

[0034] When the system state changes, the comprehensive importance index is re-evaluated through a continuous update function, and the equipment adjustment sequence is adjusted to generate an updated equipment adjustment sequence, which serves as the update input for the linkage control of the execution control layer.

[0035] Preferably, the method for cross-timescale linkage control based on partitioning results using the source-load fluctuation event chain model includes:

[0036] The indicators involved in the event chain are decomposed to extract the impact features. The indicators include photovoltaic output, load demand and voltage level. The impact features include instantaneous impact features, trend impact features and coupling impact features, which serve as the basis for correlation and integration.

[0037] Based on the partitioning results, dynamic weights are assigned to the influencing features to generate a weight sequence. The assignment is based on the degree of influence of the influencing features on the stability of the substation. If the instantaneous influencing features have a large impact on the stability of the substation, they are assigned high weights. If the trend influencing features have a small impact, they are assigned low weights. The weights are dynamically adjusted according to the strong coupling and weak coupling regions of the partitioning results and serve as inputs for conditional optimization.

[0038] By using a weighted fusion function to correlate and fuse the influencing features and weight sequences, composite triggering conditions are generated, and a comprehensive influence index is generated, which serves as the basis for coordinated control.

[0039] Based on the equipment response results and composite triggering conditions, cross-timescale linkage control is executed. The linkage control includes millisecond-level energy storage compensation, second-level load shedding and minute-level grid support, generating linkage control results as input to the execution results of the integration layer.

[0040] The trigger threshold of the composite triggering condition is dynamically optimized based on the disturbance index of the substation status. If the disturbance index of the substation status is high, the trigger threshold is lowered; if the disturbance index of the substation status is low, the trigger threshold is raised. The optimized composite triggering condition is generated and used as the update input for the execution result of the integration layer.

[0041] Preferably, the method for game arbitration and market settlement based on the execution results of the cloud platform includes:

[0042] The substation operation scenarios are decomposed in multiple dimensions to extract impact characteristics. The operation scenarios include voltage collapse risk, insufficient renewable energy absorption and load change. The impact characteristics include disturbance impact characteristics, range impact characteristics and economic impact characteristics, which serve as the basis for scenario impact assessment.

[0043] Based on the partitioning results, the impact features are prioritized and a comprehensive emergency index for the scenario is generated. The evaluation criteria are the degree of impact of the impact features on the substation operation. If the disturbance impact features show a high risk of voltage collapse and the range impact features show a strong coupling area, the emergency index is high, which serves as an input for funding optimization.

[0044] The allocation ratio of backup resource pool funds is dynamically optimized based on the comprehensive emergency index. If the comprehensive emergency index of a scenario is higher than the first threshold, funds are allocated first to purchase energy storage services. If the emergency index is lower than the second threshold, funds are allocated first to subsidize affected participants. A fund allocation plan is generated as the basis for game arbitration and market settlement.

[0045] Based on the equipment response results and linkage control results, the technical feasibility of the scheduling scheme is verified. The technical feasibility includes power flow security and voltage stability. An arbitration result is generated as input for market clearing.

[0046] Based on the arbitration results and the fund allocation plan, the settlement of reward and penalty funds and ancillary service fees is carried out. The reward and penalty funds include the forecast deviation penalty and economic compensation for new energy operators and load aggregators, and the ancillary service fees include the fees for energy storage services and load reduction services. The settlement results are generated and used as input for the edge layer update scheduling scheme.

[0047] Preferably, the method for accelerating equilibrium point search using a quantum-inspired algorithm in the edge layer and updating the scheduling scheme based on the execution results includes:

[0048] The execution results are decomposed in multiple dimensions to extract deviation features. The execution results include equipment response results and linkage control results, and the deviation features include equipment response deviation features and linkage control deviation features, which serve as the basis for strategy optimization.

[0049] Based on the scheduling scheme, the deviation characteristics are prioritized and a deviation priority sequence is generated. The evaluation criteria are the degree of impact of the deviation characteristics on the stability and economy of the substation. If the equipment response deviation characteristics show that the power regulation exceeds the range of the scheduling scheme, the priority is high. If the linkage control deviation characteristics show that the voltage fluctuation exceeds the safe range, the priority is high. This serves as the input for the equilibrium point search.

[0050] A quantum-inspired algorithm is used to accelerate the search for equilibrium points based on the liquidation results. The liquidation results are used to adjust the reward and punishment rules of the game framework and generate an optimized strategy set, which serves as the basis for updating the scheduling scheme.

[0051] Based on the optimized strategy set and deviation priority sequence, the scheduling scheme is optimized. The optimization includes adjusting the active / reactive power regulation range, charging and discharging strategy, load transfer plan and power purchase and sale plan of each subject, and generating an updated scheduling scheme as input for the terminal layer to adjust the operating status of equipment.

[0052] When the system state changes, the deviation priority sequence is re-evaluated through the continuous optimization function, and the optimized strategy set is adjusted to generate an adjusted update scheduling scheme, which serves as the update input for the terminal layer to adjust the device operating state.

[0053] Preferably, the method for updating the scheduling scheme based on the execution result further includes:

[0054] After adding the settlement results to the execution results, multi-dimensional integration is performed to extract optimization features, including equipment response optimization features, linkage control optimization features, and economic optimization features, which serve as the basis for rolling optimization.

[0055] Based on the scheduling scheme, the optimization features are prioritized and an optimization priority sequence is generated. The evaluation criteria are the degree to which the optimization features improve the stability and economy of the substation. If the equipment response optimization features show a reduction in power regulation deviation, the priority is high. If the economic optimization features show a reduction in cost, the priority is high. This serves as the input for rolling optimization.

[0056] Based on the optimized priority sequence and the updated scheduling scheme, the scheduling scheme is rolled out for optimization. Rolling optimization includes refreshing the three-dimensional uncertainty cloud map and the scheduling scheme at preset time intervals to generate an optimized scheduling scheme, which serves as the input for the terminal layer to adjust the device's operating status.

[0057] When a sudden event is detected, the system immediately switches to emergency mode based on the linkage control results. In emergency mode, rapid response resources are prioritized, including energy storage devices and adjustable load resources. An emergency management plan is generated as a temporary input for the terminal layer to adjust the operating status of the equipment.

[0058] When the substation status changes, the optimization priority sequence is re-evaluated through a continuous optimization function, and the optimized scheduling scheme is adjusted to generate an adjusted optimized scheduling scheme, which serves as the update input for the terminal layer to adjust the operating status of the equipment.

[0059] Preferably, the terminal layer dynamically reconstructs the device impedance characteristics through a power electronic switch array and adjusts the device operating status based on the updated scheduling scheme. The method includes:

[0060] The updated scheduling scheme is decomposed in multiple dimensions to extract adjustment features. The updated scheduling scheme includes the updated scheduling scheme and the optimized scheduling scheme. The adjustment features include impedance adjustment features, power adjustment features and time adjustment features, which serve as the basis for dynamic reconfiguration.

[0061] Based on the adjustment characteristics, the impedance characteristics of the equipment are dynamically reconstructed through the power electronic switch array. The impedance characteristics include the real and imaginary parts of the virtual impedance. The reconstruction includes the simulated synchronous machine inertial response of the new energy equipment and the short-circuit capacity support provided by the energy storage equipment. The reconstructed impedance characteristics are generated as the input for adjusting the operating state of the equipment.

[0062] Based on the reconstructed impedance characteristics, the equipment operating status is adjusted, which includes actual output, state of charge, and impedance value, to generate the adjusted equipment operating status as the input for the next round of prediction.

[0063] The adjusted device operating status is fed back to the edge layer in real time to update the execution results, generate the feedback execution results, and serve as the input for the next round of game arbitration and market settlement in the integration layer;

[0064] When the substation status changes, the adjustment characteristics are re-evaluated through a continuous optimization function, and the reconstructed impedance characteristics are adjusted to generate an updated equipment operating status, which serves as the update input for the next round of prediction.

[0065] The technical effects and advantages of the cloud computing-based intelligent substation energy management method of the present invention are as follows:

[0066] By constructing a "prediction-scheduling" game framework, economic incentives are used to incentivize new energy operators and load aggregators to improve prediction accuracy. The dynamic calibration logic of the prediction commitment interval can adaptively adjust the slackness according to the hierarchical characteristics of environmental disturbances, significantly reducing prediction errors. By transforming uncertainty into an economically driven problem, participants are incentivized to actively optimize the prediction model, reducing reliance on algorithmic complexity and improving the robustness and fairness of predictions.

[0067] By dividing substation resources into four types of autonomous decision-making entities—new energy clusters, energy storage alliances, adjustable load groups, and grid interface agents—and employing a multi-stage negotiation mechanism and Nash equilibrium solution, distributed autonomous decision-making is achieved. The dynamic electrical zoning logic adaptively divides the control granularity based on the feeder voltage-power sensitivity matrix, optimizing scheduling efficiency. This significantly improves the real-time performance and relevance of scheduling scheme generation, making it particularly suitable for complex fluctuation management in scenarios with a high proportion of new energy integration.

[0068] By embedding programmable impedance characteristics into photovoltaic inverters and energy storage power conversion systems, the equipment can autonomously adjust its electrical characteristics according to the scheduling scheme, enabling new energy equipment to simulate the inertial response of a synchronous machine and providing short-circuit capacity support for energy storage equipment. Cross-timescale linkage control logic optimizes response speed and stability through composite triggering conditions. Physical-level intelligent reconfiguration of equipment characteristics eliminates command latency issues, significantly improving second-level response capabilities and enhancing the system's dynamic stability.

[0069] By leveraging game arbitration and market clearing on the cloud platform, combined with quantum-inspired algorithms at the edge layer to accelerate equilibrium point search, real-time updates and rolling optimization of the scheduling scheme are achieved. At the terminal layer, dynamic reconstruction of device impedance characteristics via power electronic switch arrays ensures efficient execution of the scheduling scheme. This forms a closed-loop management mechanism encompassing prediction, scheduling, execution, and integration, significantly improving the system's economic self-consistency and physical executability, and optimizing resource utilization efficiency. Attached Figure Description

[0070] Figure 1 This is a schematic diagram illustrating the steps of a cloud computing-based intelligent substation energy management method according to the present invention.

[0071] Figure 2 This is a schematic diagram of the structure of a cloud computing-based intelligent substation energy management method according to the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Example 1

[0074] Please see Figure 1 As shown in this embodiment, a cloud computing-based intelligent substation energy management method includes:

[0075] Traditional scheduling schemes often employ a centralized optimization paradigm, using a single control center to uniformly schedule all resources. However, this centralized scheduling approach suffers from high computational complexity in scenarios with a high proportion of renewable energy integration, making it difficult to meet the rapid response requirements of distributed energy systems. In particular, its scheduling efficiency is significantly insufficient when facing real-time power deficits or redundancies.

[0076] Existing execution control layers primarily rely on traditional power command control methods, responding by issuing specific power adjustment commands to devices. However, this approach suffers from response latency, especially in scenarios requiring millisecond-level response times, making it difficult to meet system stability requirements. Furthermore, existing devices lack intelligent features and cannot autonomously adjust their operational behavior based on system status.

[0077] Existing technologies lack a closed-loop management mechanism at the system integration level that ensures economic consistency and physical feasibility. Traditional game arbitration and market settlement methods often fail to adequately consider feedback on execution results, leading to a lack of real-time and targeted optimization of scheduling schemes. Furthermore, the allocation of funds in the backup resource pool is relatively simplistic, making it difficult to dynamically optimize based on the urgency of different scenarios, thus reducing the system's economic efficiency and stability.

[0078] Therefore, there is an urgent need for a cloud-based intelligent substation energy management method. This method should address the core challenges posed by source-load uncertainty through innovative prediction mechanisms, scheduling strategies, execution control, and system integration, achieving a paradigm shift from "scheduling systems adapting to fluctuations" to "fluctuations being tamed into dispatchable resources." It should effectively address the complex fluctuation issues in scenarios with high proportions of renewable energy integration, providing an innovative scheduling path for intelligent substation operation that is both economically self-consistent and physically feasible.

[0079] The output of new energy sources (such as photovoltaics and wind power) is affected by meteorological conditions such as sunlight and wind speed, exhibiting power fluctuations on a minute-by-minute basis. Prediction errors directly impact the accuracy of dispatch plans, potentially leading to back-feeding overvoltages to the grid or the emergency use of reserve resources, increasing operating costs. Sudden increases / decreases in industrial user loads and uncertainties in residential electricity consumption behavior cause load curves to deviate from predictions. Unplanned power deficits or redundancies need to be balanced in real time, placing extreme demands on the response speed of the dispatch system. Existing forecasting methods lack economic incentives for forecasters, making it difficult to fundamentally solve the forecasting error problem, especially when facing minute-by-minute power fluctuations, where forecast accuracy is limited.

[0080] This design constructs a "prediction-scheduling" game framework, introduces hierarchical perception of environmental disturbances and interval reshaping logic, dynamically adjusts the prediction commitment interval, and generates a three-dimensional uncertainty cloud map with spatiotemporal coupling characteristics, thereby improving prediction accuracy, quantifying uncertainty, and providing a reliable decision-making basis for the scheduling and allocation layer.

[0081] The specific content includes:

[0082] At the source-load forecasting layer, a "forecast-scheduling" game framework is constructed, in which new energy operators and load aggregators submit forecast commitment intervals to the cloud platform, and reward and punishment rules are implemented based on actual deviations; by integrating regional meteorological data and the output correlation of adjacent substations, a three-dimensional uncertainty cloud map with spatiotemporal coupling characteristics is generated as the decision basis for the scheduling and allocation layer.

[0083] The method for generating the three-dimensional uncertainty cloud map includes:

[0084] Initializing game participants involves registering participants on the cloud platform, submitting basic information, and establishing data interfaces. The steps are as follows:

[0085] Step A1: New energy operators and load aggregators register accounts through the cloud platform's user management system to obtain a unique participant identifier (such as an ID number).

[0086] Step A2: Participants need to submit their basic information, including name, geographical location, device capacity, historical operating data, etc. This information is submitted through the cloud platform's form or data upload interface and stored in the cloud database;

[0087] Step A3: The cloud platform allocates a dedicated data interface to each participant for real-time uploading of prediction data, receiving scheduling instructions, and obtaining reward and penalty results. The data interface uses standard communication protocols (such as MQTT or REST API) to ensure the security and real-time performance of data transmission.

[0088] New energy operators [referring to entities responsible for the operation and management of new energy power generation equipment (such as photovoltaic power plants and wind farms), whose main task is to predict new energy output and participate in power dispatch. For example, company A, the operator of a photovoltaic power plant, is responsible for managing a photovoltaic power plant with an installed capacity of 50MW, predicting its output for the next 15 minutes, and submitting a prediction commitment range to the cloud platform (e.g., "the probability of output ≥ 40MW is 90%")) and load aggregators [referring to entities that represent multiple load users (such as industrial users and residential users) to manage and predict loads, whose main task is to aggregate load demand, predict load curves, and participate in power dispatch. For example, company B, representing 10 factories in an industrial park, predicts their total load demand for the next hour and submits a prediction commitment range to the cloud platform (e.g., "the probability of load demand ≤ 100MW is 85%")] are the participating entities in the game framework;

[0089] The game rules are defined, including: if the actual value falls within the predicted commitment range [referring to the range of output or load predicted by a participant and its corresponding probability. For example, new energy operator A predicts that its photovoltaic power station's output will be 40MW in the next 15 minutes and commits that "the probability of output ≥ 40MW is 90%", i.e., the predicted commitment range is [40MW, ∞) and the probability is 90%], then it will obtain dispatch priority [referring to giving priority to the participant's output or load demand in power dispatch. For example, if A's actual output is within the predicted commitment range (e.g., actual output is 45MW), then in the dispatch allocation layer, A's photovoltaic power station will have priority in obtaining consumption rights, i.e., its output will be prioritized for inclusion in the dispatch plan and will not be reduced], and economic compensation [referring to the economic reward given to the participant with accurate predictions. For example, if A's actual output is within the predicted commitment range, the cloud platform will pay compensation based on the output, such as a reward of 0.05 yuan per kWh of output]. If A's actual output is 45MW, the compensation will be 45,000kWh × 0.05 yuan / kWh = 2,250 yuan. If the actual value exceeds the predicted commitment range, a deviation penalty will be paid. [The deviation penalty is a fine levied on a participant for inaccurate predictions. For example, if A's actual output is 35MW, lower than the lower limit of the predicted commitment range of 40MW, a deviation penalty will be paid, such as 0.1 yuan per kWh deviation. If the deviation is 40MW - 35MW = 5MW, the penalty will be 5,000kWh × 0.1 yuan / kWh = 500 yuan.] The penalty will be injected into the reserve resource pool [a fund pool used to smooth system fluctuations, consisting of the deviation penalty and other income]. For example, the 500 yuan penalty paid by A is injected into the reserve resource pool, which can be used to purchase energy storage services (such as paying energy storage operator C to provide 50kW of emergency compensation power) or to subsidize affected participants (such as subsidizing load aggregator B for increased electricity costs due to a sudden load surge). This serves as the incentive mechanism for the game framework; participants submit initial prediction commitment intervals as input to the game framework, constructing a "prediction-scheduling" game framework.

[0090] Specifically, taking a smart substation as an example, the main participants in the game theory framework include new energy operator A (managing a 50MW photovoltaic power station) and load aggregator B (representing an industrial park). During initialization, A and B register accounts on the cloud platform. A submits the installed capacity, historical output data, and geographical location information of the photovoltaic power station, while B submits the historical load data and load characteristic information of the industrial park. The cloud platform allocates data interfaces to A and B for uploading subsequent prediction data and receiving reward / penalty results.

[0091] Based on a game theory framework, environmental data is hierarchically decomposed to extract disturbance features at different time and spatial scales. The environmental data (acquired through sensors, monitoring equipment, and external data interfaces) includes wind speed (collected in real-time by wind speed sensors installed at wind farms, uploaded to the cloud platform at a rate of seconds), solar irradiance (collected in real-time by solar irradiance sensors installed at photovoltaic power stations, uploaded to the cloud platform at a rate of minutes), and load demand (collected from real-time load data of industrial and residential users through smart meters and load monitoring equipment, uploaded to the cloud platform at a rate of minutes). Disturbance features include minute-level abrupt changes (referring to drastic changes within a short period (minutes). For example, a sudden increase in wind speed from 5 m / s to 15 m / s within one minute represents a minute-level abrupt change), hourly-level trend features (referring to stable changes over a longer period (hours). For example, a continuous decrease in solar irradiance from 800 W / m² to 600 W / m² within one hour represents an hourly trend), and localized abrupt changes (referring to drastic changes in a specific local area). For example, a 15% decrease in solar irradiance in a photovoltaic panel area while other areas remain stable indicates a localized abrupt change. [This refers to] regional trend characteristics [referring to smooth changes across the entire region. For example, a decrease in wind speed from 10 m / s to 8 m / s across the entire substation area within one hour indicates a regional trend characteristic]. These serve as the basis for predicting the reshaping of the commitment interval.

[0092] Hierarchical decomposition refers to breaking down environmental data into multiple levels according to temporal and spatial scales in order to extract different features. Specific methods include:

[0093] For time-scale decomposition: wavelet transform can be used to decompose environmental data into components at different time scales. For example, wavelet transform can be applied to wind speed data to decompose it into minute-level components (capturing abrupt changes), hour-level components (capturing trends), and day-level components (capturing long-term changes).

[0094] Corresponding spatial scale decomposition: Spatial interpolation methods (such as Kriging interpolation) can be used to decompose environmental data into components at different spatial scales. For example, spatial interpolation of light intensity data can decompose it into local region components (capturing changes in a specific photovoltaic panel area) and regional components (capturing changes in the entire substation area).

[0095] The perturbation features are extracted through statistical analysis and feature detection methods. The specific steps are as follows:

[0096] Statistical analysis: Calculating statistical indicators for environmental data, such as mean, variance, and rate of change. For example, calculating the minute-level rate of change of wind speed; if the rate of change exceeds 10%, it is extracted as a minute-level abrupt change feature.

[0097] Feature detection: Threshold detection or pattern recognition methods are used to identify abnormal patterns in environmental data. For example, if the light intensity decreases by 15% in a local area, it is extracted as a local mutation feature.

[0098] Of course, other methods can be used to achieve the purpose of this technology.

[0099] Disturbance features quantify environmental uncertainties and guide the dynamic adjustment of the forecast commitment range. For example, if minute-level abrupt changes are detected (such as a sudden increase in wind speed), it indicates severe environmental fluctuations and high forecasting difficulty, so the forecast commitment range should be widened (e.g., from ±10% to ±20%) to avoid unfair penalties; if hourly-level trends are detected (such as a steady decrease in light intensity), it indicates relatively small environmental fluctuations and low forecasting difficulty, so the forecast commitment range should be tightened (e.g., from ±10% to ±5%) to improve forecast accuracy requirements.

[0100] For perturbation features of different standards, minute-level mutation features have higher priority than hour-level trend features, local mutation features have higher priority than regional trend features, and for perturbation features of the same standard, priority is ranked according to the degree of impact of the perturbation features on the prediction accuracy.

[0101] The degree of impact can be quantified using the following indicators:

[0102] Fluctuation amplitude: The greater the amplitude of the disturbance feature, the greater the impact on prediction accuracy. For example, a minute-level abrupt change in wind speed of 10 m / s has a greater impact than an hourly trend feature of wind speed decrease of 2 m / s.

[0103] Duration: The shorter the duration of the perturbation feature, the greater its impact on prediction accuracy. For example, minute-level mutation features have short durations and are difficult to predict, therefore they have a higher priority than hourly-level trend features.

[0104] Spatial range: The smaller the range of influence of a disturbance feature, the greater its impact on prediction accuracy. For example, local abrupt changes (such as a 15% decrease in sunlight in a photovoltaic panel area) are more difficult to predict than regional trend features (such as a 5% decrease in sunlight across the entire region), and therefore have a higher priority.

[0105] A weighted scoring method can be used to calculate the degree of influence score by comprehensively considering the fluctuation amplitude, duration, and spatial range. For example, the score for minute-level mutation features = fluctuation amplitude weight (0.5) × amplitude score (10) + duration weight (0.3) × time score (8) + spatial range weight (0.2) × range score (6) = 8.3; the score for hourly-level trend features = 4.5. The higher the score, the higher the priority.

[0106] Based on the priority sequence and the strength of the perturbation features, the leniency of the initial prediction commitment interval is dynamically adjusted. That is, if a high-priority minute-level mutation feature is detected, the tolerance range of the prediction commitment interval is significantly widened and the penalty intensity is reduced [the penalty amount can be reduced by adjusting the calculation coefficient of the penalty amount. For example, if the initial penalty coefficient is 0.1 yuan / kWh, it is significantly reduced to 0.05 yuan / kWh if a minute-level mutation feature is detected; if an hour-level trend feature is detected, it is moderately reduced to 0.08 yuan / kWh]. If only a medium-priority hour-level trend feature is detected, the prediction commitment interval is moderately widened, generating an adjusted prediction commitment interval, which serves as the update input for the game framework.

[0107] Based on the adjusted forecast commitment range [referring to the final forecast result submitted to the cloud platform by the participating parties based on the adjusted forecast commitment range. For example, new energy operator A, based on the adjusted forecast commitment range "the probability of output ≥ 35MW is 90%", submits a forecast commitment of "output of 40MW in the next 15 minutes, with a probability of 90%"]., the new energy operator and load aggregator submit forecast commitments to the cloud platform, and implement reward and punishment rules based on the actual deviation [referring to the difference between the forecast value and the actual value. For example, if A's forecast commitment is "output of 40MW", and the actual output is 45MW, then the actual deviation is 45MW - 40MW = 5MW (positive deviation); if the actual output is 30MW, then the actual deviation is 30MW - 40MW = -10MW (negative deviation).]. For example, if A's actual output is 45MW, within the predicted commitment range, it will receive dispatch priority (priority for output absorption) and economic compensation (2,250 yuan); if the actual output is 30MW, exceeding the predicted commitment range, it will pay a deviation penalty (deviation amount 10MW × 0.05 yuan / kWh = 500 yuan). Prediction deviation data is generated as the basis for correcting the generated three-dimensional uncertainty cloud map. [By quantifying the prediction error, the probability distribution of the three-dimensional uncertainty cloud map is corrected. For example, if A's prediction deviation is -10MW, indicating that the predicted value is too high, the probability of the power distribution will be adjusted towards lower power when generating the cloud map (e.g., correcting "40MW probability is 90%" to "35MW probability is 90%)", thereby improving the accuracy of the cloud map.]

[0108] This method integrates prediction bias data, regional meteorological data [referring to meteorological information covering the substation area, including wind speed, solar intensity, and temperature. For example, wind speed distribution (e.g., 10 m / s in one area, 8 m / s in neighboring areas) and solar intensity distribution (e.g., 800 W / m² in one area, 600 W / m² in neighboring areas)], and output correlation data from adjacent substations [referring to the mutual influence between the outputs of neighboring substations. For example, if the photovoltaic output of substation A decreases by 10%, historical data analysis suggests that the wind power output of neighboring substation B may decrease by 8%, with a correlation coefficient of 0.8. This correlation is calculated using a correlation coefficient matrix]. This generates a model with spatiotemporal coupling characteristics [referring to the interrelationship between the time and spatial dimensions in a three-dimensional uncertainty cloud map]. For example, the output of a photovoltaic power station is represented in the time dimension as "output will decrease by 10% in the next 15 minutes," and in the spatial dimension as "the decrease is mainly concentrated in a local area." The spatiotemporal coupling characteristic is represented as "the probability of a 10% decrease in output in a local area in the next 15 minutes is 80%." The three-dimensional uncertainty cloud map is shown, where the three-dimensional uncertainty cloud map includes the distribution of time, power, and probability of occurrence.

[0109] By employing a game-theoretic framework with reward and penalty rules, the prediction accuracy problem is transformed into an economic incentive problem, motivating new energy operators and load aggregators to proactively improve prediction accuracy. Through hierarchical perception and interval reshaping logic of environmental disturbances, the prediction commitment interval is dynamically adjusted based on the priority and intensity of disturbance characteristics, improving the fairness and adaptability of predictions. By generating a three-dimensional uncertainty cloud map with spatiotemporal coupling characteristics, replacing traditional point predictions, it provides distribution information on time, power, and probability of occurrence, offering a more comprehensive decision-making basis for the scheduling and allocation layer. Through a feedback correction mechanism for prediction deviation data, a closed-loop optimization of prediction-scheduling-execution is formed, improving the overall prediction accuracy and stability of the system.

[0110] Due to the volatility of renewable energy output and the randomness of load demand, various resources within a substation (such as renewable energy, energy storage, loads, and grid interfaces) face conflicting objectives during dispatching. For example, renewable energy's pursuit of maximizing absorption may lead to over-discharge of energy storage devices, affecting their lifespan; adjustable loads' pursuit of cost minimization may conflict with grid stability requirements. Traditional centralized optimization dispatching methods suffer from high computational complexity and struggle to respond quickly to real-time fluctuations. However, distributed autonomous decision-making and multi-stage negotiation mechanisms resolve these policy conflicts, improving dispatching efficiency and system stability.

[0111] At the dispatch and allocation layer, substation resources are divided into: New Energy Clusters [referring to the collection of all new energy power generation equipment within a substation, such as photovoltaic power plants and wind farms. These devices are connected to the grid via inverters and can regulate active and reactive power output. For example, a substation may contain 10 photovoltaic power generation units and 5 wind turbine units, which together form a new energy cluster], Energy Storage Alliances [referring to the collection of all energy storage devices within a substation, such as lithium battery banks and supercapacitors. These devices are connected to the grid via power conversion systems (PCS) and can perform charging and discharging operations. For example, a substation may contain 3 large lithium battery banks and 2 supercapacitors, which together form an energy storage alliance], and Adjustable Load Groups [referring to the collection of all adjustable loads within a substation, such as interruptible loads from industrial users and smart home appliances from residential users. These loads can adjust electricity consumption through demand response mechanisms. For example, a substation may contain interruptible production lines from 5 industrial users and smart air conditioners for 1000 residential households, which together form an adjustable load group]. The system comprises four types of autonomous decision-making entities: [1] Grid interface agent [refers to the interface device or virtual agent between the substation and the upper-level power grid, responsible for executing the upper-level power grid's dispatch instructions, such as power purchase or sales plans. The goal of the grid interface agent is to ensure that the substation's power exchange meets the requirements of the upper-level power grid. For example, a substation may be connected to the upper-level power grid through a transformer, and the grid interface agent is responsible for managing the power flow between the substation and the upper-level power grid]; [2] Entities with independent decision-making capabilities during the dispatching process. Each entity formulates its own strategy based on its own goals and constraints, and negotiates with other entities to achieve global optimization. For example, a new energy cluster autonomously decides its output plan, and an energy storage alliance autonomously decides its charging and discharging strategy. The edge layer solves for the Nash equilibrium point based on a three-dimensional uncertainty cloud map to generate a dispatching scheme; it performs dynamic electrical partitioning based on the feeder voltage-power sensitivity matrix, obtains the partitioning results, and uses the dispatching scheme and partitioning results as inputs to the execution control layer.

[0112] Methods for generating scheduling schemes based on solving the Nash equilibrium point using three-dimensional uncertainty contour maps include:

[0113] Substation resources are divided into four categories of autonomous decision-making entities: new energy clusters, energy storage alliances, adjustable load groups, and grid interface agents. The goal of the new energy clusters is to maximize power consumption [meaning the new energy clusters output as much power as possible to reduce wind and solar curtailment. For example, the goal of a photovoltaic power station is to generate as much power as possible during periods of sufficient sunlight, reaching over 90% of its installed capacity]. The goal of the energy storage alliances is to minimize lifespan losses. The goal of the adjustable load groups is to minimize electricity costs. The goal of the grid interface agents is to meet the instructions of higher-level dispatching authorities, serving as the basis for policy conflict negotiation.

[0114] Conflict features are extracted from the objective functions of each autonomous decision-making entity to generate a conflict feature set. These conflict features include objective priority features (referring to the importance of each entity's objective, used to measure the relative importance between objectives. For example, the priority of renewable energy cluster absorption may be higher than the priority of energy storage lifespan degradation, because renewable energy absorption is the core objective of system operation) and constraint boundary features (referring to the restrictions that each entity must meet to achieve its objectives. For example, the SOC protection boundary for energy storage means that the SOC must not be lower than 20% or higher than 80% to protect battery life; the load reduction capability boundary for adjustable load means that the load reduction amount must not exceed its adjustable range, for example, industrial users can reduce their electricity consumption by a maximum of 30%). Objective priority features include the priority of renewable energy cluster absorption and the priority of energy storage lifespan degradation. Constraint boundary features include the SOC protection boundary for energy storage and the load reduction capability boundary for adjustable load, serving as the basis for hierarchical negotiation.

[0115] The objective function is a mathematical expression describing the optimization goals of each autonomous decision-making entity, used to quantify their decision-making objectives. Conflict feature extraction refers to analyzing the contradictions between the objective functions and extracting the key features that lead to the conflict. For example, by comparing the output plans of the new energy cluster and the charging and discharging plans of the energy storage alliance, the conflict feature of "increased new energy output leading to a decrease in energy storage SOC" can be extracted.

[0116] The conflict feature set is decomposed into multiple conflict levels based on the severity and scope of the conflict. The conflict levels include the initial conflict level and the severe conflict level. The initial conflict level corresponds to the conflict where the SOC is close to exceeding the limit, and the severe conflict level corresponds to the conflict where the SOC actually exceeds the limit. These levels serve as the input for constraint convergence.

[0117] The decomposition is based on the severity and scope of the conflict. Severity refers to the degree of threat the conflict poses to system stability; for example, a conflict with SOC approaching its limit poses a relatively small threat, while a conflict with SOC actually exceeding its limit poses a larger threat. Scope of impact refers to the number of entities or the geographical area involved in the conflict; for example, a conflict involving only new energy sources and energy storage has a smaller scope of impact, while a conflict involving all entities has a larger scope of impact.

[0118] Conflict levels: Conflict levels include initial conflict level and severe conflict level.

[0119] Initial conflict level: This corresponds to a conflict where the SOC is close to exceeding the limit. For example, the SOC of energy storage drops to 25%, which is close to the protection boundary of 20%, but has not yet been breached.

[0120] Severe Conflict Level: This level corresponds to a conflict where the actual SOC exceeds the limit. For example, if the SOC of the energy storage drops to 18%, it has exceeded the protection boundary of 20%, which may lead to battery damage.

[0121] As input to constraint convergence: the conflict level serves as input to constraint convergence, guiding subsequent negotiation mechanisms. For example, initial conflict levels can be resolved with minor adjustments to the strategy, while severe conflict levels require significant adjustments to the strategy or even relaxation of constraint boundaries.

[0122] Based on the three-dimensional uncertainty cloud map, a multi-stage negotiation mechanism is used to resolve policy conflicts at the conflict level. In the initial conflict level, the new energy cluster adjusts its output plan to meet the SOC protection boundary of energy storage. In the severe conflict level, the energy storage alliance relaxes the SOC protection boundary. The negotiation process adopts an iterative approach, with each iteration narrowing the feasible domain of the strategy and generating a negotiated strategy set as input for solving the Nash equilibrium.

[0123] The edge layer, based on the negotiated strategy set [the adjusted strategy set of each entity, such as "185kW of new energy output, 45kW of energy storage discharge, 8kW of adjustable load reduction, and 35kW of grid power purchase"] and a three-dimensional uncertainty cloud map, solves for the Nash equilibrium point and generates a scheduling scheme. The scheduling scheme includes the active / reactive power adjustment range, charging and discharging strategies, load transfer plans, and power purchase and sale plans of each entity.

[0124] For example, suppose a renewable energy cluster at a substation plans to output 200kW of power within the next 15 minutes. An energy storage consortium plans to discharge 50kW to support renewable energy consumption, an adjustable load group plans to reduce its load by 10kW to lower costs, and a grid interface agent plans to purchase 30kW of electricity to meet superior instructions. In conflict feature extraction, the following conflict features are identified: "Renewable energy output of 200kW causes the energy storage SOC to drop to 25%, approaching the protection boundary of 20%," and "The adjustable load reduction of 10kW causes the grid purchase of electricity to increase to 40kW, exceeding the plan." These features constitute the conflict feature set. In conflict hierarchy decomposition, "SOC dropping to 25%" is classified as an initial conflict level because it has not yet exceeded the protection boundary; "Grid purchase of electricity increasing to 40kW" is classified as a severe conflict level because it has exceeded the planned range. These conflict levels serve as inputs for constraint convergence, guiding subsequent negotiations.

[0125] Based on the 3D uncertainty cloud map, the new energy cluster found an 80% probability of outputting 200kW, but this would cause the energy storage SOC to drop to 25%, falling into the initial conflict level. Therefore, in the first iteration, the new energy cluster reduced its output to 190kW, restoring the energy storage SOC to 22%, still close to the protection boundary. In the second iteration, the new energy cluster further reduced its output to 185kW, restoring the energy storage SOC to 20%, meeting the protection boundary. Simultaneously, the adjustable load group found that reducing output by 10kW would lead to an increase in grid power purchases to 40kW, falling into the severe conflict level. In the first iteration, the adjustable load group reduced the reduction to 8kW, decreasing grid power purchases to 38kW; in the second iteration, the energy storage alliance relaxed the SOC protection boundary to 15%, supporting a new energy output of 185kW, ultimately reducing grid power purchases to 35kW, meeting the planned range. After multi-stage negotiation, a negotiated strategy set of "new energy output 185kW, energy storage discharge 45kW, adjustable load reduction 8kW, grid power purchase 35kW" was generated as input for the Nash equilibrium solution.

[0126] Methods for obtaining zoning results based on feeder voltage-power sensitivity matrix dynamic electrical zoning include:

[0127] Substation status data is decomposed into multiple dimensions to extract change characteristics. This data includes voltage, power, and power flow distribution. These change characteristics serve as the basis for disturbance sensing and include instantaneous change characteristics (referring to the abrupt changes in data over a short period), trend change characteristics (referring to the changing trends of data over a longer period), and spatial distribution characteristics (referring to the spatial distribution differences of data).

[0128] Multidimensional decomposition refers to decomposing substation status data in three dimensions: time, space, and physical quantity, in order to extract change features at different scales.

[0129] Time dimension decomposition: Data is decomposed into components of different time scales, such as minutes and hours, through time series analysis methods (such as wavelet transform and Fourier transform).

[0130] Spatial dimensionality decomposition: Data is decomposed into components of different spatial scales, such as local areas or the entire substation area, through spatial interpolation methods (such as Kriging interpolation) or network topology analysis.

[0131] Physical quantity dimensional decomposition: Data is decomposed into components of different physical quantities, such as voltage, power, and power flow, through statistical analysis methods (such as principal component analysis).

[0132] Power flow distribution refers to the power flow distribution among feeders and nodes within a substation, including the direction and magnitude of active and reactive power. For example, a substation has three feeders: Feeder 1 has an active power flow of 50MW and a reactive power flow of 20Mvar; Feeder 2 has an active power flow of 30MW and a reactive power flow of 10Mvar; Feeder 3 has an active power flow of -10MW (in reverse) and a reactive power flow of 5Mvar.

[0133] Based on the three-dimensional uncertainty cloud map, the change characteristics are comprehensively evaluated to generate the disturbance index of the substation status. The comprehensive evaluation is based on the intensity and duration of the change characteristics. If the instantaneous change characteristics are strong and short in duration, the disturbance index is high. If the trend change characteristics are weak and long in duration, the disturbance index is low. This serves as the input for frequency optimization.

[0134] Comprehensive assessment refers to the use of weighted fusion methods to integrate the intensity and duration of changing characteristics into a unified disturbance index, which is used to quantify the degree of disturbance to the system state. Specific methods include:

[0135] Step B1: Calculate the intensity values ​​for instantaneous change characteristics, trend change characteristics, and spatial distribution characteristics respectively. For example, the intensity of instantaneous change characteristics can be calculated using the absolute value of the rate of change (e.g., voltage change rate -2% / minute, intensity is 2%); the intensity of trend change characteristics can be calculated using the total change (e.g., power change +50kW / hour, intensity is 50kW); and the intensity of spatial distribution characteristics can be calculated using the variance (e.g., power distribution variance is 1000kW², intensity is 1000).

[0136] Step B2: Normalize the duration of each feature. For example, the duration of an instantaneous change feature is 1 minute, and it is normalized to 0.1 (based on 10 minutes); the duration of a trend change feature is 1 hour, and it is normalized to 1.

[0137] Step B3: Determine the weights of each feature based on the probability distribution provided by the 3D uncertainty cloud map. For example, if the cloud map shows an 80% probability of power fluctuation in the next 15 minutes, then the instantaneous change feature has a higher weight (e.g., 0.6), the trend change feature has a lower weight (e.g., 0.3), and the spatial distribution feature has a moderate weight (e.g., 0.1). The disturbance index is calculated as follows: Disturbance Index = Weight 1 × Intensity 1 × (1 - Duration 1) + Weight 2 × Intensity 2 × Duration 2 + Weight 3 × Intensity 3, where Weight 1, Weight 2, and Weight 3 correspond to the instantaneous, trend, and spatial distribution features, respectively.

[0138] Suppose that the voltage of feeder A in a substation suddenly drops from 10kV to 9.8kV within 1 minute (instantaneous change characteristics: intensity 2%, duration 0.1), and the power of feeder B increases from 500kW to 550kW within 1 hour (trend change characteristics: intensity 50kW, duration 1). The variance of the power distribution of feeders A and B increases from 500kW² to 1000kW² (spatial distribution characteristics: intensity 1000, duration 0.5). Based on the three-dimensional uncertainty cloud map, the probability of power fluctuation in the next 15 minutes is 80%, and the weights are determined as instantaneous 0.6, trend 0.3, and spatial 0.1. The disturbance index is calculated as follows: Disturbance index = 0.6 × 2 × (1 - 0.1) + 0.3 × 50 × 1 + 0.1 × 1000 × 0.5 = 1.08 + 15 + 50 = 66.08. The disturbance index of 66.08 is used as the input for frequency optimization.

[0139] Based on the disturbance index, the update frequency of the feeder voltage-power sensitivity matrix is ​​dynamically adjusted. If the disturbance index is higher than a first threshold, the update frequency is increased to the first frequency value; if the disturbance index is lower than a second threshold, the update frequency is decreased to the second frequency value, generating an optimized update frequency as the basis for dynamic electrical zoning. The first threshold represents a high disturbance state, and the second threshold represents a low disturbance state. If the disturbance index is between the first and second thresholds, the current frequency is maintained (e.g., updated every 5 minutes).

[0140] Assuming a disturbance index of 66.08, a first threshold of 80, and a second threshold of 30, the system state is classified as having a moderate disturbance since 66.08 falls between 30 and 80, and the update frequency remains at once every 5 minutes. If a subsequent voltage drop intensifies, causing the disturbance index to rise to 85 (above the first threshold of 80), the system state is classified as having a high disturbance, and the update frequency is increased to once every 1 minute. If the voltage stabilizes and the disturbance index drops to 25 (below the second threshold of 30), the system state is classified as having a low disturbance, and the update frequency is reduced to once every 10 minutes. The optimized update frequency (e.g., once every 1 minute) serves as the basis for dynamic electrical zoning.

[0141] Based on the optimized update frequency and scheduling scheme, the feeder voltage-power sensitivity matrix is ​​calculated, and dynamic electrical partitioning is generated. The dynamic electrical partitioning includes a strongly coupled region and a weakly coupled region. The strongly coupled region performs centralized optimization scheduling, while the weakly coupled region performs autonomous decision scheduling, and the partitioning results are generated.

[0142] The feeder voltage-power sensitivity matrix is ​​used to quantify the impact of power variations in each feeder on node voltage. Specific calculation methods include:

[0143] Step C1: Based on the power flow equations (P = V²G - VYcosθ, Q = -V²B + VYsinθ, where P is active power, Q is reactive power, V is voltage, G is conductance, B is susceptance, Y is admittance, and θ is phase angle), calculate the sensitivity matrix using a linearization method. For example, the voltage sensitivity to power can be expressed as ∂V / ∂P and ∂V / ∂Q, typically calculated using the inverse of the Jacobian matrix.

[0144] Step C2: For a substation with n nodes and m feeders, the sensitivity matrix is ​​an n×m matrix, where each element represents the voltage change at each node when the power of a feeder changes by 1 unit. For example, a substation with 3 nodes (N1, N2, N3) and 2 feeders (F1, F2) has the following calculated sensitivity matrix:

[0145] [∂V1 / ∂P1 ∂V1 / ∂P2]

[0146] [∂V2 / ∂P1 ∂V2 / ∂P2]

[0147] [∂V3 / ∂P1 ∂V3 / ∂P2]

[0148] Assume the calculation result is: [0.05 0.01] [0.02 0.04] [0.01 0.03]

[0152] This means that when the power of feeder F1 increases by 1MW, the voltage at node N1 increases by 0.05kV, the voltage at node N2 increases by 0.02kV, and the voltage at node N3 increases by 0.01kV.

[0153] When the partition changes, the scheduling rules are adjusted through the continuous optimization function to generate the adjusted partition result, which serves as the update input for the linkage control of the execution control layer.

[0154] By dividing substation resources into four types of autonomous decision-making entities—new energy clusters, energy storage alliances, adjustable load groups, and grid interface agents—this approach breaks away from the traditional centralized optimization paradigm, achieving distributed autonomous decision-making and reducing computational complexity. Through conflict feature extraction and conflict hierarchy decomposition, complex strategy conflict problems are processed in layers, improving the targeting and efficiency of conflict resolution. A multi-stage negotiation mechanism and iterative convergence method dynamically adjust the strategies of each entity, ensuring the global optimality of the scheduling scheme. Based on the probability distribution information of the three-dimensional uncertainty cloud map, strategy conflict negotiation and Nash equilibrium solving are guided, improving the accuracy and adaptability of the scheduling scheme.

[0155] By employing multi-dimensional decomposition and comprehensive evaluation, the degree of disturbance to the system state is accurately perceived, avoiding the blindness of fixed-frequency updates. Dynamic adjustment of the update frequency optimizes the efficiency of computing resource utilization. Continuous optimization functions smooth out scheduling rule adjustments during partition changes, enhancing system stability.

[0156] At the execution control layer, programmable impedance characteristics are embedded in the photovoltaic inverter and energy storage power conversion system to enable autonomous equipment response based on the scheduling scheme; the source-load fluctuation event chain model performs cross-time scale linkage control based on the partitioning results; and the execution results are fed back to the integration layer.

[0157] Methods for autonomous equipment response based on scheduling schemes, which involve embedding programmable impedance characteristics into photovoltaic inverters and energy storage power conversion systems, include:

[0158] The equipment's operational data is decomposed into multiple levels to extract impact features. The operational data includes the operating status of the photovoltaic inverter and the energy storage power conversion system, while the impact features include geographical impact features, power impact features, and sensitivity impact features, which serve as the basis for equipment impact assessment.

[0159] Based on the partitioning results, the impact features are prioritized and evaluated to generate a comprehensive importance index for the equipment. The evaluation criteria are the contribution of the impact features to the system stability. If the geographical impact features of the equipment show that it is located in a strong coupling area and the sensitivity impact features show that it has a large impact on voltage, then the comprehensive importance index is high and it is used as the input for priority adjustment.

[0160] Based on the comprehensive importance index, the impedance parameter adjustment priority of the equipment is dynamically sorted. Among them, the equipment with a high comprehensive importance index is adjusted first, while the equipment with a low comprehensive importance index is adjusted later. This generates an equipment adjustment sequence as the basis for the equipment's autonomous response.

[0161] Based on the scheduling scheme and the equipment adjustment sequence, programmable impedance characteristics are embedded in the photovoltaic inverter and energy storage power conversion system to execute autonomous equipment response. The autonomous response includes the simulated synchronous machine inertial response of the new energy equipment and the short-circuit capacity support provided by the energy storage equipment, generating equipment response results as inputs for the linkage control of the execution control layer.

[0162] When the system state changes, the comprehensive importance index is re-evaluated through a continuous update function, and the equipment adjustment sequence is adjusted to generate an updated equipment adjustment sequence, which serves as the update input for the linkage control of the execution control layer.

[0163] For example, the substation contains 10 photovoltaic inverters (numbered PV1-PV10) and 5 energy storage power conversion systems (numbered ESS1-ESS5). Operating data includes the actual output of the photovoltaic inverters (unit: kW), the state of charge (SOC, unit: %) of the energy storage power conversion systems, and their output power (unit: kW).

[0164] Influencing features are extracted using a multi-level decomposition method.

[0165] Geographical impact characteristics: Record the distance between each device and the substation load center. For example, PV1 is 100 meters away from the load center, and ESS1 is 50 meters away from the load center.

[0166] Power impact characteristics: Calculate the rated power and actual output of each device. For example, PV1 has a rated power of 500kW and an actual output of 450kW; ESS1 has a rated power of 300kW and an actual output of 200kW.

[0167] Sensitivity impact characteristics: Calculate the sensitivity of each device to the substation bus voltage (unit: V / kW). For example, PV1 has a sensitivity of 0.05V / kW to the bus voltage, and ESS1 has a sensitivity of 0.08V / kW.

[0168] Priority assessment based on partitioning results: Assuming the dispatching and allocation layer has generated partitioning results, the substation is divided into a strongly coupled zone (including bus 1 and feeders 1-3) and a weakly coupled zone (including bus 2 and feeders 4-6). PV1, PV2, and ESS1 are located in the strongly coupled zone, while the remaining equipment is located in the weakly coupled zone. Priority assessment of impact characteristics is then performed based on the partitioning results:

[0169] Evaluation criteria: Geographical impact characteristics indicate that the device is located in a strong coupling area, thus contributing highly; sensitivity impact characteristics indicate that the device has a significant impact on voltage, thus contributing highly; power impact characteristics indicate that the device has a large rated power, thus contributing highly.

[0170] The overall importance index is calculated using a weighted summation method, with weights of 0.4 for geographical influence, 0.4 for sensitivity influence, and 0.2 for power influence. For example, the overall importance index of ESS1 is 0.4×(1 / 50)+0.4×0.08+0.2×(300 / 500)=0.16; the index of PV1 is 0.4×(1 / 100)+0.4×0.05+0.2×(500 / 500)=0.11.

[0171] The priority of equipment impedance parameter adjustment is dynamically ranked according to the comprehensive importance index: ESS1 (0.16) > PV1 (0.11) > PV2 (0.10) > ... > PV10 (0.05). Equipment with a high comprehensive importance index (such as ESS1) has its impedance parameters adjusted first, while equipment with a low comprehensive importance index (such as PV10) has its adjustment delayed.

[0172] Based on the autonomous response of the equipment according to the scheduling scheme, assuming the scheduling scheme requires an increase of 100kW in photovoltaic output and 50kW in energy storage discharge, the impedance parameters of ESS1 are adjusted first according to the equipment adjustment sequence:

[0173] Implanted programmable impedance characteristics: By using a power electronic switch array, the virtual impedance of ESS1 is adjusted to 0.1 + j0.05 ohms, which provides short-circuit capacity support and increases the discharge power by 50kW.

[0174] New energy equipment response: Adjusting the virtual impedance of PV1 to 0.15 + j0.08 ohms simulates the inertial response of the synchronous machine, increasing the output by 60kW; adjusting the impedance of PV2 increases the output by 40kW.

[0175] Equipment response results: Generate equipment response results, including ESS1 discharge of 50kW, PV1 output increase of 60kW, and PV2 output increase of 40kW, as inputs for linkage control.

[0176] When the system state changes, the sequence is updated, assuming the load center migrates to bus 2 and the weakly coupled area becomes a strongly coupled area. The comprehensive importance index is re-evaluated using a continuous update function (e.g., a linear interpolation function). For example, if the weight of the geographical impact feature of PV10 increases, the index rises from 0.05 to 0.12, and the adjustment sequence is updated to ESS1>PV10>PV1>…, generating an updated equipment adjustment sequence as the update input for the linkage control.

[0177] The source-load fluctuation event chain model includes methods for cross-timescale linkage control based on partitioning results, such as:

[0178] The indicators involved in the event chain are decomposed to extract the impact features. The indicators include photovoltaic output, load demand and voltage level. The impact features include instantaneous impact features, trend impact features and coupling impact features, which serve as the basis for correlation and integration.

[0179] Based on the partitioning results, dynamic weights are assigned to the influencing features to generate a weight sequence. The assignment is based on the degree of influence of the influencing features on the stability of the substation. If the instantaneous influencing features have a large impact on the stability of the substation, they are assigned high weights. If the trend influencing features have a small impact, they are assigned low weights. The weights are dynamically adjusted according to the strong coupling and weak coupling regions of the partitioning results and serve as inputs for conditional optimization.

[0180] By using a weighted fusion function to correlate and fuse the influencing features and weight sequences, composite triggering conditions are generated, and a comprehensive influence index is generated, which serves as the basis for coordinated control.

[0181] Based on the equipment response results and composite triggering conditions, cross-timescale linkage control is executed. The linkage control includes millisecond-level energy storage compensation, second-level load shedding and minute-level grid support, generating linkage control results as input to the execution results of the integration layer.

[0182] The trigger threshold of the composite triggering condition is dynamically optimized based on the disturbance index of the substation status. If the disturbance index of the substation status is high, the trigger threshold is lowered; if the disturbance index of the substation status is low, the trigger threshold is raised. The optimized composite triggering condition is generated and used as the update input for the execution result of the integration layer.

[0183] At the integration layer, the cloud platform performs game arbitration and market settlement based on the execution results. The edge layer uses quantum-inspired algorithms to accelerate the search for equilibrium points and updates the scheduling scheme based on the execution results. The terminal layer dynamically reconstructs the impedance characteristics of the equipment through power electronic switch arrays and adjusts the operating status of the equipment based on the updated scheduling scheme.

[0184] Assume the event chain involves a decrease in photovoltaic output, a sudden increase in load demand, and a voltage drop. Extract the impact characteristics:

[0185] Instantaneous impact characteristics: Photovoltaic output decreases by 10% within 1 minute (e.g., from 500kW to 450kW).

[0186] Trend impact characteristics: Load demand increases by 5% continuously within 1 hour (e.g., from 1000kW to 1050kW).

[0187] Coupling effect characteristics: The correlation coefficient between the decrease in photovoltaic output and the decrease in voltage is 0.8.

[0188] Dynamic weight allocation based on partitioning results

[0189] Assume the partitioning results show significant voltage fluctuations in the strongly coupled region. Based on the partitioning results, dynamically allocate weights:

[0190] Allocation criteria: Instantaneous impact features have a significant impact on system stability, with a weight of 0.6; trend impact features have a relatively small impact, with a weight of 0.3; coupling impact features have a weight of 0.1.

[0191] Strong coupling region adjustment: The weight of the strong coupling region is increased by 10%, for example, the weight of the instantaneous influence feature is adjusted to 0.66.

[0192] Generate a weight sequence: The weight sequence is [0.66, 0.27, 0.07], which serves as the input for conditional optimization.

[0193] Related fusion generates composite triggering conditions

[0194] Generate a comprehensive impact index using a weighted fusion function (e.g., weighted summation):

[0195] Calculation formula: Comprehensive impact index = 0.66 × (proportion of decrease in photovoltaic output) + 0.27 × (proportion of increase in load demand) + 0.07 × (correlation coefficient).

[0196] Example: If photovoltaic output decreases by 10% and load demand increases by 5%, the correlation coefficient is 0.8, and the comprehensive impact index is 0.66×0.1+0.27×0.05+0.07×0.8=0.125.

[0197] Composite triggering condition: If the comprehensive impact index is greater than 0.1, then the linkage control is triggered.

[0198] The linkage control is executed based on the equipment response results. Specifically, it is based on the equipment response results (ESS1 discharges 50kW, PV1 output increases by 60kW) and the composite triggering condition (exponent 0.125 > 0.1).

[0199] Millisecond-level energy storage compensation: ESS1 discharges 50kW within 10 milliseconds to compensate for the decrease in photovoltaic output.

[0200] Second-level load shedding: Shedding 10% of non-critical loads (e.g., 100kW) in a strongly coupled area within 5 seconds.

[0201] Minute-level grid support: Request 50kW support from the upper-level grid within 1 minute.

[0202] Linkage control results: Generate linkage control results, including 50kW energy storage compensation, 100kW load shedding, and 50kW grid support, as inputs to the integration layer execution results.

[0203] The trigger threshold is optimized based on the disturbance index. Assuming the disturbance index of the substation is 0.8 (above the threshold of 0.6), the trigger threshold is lowered to 0.08; if the disturbance index drops to 0.4 (below the threshold of 0.5), the trigger threshold is increased to 0.12. The optimized composite trigger condition is generated and used as the update input for the integration layer execution result.

[0204] The cloud platform's methods for game arbitration and market settlement based on execution results include:

[0205] The substation operation scenarios are decomposed in multiple dimensions to extract impact characteristics. The operation scenarios include voltage collapse risk, insufficient renewable energy absorption and load change. The impact characteristics include disturbance impact characteristics, range impact characteristics and economic impact characteristics, which serve as the basis for scenario impact assessment.

[0206] Based on the partitioning results, the impact features are prioritized and a comprehensive emergency index for the scenario is generated. The evaluation criteria are the degree of impact of the impact features on the substation operation. If the disturbance impact features show a high risk of voltage collapse and the range impact features show a strong coupling area, the emergency index is high, which serves as an input for funding optimization.

[0207] The allocation ratio of backup resource pool funds is dynamically optimized based on the comprehensive emergency index. If the comprehensive emergency index of a scenario is higher than the first threshold, funds are allocated first to purchase energy storage services. If the emergency index is lower than the second threshold, funds are allocated first to subsidize affected participants. A fund allocation plan is generated as the basis for game arbitration and market settlement.

[0208] Based on the equipment response results and linkage control results, the technical feasibility of the scheduling scheme is verified. The technical feasibility includes power flow security and voltage stability. An arbitration result is generated as input for market clearing.

[0209] Based on the arbitration results and the fund allocation plan, the settlement of reward and penalty funds and ancillary service fees is carried out. The reward and penalty funds include the forecast deviation penalty and economic compensation for new energy operators and load aggregators, and the ancillary service fees include the fees for energy storage services and load reduction services. The settlement results are generated and used as input for the edge layer update scheduling scheme.

[0210] Assume the operating scenario involves voltage collapse risk. Extract the impact features:

[0211] Disturbance impact characteristics: voltage drop of 5% (e.g., from 110kV to 104.5kV).

[0212] Impact characteristics: Affects 3 feeders in the strong coupling region.

[0213] Economic impact characteristics: The potential economic loss is 100,000 yuan.

[0214] Priority evaluation based on partitioning results; evaluation of impact characteristics based on partitioning results (large voltage fluctuations in strongly coupled regions):

[0215] Evaluation criteria: Disturbance impact characteristics show a significant voltage drop, weight 0.5; range impact characteristics show impact on the strongly coupled region, weight 0.3; economic impact characteristics show large losses, weight 0.2.

[0216] Comprehensive Emergency Index Calculation: Comprehensive Emergency Index = 0.5 × (5 / 10) + 0.3 × (3 / 6) + 0.2 × (10 / 20) = 0.5.

[0217] The fund allocation is optimized based on the emergency index. Assuming the comprehensive emergency index is 0.5 > the first threshold of 0.4, priority is given to allocating funds from the reserve resource pool for purchasing energy storage services (e.g., allocating 70% of the funds, 7000 yuan). If the emergency index drops to 0.3 < the second threshold of 0.35, priority is given to allocating funds to subsidize affected participants (e.g., allocating 60% of the funds, 6000 yuan). A fund allocation plan is generated to serve as the basis for game arbitration and market settlement.

[0218] Verify the technical feasibility based on the execution results. Verify the dispatching scheme based on the equipment response results (ESS1 discharges 50kW, PV1 output increases by 60kW) and the linkage control results (energy storage compensation 50kW, load shedding 100kW, grid support 50kW).

[0219] Power flow safety: Power flow calculation shows that the power of feeder 1 is not exceeded.

[0220] Voltage stability: The voltage of bus 1 has recovered to 105kV, which meets the safety range.

[0221] Arbitration result: The scheduling scheme is technically feasible and can be used as input for market clearing.

[0222] Market liquidation based on arbitration results and fund allocation plan.

[0223] Based on the arbitration result (technically feasible) and the funding allocation plan (purchasing energy storage services for 7,000 yuan), the settlement is as follows:

[0224] Reward and penalty funds: New energy operators will pay a penalty of 500 yuan for forecasting deviations, and load aggregators will pay 300 yuan for deviations.

[0225] Ancillary service fees: Energy storage service fee 4,000 yuan, load shedding service fee 2,000 yuan.

[0226] Settlement Result: Generate settlement result, including a penalty of 800 yuan and an auxiliary service fee of 6,000 yuan, as input for the edge layer update scheduling scheme.

[0227] The methods for accelerating equilibrium point search using quantum-inspired algorithms at the edge layer and updating the scheduling scheme based on the execution results include:

[0228] The execution results are decomposed in multiple dimensions to extract deviation features. The execution results include equipment response results and linkage control results, and the deviation features include equipment response deviation features and linkage control deviation features, which serve as the basis for strategy optimization.

[0229] Based on the scheduling scheme, the deviation characteristics are prioritized and a deviation priority sequence is generated. The evaluation criteria are the degree of impact of the deviation characteristics on the stability and economy of the substation. If the equipment response deviation characteristics show that the power regulation exceeds the range of the scheduling scheme, the priority is high. If the linkage control deviation characteristics show that the voltage fluctuation exceeds the safe range, the priority is high. This serves as the input for the equilibrium point search.

[0230] A quantum-inspired algorithm is used to accelerate the search for equilibrium points based on the liquidation results. The liquidation results are used to adjust the reward and punishment rules of the game framework and generate an optimized strategy set, which serves as the basis for updating the scheduling scheme.

[0231] Based on the optimized strategy set and deviation priority sequence, the scheduling scheme is optimized. The optimization includes adjusting the active / reactive power regulation range, charging and discharging strategy, load transfer plan and power purchase and sale plan of each subject, and generating an updated scheduling scheme as input for the terminal layer to adjust the operating status of equipment.

[0232] When the system state changes, the deviation priority sequence is re-evaluated through the continuous optimization function, and the optimized strategy set is adjusted to generate an adjusted update scheduling scheme, which serves as the update input for the terminal layer to adjust the device operating state.

[0233] Assuming the integration layer has generated execution results, including device response results (e.g., PV1 actual output increases by 60kW, ESS1 energy storage power conversion system discharges 50kW) and linkage control results (e.g., energy storage compensation of 50kW, load shedding of 100kW, grid support of 50kW), deviation features are extracted using a multi-dimensional decomposition method.

[0234] Equipment response deviation characteristics: Compare the difference between the actual equipment response and the scheduling plan. For example, the scheduling plan requires PV1 output to increase by 50kW, and the actual increase is 60kW, with a deviation of +10kW; ESS1 requires discharge of 50kW, and the actual discharge is 50kW, with a deviation of 0kW.

[0235] Interlocking control deviation characteristics: Compare the difference between the actual effect of the interlocking control and the target. For example, the target is to restore the voltage of bus 1 to 105kV, and the actual voltage is restored to 104kV, with a deviation of -1kV.

[0236] Prioritization of dispatch schemes: Based on dispatch schemes (e.g., increasing photovoltaic output by 50kW, discharging energy storage by 50kW, and shedding load by 100kW), prioritize the deviation characteristics.

[0237] Evaluation criteria: If the equipment response deviation characteristics show that the power adjustment exceeds the range of the scheduling scheme (e.g., PV1 deviation +10kW), then the priority is high; if the linkage control deviation characteristics show that the voltage fluctuation exceeds the safe range (e.g., voltage deviation -1kV, safe range ±0.5kV), then the priority is high.

[0238] Deviation priority sequence: Generate a sequence, such as voltage deviation (-1kV, priority 0.8) > PV1 power deviation (+10kW, priority 0.6) > ESS1 power deviation (0kW, priority 0.1), as input for the equilibrium point search.

[0239] Based on the settlement results, a quantum-inspired algorithm is used to accelerate the equilibrium point search. Assume the settlement results at the integration layer show that the new energy operator pays a prediction deviation penalty of 500 yuan and an energy storage service fee of 4000 yuan. The reward and penalty rules of the game framework are adjusted based on the settlement results, for example, increasing the penalty ratio by 10% (from 0.1 yuan per kW deviation to 0.11 yuan). A quantum-inspired algorithm (such as quantum annealing) is used to accelerate the equilibrium point search.

[0240] Search process: Using the deviation priority sequence as a constraint, optimize the strategies of each entity (e.g., new energy clusters reduce output deviation, energy storage alliances adjust discharge strategies).

[0241] Optimized strategy set: Generate an optimized strategy set, such as controlling the output deviation of the new energy cluster within ±5kW and the discharge deviation of the energy storage alliance within ±2kW, as the basis for updating the scheduling scheme.

[0242] The scheduling scheme is optimized based on the optimized policy set and the bias priority sequence.

[0243] Optimization details: Adjust the active power regulation range of the new energy cluster to ±5kW, the charging and discharging strategy of the energy storage alliance to discharge deviation ±2kW, the load transfer plan to cut off deviation ±10kW, and the power purchase and sale plan to support deviation ±5kW.

[0244] Updated scheduling scheme: Generate an updated scheduling scheme, such as PV1 output increase of 45kW, ESS1 discharge of 48kW, load shedding of 90kW, and grid support of 45kW, as input for the terminal layer to adjust the operating status of equipment.

[0245] When the system state changes, the strategy set is adjusted. Assuming a change in system state (e.g., a sudden 10% increase in load demand), the deviation priority sequence is re-evaluated using a continuous optimization function (e.g., an exponential decay function). For example, if a sudden load increase causes the voltage deviation priority to rise from 0.8 to 0.9, the optimized strategy set is adjusted (e.g., the load transfer plan deviation is relaxed to ±15kW), generating an updated scheduling scheme, which serves as the update input for the terminal layer to adjust the operating status of equipment.

[0246] Other methods for updating scheduling schemes based on execution results include:

[0247] After adding the settlement results to the execution results, multi-dimensional integration is performed to extract optimization features, including equipment response optimization features, linkage control optimization features, and economic optimization features, which serve as the basis for rolling optimization.

[0248] Based on the scheduling scheme, the optimization features are prioritized and an optimization priority sequence is generated. The evaluation criteria are the degree to which the optimization features improve the stability and economy of the substation. If the equipment response optimization features show a reduction in power regulation deviation, the priority is high. If the economic optimization features show a reduction in cost, the priority is high. This serves as the input for rolling optimization.

[0249] Based on the optimized priority sequence and the updated scheduling scheme, the scheduling scheme is rolled out for optimization. Rolling optimization includes refreshing the three-dimensional uncertainty cloud map and the scheduling scheme at preset time intervals to generate an optimized scheduling scheme, which serves as the input for the terminal layer to adjust the device's operating status.

[0250] When a sudden event is detected, the system immediately switches to emergency mode based on the linkage control results. In emergency mode, rapid response resources are prioritized, including energy storage devices and adjustable load resources. An emergency management plan is generated as a temporary input for the terminal layer to adjust the operating status of the equipment.

[0251] When the substation status changes, the optimization priority sequence is re-evaluated through a continuous optimization function, and the optimized scheduling scheme is adjusted to generate an adjusted optimized scheduling scheme, which serves as the update input for the terminal layer to adjust the operating status of the equipment.

[0252] Add liquidation results (e.g., penalty of 800 yuan, auxiliary service fee of 6000 yuan) to the execution results for multi-dimensional integration and extraction of optimization features:

[0253] Equipment response optimization features: PV1 power regulation deviation decreased from +10kW to +5kW, an optimization of 50%.

[0254] Linkage control optimization features: voltage deviation reduced from -1kV to -0.5kV, with an optimization range of 50%.

[0255] Economic optimization features: Operating costs were reduced from 100,000 yuan to 90,000 yuan, an optimization rate of 10%.

[0256] Based on the updated scheduling scheme (e.g., PV1 output increased by 45kW, ESS1 discharge increased by 48kW), evaluate the priority of the optimization features:

[0257] Evaluation criteria: If the device response optimization feature shows a reduction in power adjustment deviation (e.g., PV1 optimization by 50%), it has a high priority; if the economic optimization feature shows a reduction in cost (e.g., optimization by 10%), it has a high priority.

[0258] Optimize priority sequence: Generate a sequence, such as PV1 power optimization (priority 0.7) > voltage optimization (priority 0.6) > cost optimization (priority 0.4), as input for rolling optimization.

[0259] Based on the optimized priority sequence and the updated scheduling scheme, perform rolling optimization:

[0260] Rolling optimization process: The 3D uncertainty cloud map (e.g., updating the photovoltaic output probability distribution) and scheduling scheme are refreshed every 5 minutes. For example, the PV1 output plan is adjusted to increase by 40kW, and the ESS1 discharge plan is adjusted to 46kW.

[0261] Optimized scheduling scheme: Generate an optimized scheduling scheme, such as increasing PV1 output by 40kW, discharging ESS1 by 46kW, cutting off 85kW of load, and providing grid support of 40kW, as input for the terminal layer to adjust the operating status of equipment.

[0262] If a sudden event is detected (e.g., a 20% drop in photovoltaic output), the system will immediately switch to emergency mode based on the linkage control results (e.g., energy storage compensation of 50kW).

[0263] Emergency mode: Prioritize the use of fast response resources, such as increasing ESS1 discharge to 60kW and adjusting load shedding to 120kW.

[0264] Emergency Management Plan: Generate emergency management plans, such as ESS1 discharging 60kW and load shedding 120kW, as temporary inputs for the terminal layer to adjust the operating status of equipment.

[0265] Assuming a change in system state (e.g., load demand returns to normal), the optimization priority sequence is re-evaluated using a continuous optimization function (e.g., a linear interpolation function). For example, if the cost optimization priority increases from 0.4 to 0.6, the optimized scheduling scheme is adjusted (e.g., the load shedding plan is reduced to 80kW), generating an adjusted optimized scheduling scheme as an update input for the terminal layer to adjust the equipment operating status.

[0266] The terminal layer dynamically reconstructs the impedance characteristics of the equipment through a power electronic switch array and adjusts the equipment operating status based on the updated scheduling scheme. The methods include:

[0267] The updated scheduling scheme is decomposed in multiple dimensions to extract adjustment features. The updated scheduling scheme includes the updated scheduling scheme and the optimized scheduling scheme. The adjustment features include impedance adjustment features, power adjustment features and time adjustment features, which serve as the basis for dynamic reconfiguration.

[0268] Based on the adjustment characteristics, the impedance characteristics of the equipment are dynamically reconstructed through the power electronic switch array. The impedance characteristics include the real and imaginary parts of the virtual impedance. The reconstruction includes the simulated synchronous machine inertial response of the new energy equipment and the short-circuit capacity support provided by the energy storage equipment. The reconstructed impedance characteristics are generated as the input for adjusting the operating state of the equipment.

[0269] Based on the reconstructed impedance characteristics, the equipment operating status is adjusted, which includes actual output, state of charge, and impedance value, to generate the adjusted equipment operating status as the input for the next round of prediction.

[0270] The adjusted device operating status is fed back to the edge layer in real time to update the execution results, generate the feedback execution results, and serve as the input for the next round of game arbitration and market settlement in the integration layer;

[0271] When the substation status changes, the adjustment characteristics are re-evaluated through a continuous optimization function, and the reconstructed impedance characteristics are adjusted to generate an updated equipment operating status, which serves as the update input for the next round of prediction.

[0272] The updated scheduling scheme (including the updated and optimized scheduling schemes, such as increasing PV1 output by 40kW and ESS1 discharge by 46kW) is decomposed in multiple dimensions to extract adjustment features:

[0273] Impedance adjustment characteristics: PV1 needs to be adjusted to virtual impedance of 0.12 + j0.06 ohms, and ESS1 needs to be adjusted to 0.09 + j0.04 ohms.

[0274] Power adjustment features: PV1 output is adjusted to 490kW, and ESS1 discharge is adjusted to 46kW.

[0275] Time adjustment feature: The adjustment time is completed within 5 seconds.

[0276] Based on adjustment characteristics, the impedance characteristics of the device are dynamically reconstructed using a power electronic switch array:

[0277] PV1 Reconfiguration: Adjust the virtual impedance to 0.12 + j0.06 ohms to simulate the inertial response of the synchronous machine, increasing the output to 490kW.

[0278] ESS1 reconfiguration: Adjust the virtual impedance to 0.09 + j0.04 ohms to provide short-circuit capacity support, with a discharge capacity of 46kW.

[0279] Reconstructed impedance characteristics: Generate reconstructed impedance characteristics, such as PV1 impedance 0.12 + j0.06 ohms and ESS1 impedance 0.09 + j0.04 ohms, as inputs for adjusting the device operating status.

[0280] Based on the reconstructed impedance characteristics, adjust the equipment operating status:

[0281] PV1 operating status: actual output 490kW, impedance 0.12+j0.06 ohms.

[0282] ESS1 operating status: actual discharge 46kW, state of charge (SOC) dropped from 80% to 78%, impedance 0.09+j0.04 ohms.

[0283] Adjusted equipment operating status: Generate the adjusted equipment operating status, such as PV1 output 490kW and ESS1 state of charge 78%, as input for the next round of prediction.

[0284] The adjusted equipment operating status is fed back to the edge layer in real time, such as PV1 output of 490kW and ESS1 state of charge of 78%, to update the execution results. The feedback execution results are generated, such as equipment response deviation reduced to +2kW and voltage deviation reduced to -0.3kV, as input for the next round of game arbitration and market settlement in the integration layer.

[0285] Assuming a change in system state (e.g., a sudden 10% increase in load demand), the adjustment characteristics are reassessed using a continuous optimization function (e.g., an exponential decay function). For example, the impedance adjustment characteristics of PV1 need to be adjusted to 0.10 + j0.05 ohms. The reconstructed impedance characteristics are then used to generate updated equipment operating states, such as PV1 output of 495kW and ESS1 state of charge of 76%, which serve as the update input for the next round of prediction.

[0286] Example 2

[0287] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the cloud computing-based intelligent substation energy management method described above.

[0288] Since the electronic device described in this embodiment is the electronic device used to implement the cloud computing-based intelligent substation energy management method described in this application embodiment, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the cloud computing-based intelligent substation energy management method described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. Any electronic device used by those skilled in the art to implement the cloud computing-based intelligent substation energy management method described in this application embodiment falls within the scope of protection of this application.

[0289] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0290] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A cloud computing-based intelligent substation energy management method, characterized in that, include: At the source-load forecasting layer, a "forecasting-scheduling" game framework is constructed, in which new energy operators and load aggregators submit forecast commitment intervals to the cloud platform, and reward and punishment rules are implemented based on actual deviations; by integrating regional meteorological data and the output correlation of adjacent substations, a three-dimensional uncertainty cloud map with spatiotemporal coupling characteristics is generated as the decision basis for the scheduling and allocation layer. At the dispatch and allocation layer, substation resources are divided into four types of autonomous decision-making entities: new energy clusters, energy storage alliances, adjustable load groups, and grid interface agents. The edge layer solves the Nash equilibrium point based on the three-dimensional uncertainty cloud map to generate a dispatch scheme. Dynamic electrical partitioning is performed based on the feeder voltage-power sensitivity matrix, and the partitioning results are obtained. The scheduling scheme and partitioning results are used as inputs to the execution control layer. At the execution control layer, programmable impedance characteristics are embedded in photovoltaic inverters and energy storage power conversion systems to enable autonomous equipment response based on scheduling schemes; The source-load fluctuation event chain model performs cross-timescale linkage control based on partitioning results; and feeds back the execution results to the integration layer. At the integration layer, the cloud platform performs game arbitration and market settlement based on the execution results. The edge layer uses quantum-inspired algorithms to accelerate the search for equilibrium points and updates the scheduling scheme based on the execution results. The terminal layer dynamically reconstructs the impedance characteristics of the equipment through power electronic switch arrays and adjusts the operating status of the equipment based on the updated scheduling scheme.

2. The intelligent substation energy management method based on cloud computing according to claim 1, characterized in that, The method for generating the three-dimensional uncertainty cloud map includes: The game participants are initialized, with new energy operators and load aggregators as the main participants in the game framework; the game rules are set, including that if the actual value is within the predicted commitment range, the participant will receive scheduling priority and economic compensation, and if the actual value exceeds the predicted commitment range, the participant will pay a deviation penalty, which is injected into the standby resource pool as an incentive mechanism for the game framework; the participants submit the initial predicted commitment range as the input to the game framework to construct the "prediction-scheduling" game framework. Based on the game theory framework, environmental data is decomposed hierarchically to extract disturbance features at different time and spatial scales. The environmental data includes wind speed, light intensity and load demand, and the disturbance features include minute-level abrupt change features, hourly-level trend features, local abrupt change features and regional trend features, which serve as the basis for predicting the reshaping of the commitment interval. For perturbation features of different standards, minute-level mutation features have higher priority than hour-level trend features, local mutation features have higher priority than regional trend features, and for perturbation features of the same standard, priority is ranked according to the degree of impact of the perturbation features on the prediction accuracy. Based on the priority sequence and the strength of the perturbation features, the leniency of the initial prediction commitment interval is dynamically calibrated. That is, if a high-priority minute-level mutation feature is detected, the tolerance range of the prediction commitment interval is significantly widened and the penalty intensity is reduced. If only a medium-priority hour-level trend feature is detected, the prediction commitment interval is moderately widened to generate an adjusted prediction commitment interval, which serves as the update input for the game framework. Based on the adjusted forecast commitment range, new energy operators and load aggregators submit the forecast commitment range to the cloud platform, and implement reward and penalty rules according to the actual deviation to generate forecast deviation data, which serves as the basis for correcting the generation of the three-dimensional uncertainty cloud map. By integrating prediction bias data, regional meteorological data, and the output correlation of adjacent substations, a three-dimensional uncertainty cloud map with spatiotemporal coupling characteristics is generated. The three-dimensional uncertainty cloud map includes the distribution of time, power, and probability of occurrence.

3. The intelligent substation energy management method based on cloud computing according to claim 2, characterized in that, The method for solving the Nash equilibrium point based on a three-dimensional uncertainty cloud map and generating a scheduling scheme includes: Substation resources are divided into four types of autonomous decision-making entities: new energy clusters, energy storage alliances, adjustable load groups, and grid interface agents. The goal of the new energy clusters is to maximize the absorption capacity, the goal of the energy storage alliances is to minimize the lifespan loss, the goal of the adjustable load groups is to minimize the electricity cost, and the goal of the grid interface agents is to meet the upper-level dispatch instructions, serving as the basis for strategy conflict negotiation. Conflict features are extracted from the objective functions of each autonomous decision-making body to generate a conflict feature set. The conflict features include objective priority features and constraint boundary features. The objective priority features include the priority of the absorption capacity of the new energy cluster and the priority of the life loss of energy storage. The constraint boundary features include the SOC protection boundary of energy storage and the reduction capacity boundary of adjustable load, which serve as the basis for hierarchical negotiation. The conflict feature set is decomposed into multiple conflict levels based on the severity and scope of the conflict. The conflict levels include the initial conflict level and the severe conflict level. The initial conflict level corresponds to the conflict where the SOC is close to exceeding the limit, and the severe conflict level corresponds to the conflict where the SOC actually exceeds the limit. These levels serve as the input for constraint convergence. Based on the three-dimensional uncertainty cloud map, a multi-stage negotiation mechanism is used to resolve policy conflicts at the conflict level. In the initial conflict level, the new energy cluster adjusts its output plan to meet the SOC protection boundary of energy storage. In the severe conflict level, the energy storage alliance relaxes the SOC protection boundary. The negotiation process adopts an iterative approach, with each iteration narrowing the feasible domain of the strategy and generating a negotiated strategy set as input for solving the Nash equilibrium. The edge layer solves for the Nash equilibrium point based on the negotiated strategy set and the three-dimensional uncertainty cloud map, and generates a scheduling scheme. The scheduling scheme includes the active / reactive power adjustment range, charging and discharging strategy, load transfer plan and power purchase and sale plan of each entity.

4. The intelligent substation energy management method based on cloud computing according to claim 3, characterized in that, The method for obtaining the partitioning results based on the feeder voltage-power sensitivity matrix for dynamic electrical partitioning includes: The substation status data is decomposed into multiple dimensions to extract change features. The substation status data includes voltage, power and power flow distribution. The change features serve as the basis for disturbance perception, including instantaneous change features, trend change features and spatial distribution features. Based on the three-dimensional uncertainty cloud map, the change characteristics are comprehensively evaluated to generate the disturbance index of the substation status. The comprehensive evaluation is based on the intensity and duration of the change characteristics. If the instantaneous change characteristics are strong and short in duration, the disturbance index is high. If the trend change characteristics are weak and long in duration, the disturbance index is low. This serves as the input for frequency optimization. Based on the disturbance index, the update frequency of the feeder voltage-power sensitivity matrix is ​​dynamically adjusted. If the disturbance index is higher than the first threshold, the update frequency is increased to the first frequency value. If the disturbance index is lower than the second threshold, the update frequency is decreased to the second frequency value, generating an optimized update frequency as the basis for dynamic electrical zoning. Based on the optimized update frequency and scheduling scheme, the feeder voltage-power sensitivity matrix is ​​calculated, and dynamic electrical partitioning is generated. The dynamic electrical partitioning includes a strongly coupled region and a weakly coupled region. The strongly coupled region performs centralized optimization scheduling, while the weakly coupled region performs autonomous decision scheduling, and the partitioning results are generated. When the partition changes, the scheduling rules are adjusted through the continuous optimization function to generate the adjusted partition result, which serves as the update input for the linkage control of the execution control layer.

5. The intelligent substation energy management method based on cloud computing according to claim 4, characterized in that, The method for embedding programmable impedance characteristics into photovoltaic inverters and energy storage power conversion systems, and enabling autonomous equipment response based on scheduling schemes, includes: The equipment's operational data is decomposed into multiple levels to extract impact features. The operational data includes the operating status of the photovoltaic inverter and the energy storage power conversion system, while the impact features include geographical impact features, power impact features, and sensitivity impact features, which serve as the basis for equipment impact assessment. Based on the partitioning results, the impact features are prioritized and evaluated to generate a comprehensive importance index for the equipment. The evaluation criteria are the contribution of the impact features to the system stability. If the geographical impact features of the equipment show that it is located in a strong coupling area and the sensitivity impact features show that it has a large impact on voltage, then the comprehensive importance index is high and it is used as the input for priority adjustment. Based on the comprehensive importance index, the impedance parameter adjustment priority of the equipment is dynamically sorted. Among them, the equipment with a high comprehensive importance index is adjusted first, while the equipment with a low comprehensive importance index is adjusted later. This generates an equipment adjustment sequence as the basis for the equipment's autonomous response. Based on the scheduling scheme and the equipment adjustment sequence, programmable impedance characteristics are embedded in the photovoltaic inverter and energy storage power conversion system to execute autonomous equipment response. The autonomous response includes the simulated synchronous machine inertial response of the new energy equipment and the short-circuit capacity support provided by the energy storage equipment, generating equipment response results as inputs for the linkage control of the execution control layer. When the system state changes, the comprehensive importance index is re-evaluated through a continuous update function, and the equipment adjustment sequence is adjusted to generate an updated equipment adjustment sequence, which serves as the update input for the linkage control of the execution control layer.

6. The intelligent substation energy management method based on cloud computing according to claim 5, characterized in that, The method for cross-timescale linkage control based on partitioning results in the source-load fluctuation event chain model includes: The indicators involved in the event chain are decomposed to extract the impact features. The indicators include photovoltaic output, load demand and voltage level. The impact features include instantaneous impact features, trend impact features and coupling impact features, which serve as the basis for correlation and integration. Based on the partitioning results, dynamic weights are assigned to the influencing features to generate a weight sequence. The assignment is based on the degree of influence of the influencing features on the stability of the substation. If the instantaneous influencing features have a large impact on the stability of the substation, they are assigned high weights. If the trend influencing features have a small impact, they are assigned low weights. The weights are dynamically adjusted according to the strong coupling and weak coupling regions of the partitioning results and serve as inputs for conditional optimization. By using a weighted fusion function to correlate and fuse the influencing features and weight sequences, composite triggering conditions are generated, and a comprehensive influence index is generated, which serves as the basis for coordinated control. Based on the equipment response results and composite triggering conditions, cross-timescale linkage control is executed. The linkage control includes millisecond-level energy storage compensation, second-level load shedding and minute-level grid support, generating linkage control results as input to the execution results of the integration layer. The trigger threshold of the composite triggering condition is dynamically optimized based on the disturbance index of the substation status. If the disturbance index of the substation status is high, the trigger threshold is lowered; if the disturbance index of the substation status is low, the trigger threshold is raised. The optimized composite triggering condition is generated and used as the update input for the execution result of the integration layer.

7. The intelligent substation energy management method based on cloud computing according to claim 6, characterized in that, The cloud platform's methods for game arbitration and market settlement based on execution results include: The substation operation scenarios are decomposed in multiple dimensions to extract impact characteristics. The operation scenarios include voltage collapse risk, insufficient renewable energy absorption and load change. The impact characteristics include disturbance impact characteristics, range impact characteristics and economic impact characteristics, which serve as the basis for scenario impact assessment. Based on the partitioning results, the impact features are prioritized and a comprehensive emergency index for the scenario is generated. The evaluation criteria are the degree of impact of the impact features on the substation operation. If the disturbance impact features show a high risk of voltage collapse and the range impact features show a strong coupling area, the emergency index is high, which serves as an input for funding optimization. The allocation ratio of backup resource pool funds is dynamically optimized based on the comprehensive emergency index. If the comprehensive emergency index of a scenario is higher than the first threshold, funds are allocated first to purchase energy storage services. If the emergency index is lower than the second threshold, funds are allocated first to subsidize affected participants. A fund allocation plan is generated as the basis for game arbitration and market settlement. Based on the equipment response results and linkage control results, the technical feasibility of the scheduling scheme is verified. The technical feasibility includes power flow security and voltage stability. An arbitration result is generated as input for market clearing. Based on the arbitration results and the fund allocation plan, the settlement of reward and penalty funds and ancillary service fees is carried out. The reward and penalty funds include the forecast deviation penalty and economic compensation for new energy operators and load aggregators, and the ancillary service fees include the fees for energy storage services and load reduction services. The settlement results are generated and used as input for the edge layer update scheduling scheme.

8. The intelligent substation energy management method based on cloud computing according to claim 7, characterized in that, The method for accelerating equilibrium point search using quantum-inspired algorithms in the edge layer and updating the scheduling scheme based on the execution results includes: The execution results are decomposed in multiple dimensions to extract deviation features. The execution results include equipment response results and linkage control results, and the deviation features include equipment response deviation features and linkage control deviation features, which serve as the basis for strategy optimization. Based on the scheduling scheme, the deviation characteristics are prioritized and a deviation priority sequence is generated. The evaluation criteria are the degree of impact of the deviation characteristics on the stability and economy of the substation. If the equipment response deviation characteristics show that the power regulation exceeds the range of the scheduling scheme, the priority is high. If the linkage control deviation characteristics show that the voltage fluctuation exceeds the safe range, the priority is high. This serves as the input for the equilibrium point search. A quantum-inspired algorithm is used to accelerate the search for equilibrium points based on the liquidation results. The liquidation results are used to adjust the reward and punishment rules of the game framework and generate an optimized strategy set, which serves as the basis for updating the scheduling scheme. Based on the optimized strategy set and deviation priority sequence, the scheduling scheme is optimized. The optimization includes adjusting the active / reactive power regulation range, charging and discharging strategy, load transfer plan and power purchase and sale plan of each subject, and generating an updated scheduling scheme as input for the terminal layer to adjust the operating status of equipment. When the system state changes, the deviation priority sequence is re-evaluated through the continuous optimization function, and the optimized strategy set is adjusted to generate an adjusted update scheduling scheme, which serves as the update input for the terminal layer to adjust the device operating state.

9. The intelligent substation energy management method based on cloud computing according to claim 8, characterized in that, The method for updating the scheduling scheme based on the execution result also includes: After adding the settlement results to the execution results, multi-dimensional integration is performed to extract optimization features, including equipment response optimization features, linkage control optimization features, and economic optimization features, which serve as the basis for rolling optimization. Based on the scheduling scheme, the optimization features are prioritized and an optimization priority sequence is generated. The evaluation criteria are the degree to which the optimization features improve the stability and economy of the substation. If the equipment response optimization features show a reduction in power regulation deviation, the priority is high. If the economic optimization features show a reduction in cost, the priority is high. This serves as the input for rolling optimization. Based on the optimized priority sequence and the updated scheduling scheme, the scheduling scheme is rolled out for optimization. Rolling optimization includes refreshing the three-dimensional uncertainty cloud map and the scheduling scheme at preset time intervals to generate an optimized scheduling scheme, which serves as the input for the terminal layer to adjust the device's operating status. When a sudden event is detected, the system immediately switches to emergency mode based on the linkage control results. In emergency mode, rapid response resources are prioritized, including energy storage devices and adjustable load resources. An emergency management plan is generated as a temporary input for the terminal layer to adjust the operating status of the equipment. When the substation status changes, the optimization priority sequence is re-evaluated through a continuous optimization function, and the optimized scheduling scheme is adjusted to generate an adjusted optimized scheduling scheme, which serves as the update input for the terminal layer to adjust the operating status of the equipment.

10. The intelligent substation energy management method based on cloud computing according to claim 9, characterized in that, The terminal layer dynamically reconstructs the device impedance characteristics through a power electronic switch array and adjusts the device operating status based on the updated scheduling scheme. The method includes: The updated scheduling scheme is decomposed in multiple dimensions to extract adjustment features. The updated scheduling scheme includes the updated scheduling scheme and the optimized scheduling scheme. The adjustment features include impedance adjustment features, power adjustment features and time adjustment features, which serve as the basis for dynamic reconfiguration. Based on the adjustment characteristics, the impedance characteristics of the equipment are dynamically reconstructed through the power electronic switch array. The impedance characteristics include the real and imaginary parts of the virtual impedance. The reconstruction includes the simulated synchronous machine inertial response of the new energy equipment and the short-circuit capacity support provided by the energy storage equipment. The reconstructed impedance characteristics are generated as the input for adjusting the operating state of the equipment. Based on the reconstructed impedance characteristics, the equipment operating status is adjusted, which includes actual output, state of charge, and impedance value, to generate the adjusted equipment operating status as the input for the next round of prediction. The adjusted device operating status is fed back to the edge layer in real time to update the execution results, generate the feedback execution results, and serve as the input for the next round of game arbitration and market settlement in the integration layer; When the substation status changes, the adjustment characteristics are re-evaluated through a continuous optimization function, and the reconstructed impedance characteristics are adjusted to generate an updated equipment operating status, which serves as the update input for the next round of prediction.

Images