Virtual power plant-based power distribution area collaborative treatment system and method

Abstract

The invention discloses a virtual power plant-based power distribution area collaborative governance system and method, and relates to the technical field of power system automation and intelligent power distribution networks, and the system comprises a dynamic autonomous partition governance unit which is used for dividing a power distribution area node into a plurality of subunits, correcting abnormal states in the subunits based on a power mutual aid strategy between adjacent subunits; the multi-time-scale collaborative optimization unit is used for constructing a second-level, minute-level and hour-level multi-time-scale collaborative mechanism and coordinating control targets of the power distribution area at different time scales; the edge-cloud collaborative decision-making unit is used for complementing the edge side real-time control strategy and the cloud global strategy; and the elastic resource pool unit is used for calculating the elastic contribution degree of the resources and formulating priorities, and calling the resources according to the priorities to meet load requirements. By integrating the virtual power plant technology, the edge calculation, the digital twinning and the multi-time scale optimization strategy, the intelligent and efficient treatment of the power distribution area is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power system automation and intelligent distribution network technology, and in particular to a distribution network collaborative management system and method based on a virtual power plant. Background Art

[0002] A distribution substation is an area within the power distribution system responsible for converting electrical energy from substations through transformers into low-voltage power suitable for consumer use. This power is then distributed to individual users via distribution equipment and lines. It typically includes equipment such as distribution transformers, distribution switches, and circuit breakers, which control and protect the flow of power, ensuring a secure and stable power supply. The division and management of distribution substations are typically handled by power supply companies, with design and maintenance based on regional electricity demand and technical standards.

[0003] With the integration of new loads such as distributed energy resources, energy storage devices, and electric vehicles, the operation of distribution substations has become increasingly complex. Traditional distribution substation management methods struggle to meet the growing demand for electricity and the requirements for high-quality power supply. They suffer from voltage fluctuations, excessive harmonics, supply-demand imbalances, and insufficient resilience under extreme operating conditions. Therefore, a new collaborative distribution substation management system and method is urgently needed to improve the operational efficiency and power supply reliability of the distribution network. Summary of the Invention

[0004] (1) Technical problems solved

[0005] In response to the shortcomings of the existing technology, the present invention provides a distribution substation collaborative management system and method based on a virtual power plant, which has the advantages of voltage stability, cost reduction and enhanced resilience to extreme working conditions, thereby solving the problem of insufficient resilience of traditional distribution substations under voltage fluctuations, excessive harmonics, supply and demand imbalance and extreme working conditions.

[0006] (2) Technical solution

[0007] To achieve the above advantages of voltage stability, cost reduction, and enhanced resilience under extreme working conditions, the specific technical solutions adopted by the present invention are as follows:

[0008] According to one aspect of the present invention, a distribution substation collaborative management system based on a virtual power plant is provided, the system comprising:

[0009] Data collection unit, used to collect real-time operation data of nodes in the distribution area and virtualize resource attributes;

[0010] Dynamic autonomous partition management unit, used to divide the distribution area node into several sub-units and correct abnormal conditions within the sub-unit based on the power mutual assistance strategy between adjacent sub-units;

[0011] Multi-time scale collaborative optimization unit, used to build multi-time scale collaborative mechanisms at the second, minute, and hour levels, and coordinate the control objectives of distribution substations at different time scales;

[0012] Edge-cloud collaborative decision-making unit, which complements edge-side real-time control strategies with cloud-based global strategies to achieve maintenance of edge devices;

[0013] The elastic resource pool unit is used to calculate the elastic contribution of resources and set priorities, and call resources according to the priorities to meet load requirements.

[0014] Preferably, the dynamic autonomous partition management unit divides the distribution area node into several sub-units and corrects the abnormal state in the sub-unit based on the power mutual assistance strategy between adjacent sub-units, including:

[0015] The improved K-means clustering algorithm is used to divide the distribution area nodes with the same similar characteristics into the same group to obtain several sub-units;

[0016] Identify local abnormal scenarios of subunits, including voltage exceeding the limit and harmonic exceeding the limit;

[0017] Triggering the autonomous control mechanism to correct abnormal scenarios, including:

[0018] Based on the voltage over-limit scenario, if the voltage of the distribution area node exceeds the preset range, the energy storage device of the sub-unit is triggered to provide voltage support;

[0019] Based on the harmonic exceeding standard scenario, if the total harmonic distortion rate of the distribution area node exceeds the preset range, the adjacent sub-units will be triggered to perform harmonic compensation.

[0020] Preferably, the improved K-means clustering algorithm is used to divide the distribution area nodes with the same similar characteristics into the same group, and the obtained sub-units include:

[0021] Based on historical operating data, the distribution substation node with the largest load fluctuation is selected as the initial cluster center, and the weight coefficients of power imbalance and energy storage regulation potential are set;

[0022] Iteratively calculate the distance between each distribution area node and the cluster center, and assign the distribution area node to the subunit with the nearest cluster center;

[0023] The power imbalance and energy storage regulation potential of each sub-unit are calculated based on the clustering results, and an objective function is constructed with the goal of minimizing the power imbalance and maximizing the energy storage regulation potential.

[0024] The objective function is solved using a local search strategy to obtain the optimal solution, the clustering results are optimized based on the optimal solution, and the iteration is stopped when the clustering results converge or the maximum number of iterations is met.

[0025] Preferably, the multi-timescale collaborative optimization unit includes:

[0026] Second-level control module, used to suppress voltage flicker and frequency fluctuation in the distribution area based on inverter reactive power regulation and energy storage reactive power regulation;

[0027] The minute-level scheduling module is used to optimize the operation strategy of energy storage and adjustable load in the distribution station area on a preset minute time scale with the goal of short-term economy;

[0028] The hourly planning module is used to formulate the day-ahead resource allocation strategy for the distribution substation based on the load forecast and photovoltaic forecast for the distribution substation in the future time period.

[0029] Preferably, the hourly planning module includes the following when formulating a day-ahead resource allocation strategy for the distribution substation based on the load forecast and photovoltaic forecast for the distribution substation in the future time period:

[0030] Based on historical operating data, the long short-term memory network model is used to predict the load curve of the distribution substation in the future time period;

[0031] Based on cloud cover and irradiance data, a convolutional neural network is combined with a long short-term memory network model to predict the photovoltaic output curve of the distribution station area in the future time period.

[0032] An economic dispatch model is constructed based on the load curve and photovoltaic output curve of the distribution substation in the future time period. With the goal of minimizing the sum of the grid's electricity purchase cost and energy storage cycle loss, the economic dispatch model is solved in combination with predefined constraints to output the day-ahead resource allocation strategy for the distribution substation.

[0033] Preferably, the expression of the economic dispatch model includes:

[0034]

[0035] constraint

[0036] Where, P grid (t) represents the actual grid power at time t; L j represents the set of nodes connected by line j; P i (t) represents the power of node i at time t; S j represents the maximum transmission capacity of line j; C grid (t) represents the time-of-use electricity price of the power grid at time t; C ESS Indicates the rated capacity of the energy storage device; Indicates the upper limit of the maximum allowable charging power of energy storage; Indicates the upper limit of the maximum allowable discharge power of energy storage; P PV (t) represents the output of photovoltaic equipment; P ESS (t) represents the actual charging and discharging power of the energy storage at time t; P load (t) represents the load power.

[0037] Preferably, the edge-cloud collaborative decision-making unit includes the following steps when complementing the edge-side real-time control strategy with the cloud-side global strategy to achieve maintenance of edge devices:

[0038] Simplify the dimensions of the system matrix, input matrix, and output matrix, construct state space equations to perform lightweight calculations on edge devices, and output state estimates and control instructions for edge devices;

[0039] Among them, the system matrix is ​​used to analyze the natural evolution of the internal state of the subunit;

[0040] The input matrix is ​​used to control the effect of the state space equation input on the state variables;

[0041] The output matrix is ​​used to analyze the mapping relationship between state variables and observable outputs;

[0042] Obtain the historical operating data of edge devices, and comprehensively calculate the health of edge devices in combination with the state estimation and control instructions of edge devices, and generate maintenance strategies for edge devices based on the health calculation results.

[0043] Preferably, the elastic resource pool unit calculates the elastic contribution of resources and sets priorities, and calls resources according to the priorities to meet the load demand, including:

[0044] Build an elastic resource pool based on energy storage and adjustable load resources, calculate the elastic contribution of resources in the elastic resource pool, and sort the elastic contribution of resources from high to low to obtain the priority order;

[0045] Detect the connection status between the distribution substation and the main grid. If the distribution substation is disconnected from the main grid and in tinkering mode, call the island mode resources in order of priority to meet the critical load requirements;

[0046] Construct constraints for resource power limits, energy storage state of charge safety ranges, resource response time constraints, and line mutual aid capacity limits.

[0047] The power balance after calling resources is verified based on the constraint conditions. If the verification fails, the remaining resources are re-called until the preset conditions are met.

[0048] Preferably, the calculation formula for resource elasticity contribution is:

[0049]

[0050] Where η i (t) represents the resource elasticity contribution of resource i at time t; E flex,i represents the adjustable energy of resource i; Indicates regulatory potential; Indicates economy; T resp,i represents the response time of resource i; α and β represent weight coefficients; λ represents the penalty coefficient; f(SOC i ,N cycle,i ) represents the aging factor.

[0051] According to another aspect of the present invention, a method for collaborative management of distribution substations based on a virtual power plant is provided, the method comprising:

[0052] Collect real-time operating data of nodes in the distribution area and virtualize resource attributes;

[0053] The distribution area node is divided into several sub-units, and the abnormal state in the sub-unit is corrected based on the power mutual assistance strategy between adjacent sub-units;

[0054] Build a multi-time scale coordination mechanism at the second, minute, and hour levels to coordinate the control targets of distribution substations at different time scales;

[0055] Complement edge-side real-time control with cloud-based global policies to achieve maintenance of edge devices;

[0056] Calculate the elastic contribution of resources and set priorities, and call resources according to the priorities to meet load requirements.

[0057] (3) Beneficial effects

[0058] Compared with the existing technology, the present invention provides a system and method for collaborative management of distribution substations based on virtual power plants, which has the following beneficial effects:

[0059] 1. This invention uses an elastic partitioning strategy to respond to load and resource fluctuations in real time, avoiding the overload or resource idleness problems of traditional fixed partitions. It also improves control accuracy and response speed by building a second-minute-hour closed-loop optimization to match different time scale requirements. At the same time, it integrates harmonic coordinated compensation and rapid voltage support to reduce equipment losses and user complaints.

[0060] 2. In the island mode, the elastic resource pool of the present invention can maintain the power supply of critical loads, reduce the risk of large-scale power outages, combine day-ahead scheduling with real-time frequency regulation, reduce the power grid's electricity purchase costs and energy storage cycle losses, and realize equipment health prediction and line loss anomaly location through digital twins, thereby reducing the workload of manual inspections.

[0061] 3. This invention makes resource calling rules transparent through a contribution quantification model, thereby enhancing users' enthusiasm for participating in VPP sharing. At the same time, it supports access to multiple resources such as photovoltaics / energy storage / charging piles, adapts to the rural low-voltage substation to the urban high-density distribution network, and reduces the carbon emission intensity of the substation by optimizing the consumption of renewable energy. The modular design supports plug-and-play and reduces the cost of system transformation and upgrade.

[0062] 4. The present invention realizes intelligent and efficient management of distribution substations by integrating virtual power plant technology, edge computing, digital twins and multi-time scale optimization strategies, and innovatively adopts dynamic autonomous zoning management. Through the division of flexible autonomous sub-units and optimal resource allocation, the response speed and resource utilization of distribution substations are improved. At the same time, the multi-time scale collaborative optimization framework effectively solves the problem of mismatch between long-term planning and short-term fluctuations, thereby improving the power supply quality and system resilience, and solving the problem of insufficient resilience of traditional distribution substations under voltage fluctuations, excessive harmonics, imbalance between supply and demand and extreme working conditions.

[0063] 5. The present invention provides real-time monitoring and precise decision-making support for distribution substations through the edge-cloud collaborative digital twin governance model, further enhancing the operating efficiency and reliability of the distribution network, and thus providing a new governance idea for the power system field. It has important practical application value and broad market prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0065] Figure 1 This is a principle block diagram of a distribution substation collaborative management system based on a virtual power plant according to an embodiment of the present invention;

[0066] Figure 2 This is a flow chart of a dynamic autonomous partition management unit in a distribution substation collaborative management system based on a virtual power plant according to an embodiment of the present invention when dividing a distribution substation node into several subunits;

[0067] Figure 3 This is a flow chart of a multi-time-scale collaborative optimization unit in a distribution substation collaborative governance system based on a virtual power plant according to an embodiment of the present invention when constructing a multi-time-scale collaborative mechanism at the second, minute, and hour levels;

[0068] Figure 41 is a schematic diagram of an edge-cloud collaborative decision-making unit in a distribution substation collaborative governance system based on a virtual power plant according to an embodiment of the present invention;

[0069] Figure 5 This is a flow chart of an elastic resource pool unit in a distribution substation collaborative management system based on a virtual power plant according to an embodiment of the present invention when calling resources according to priority to meet load demand;

[0070] Figure 6 It is a flowchart of a method for collaborative management of distribution substation areas based on a virtual power plant according to an embodiment of the present invention.

[0071] In the picture:

[0072] 1. Data collection unit; 2. Dynamic autonomous partition governance unit; 3. Multi-time scale collaborative optimization unit; 4. Edge-cloud collaborative decision-making unit; 5. Elastic resource pool unit; 6. Governance effect evaluation unit. DETAILED DESCRIPTION

[0073] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. By referring to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0074] According to an embodiment of the present invention, a system and method for collaborative management of distribution substations based on a virtual power plant are provided.

[0075] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to an embodiment of the present invention, a distribution substation collaborative management system based on a virtual power plant includes:

[0076] The data collection unit 1 is used to collect real-time operation data of the nodes in the distribution area and virtualize resource attributes.

[0077] It should be noted that the data collection unit 1 includes the following steps when collecting real-time operation data of the distribution area nodes and virtualizing resource attributes:

[0078] Collect real-time data and obtain measurement values ​​from all nodes in the distribution area, and the measurement values ​​include: load power P load,i (t); Photovoltaic output P PV,i (t); energy storage state of charge SOC i (t); node voltage V i (t); harmonic current hth harmonic component.

[0079] Construct virtual resource nodes: Map physical devices to dispatchable resources of the virtual power plant (VPP) and define the attributes of each virtual node i;

[0080] Adjustable power range: in, For adjustable power cap, The lower limit of adjustable power, such as minimum energy storage (discharge), maximum energy storage (Charge);

[0081] Response time: T resp,i (For example, PV inverter is 200ms and energy storage is 1s);

[0082] Adjustment cost: The adjustment cost c of the i-th virtual node resource (such as controllable load, energy storage, distributed power supply, etc.) at time t i (t), unit: yuan / kWh, grid electricity purchase cost c grid (t) Time-sharing pricing.

[0083] The dynamic autonomous partition management unit 2 is used to divide the distribution area node into several sub-units and correct the abnormal state in the sub-unit based on the power mutual assistance strategy between adjacent sub-units.

[0084] The dynamic autonomous partition management unit 2 divides the distribution area node into several sub-units and corrects the abnormal state in the sub-unit based on the power mutual assistance strategy between adjacent sub-units, including:

[0085] The improved K-means clustering algorithm is used to divide the distribution area nodes with the same similar characteristics into the same group to obtain several sub-units.

[0086] Among them, the improved K-means clustering algorithm is used to divide the distribution area nodes with the same similar characteristics into the same group, and several sub-units are obtained, including:

[0087] Based on historical operating data, the distribution substation node with the largest load fluctuation is selected as the initial cluster center, and the weight coefficients of power imbalance and energy storage regulation potential are set;

[0088] Iteratively calculate the distance between each distribution area node and the cluster center, and assign the distribution area node to the subunit with the nearest cluster center;

[0089] The power imbalance and energy storage regulation potential of each sub-unit are calculated based on the clustering results, and an objective function is constructed with the goal of minimizing the power imbalance and maximizing the energy storage regulation potential.

[0090] The objective function is solved using a local search strategy to obtain the optimal solution, the clustering results are optimized based on the optimal solution, and the iteration is stopped when the clustering results converge or the maximum number of iterations is met.

[0091] Identify local abnormal scenarios of subunits, including voltage exceeding the limit and harmonic exceeding the limit;

[0092] Triggering the autonomous control mechanism to correct abnormal scenarios, including:

[0093] Based on the voltage over-limit scenario, if the voltage of the distribution area node exceeds the preset range, the energy storage device of the sub-unit is triggered to provide voltage support;

[0094] Based on the harmonic exceeding standard scenario, if the total harmonic distortion rate of the distribution area node exceeds the preset range, the adjacent sub-units will be triggered to perform harmonic compensation.

[0095] It should be noted that, the following is a further explanation of the dynamic autonomous partition management unit 2 dividing the distribution station area node into several sub-units and correcting the abnormal state in the sub-unit based on the power mutual assistance strategy between adjacent sub-units in combination with a specific implementation method.

[0096] Step 1: Calculate partition index:

[0097] 1. Power imbalance, measuring the difference between the net load in the sub-unit and the distributed power supply;

[0098]

[0099] Where, P PV,i (t) represents the distributed power supply of the ith distribution area node at time t, P load,i (t) represents the net load of the node in the ith distribution area at time t; ΔP k (t) represents the power imbalance of subunit k, that is, the difference between the net load and the distributed generation output; C k Represents the node set contained in subunit k.

[0100] 2. Energy storage regulation potential and evaluation of the dispatchable capacity of energy storage within the sub-unit:

[0101]

[0102] Where, Indicates the rated capacity of energy storage (unit: kWh); E flex,k (t) represents the dispatchable capacity of energy storage in the sub-unit at time t; SOC i (t) represents the state of charge of the energy storage.

[0103] Step 2: Dynamic partition optimization:

[0104] Objective function: Minimize inter-regional power imbalance and maximize energy storage regulation potential

[0105]

[0106] Constraints

[0107] Where ΔP k (t) represents the power imbalance of subunit k, that is, the difference between the net load and the distributed generation output; E flex,k (t) represents the energy storage regulation potential of subunit k; λ represents the weight coefficient, which is the weight of balancing power imbalance and energy storage regulation capability; C k represents the node set contained in subunit k; C m Represents the node set contained in subunit m; N max Indicates the maximum number of nodes in a single partition.

[0108] Solution: Improved K-means clustering algorithm (the eigenvector is [ΔP k ,E flex,k ]), in order to overcome the shortcomings of the existing K-means clustering algorithm in the dynamic partitioning process, such as slow convergence speed and susceptibility to initial value influence, the present invention adopts the following improvement measures:

[0109] 1. Introducing dynamic weighting: When calculating partition indicators, the fluctuating characteristics of load and resources at different times are taken into account, and the weight coefficients of power imbalance and energy storage regulation potential are dynamically adjusted. For example, during peak load periods, the weight of power imbalance can be increased to ensure partition balance; during low load periods, the weight of energy storage regulation potential can be increased to improve resource utilization.

[0110] 2. Adaptive selection of cluster centers: Avoid randomly selecting initial cluster centers. Instead, select representative nodes as initial cluster centers based on historical or real-time data. For example, nodes with large load fluctuations can be selected as initial cluster centers to better divide load fluctuation areas.

[0111] 3. Iterative Optimization Algorithm: During the iteration process, local search strategies, such as hill climbing, are introduced to locally optimize the clustering results, improving the convergence speed and stability of the algorithm. For example, after each iteration, each node can be checked to see if it is correctly assigned to a subunit. If an assignment error is found, the node is reassigned to the subunit that is most similar to it.

[0112] Implementation steps of the improved K-means clustering algorithm:

[0113] 1. Initialization: Based on historical or real-time data, select representative nodes as initial cluster centers. Set dynamic weight coefficients, such as weighting power imbalance and energy storage regulation potential based on load fluctuation characteristics.

[0114] 2. Iterative calculation: Calculate the distance from each node to each cluster center and assign it to the subunit with the closest cluster center. Based on the clustering results, update the power imbalance and energy storage regulation potential of each subunit. Calculate the partition index of each subunit based on the dynamic weight coefficient. Use a local search strategy, such as a hill climbing algorithm, to locally optimize the clustering results.

[0115] 3. Termination condition: When the clustering results converge, that is, when the clustering results no longer change during the iteration process, terminate the iteration or set a maximum number of iterations to prevent the algorithm from falling into a local optimal solution.

[0116] Improved effect:

[0117] 1. Improve convergence speed: By introducing dynamic weights and adaptively selecting cluster centers, the convergence speed of the algorithm can be accelerated and the number of iterations can be reduced.

[0118] 2. Improve stability: By iteratively optimizing the algorithm, the stability of the algorithm can be improved and the algorithm can be prevented from falling into a local optimal solution.

[0119] 3. Improve adaptability: By dynamically adjusting the weight coefficient, it can better adapt to the fluctuating characteristics of load and resources and improve the adaptability of the partition.

[0120] Implementation example:

[0121] Assume that the distribution substation area contains 10 nodes. Each node contains information such as load power, PV output, energy storage charge state, node voltage, and harmonic current. Dynamic partitioning is performed according to the following steps:

[0122] Initialization: Based on historical data, select the node with the largest load fluctuation as the initial cluster center. Set the weight coefficient of power imbalance and energy storage regulation potential to 0.5.

[0123] Iterative calculation: Calculate the distance from each node to the cluster center and assign it to the subunit with the closest cluster center. Based on the clustering results, calculate the power imbalance and energy storage regulation potential of each subunit, and use the hill climbing algorithm to locally optimize the clustering results.

[0124] Termination condition: When the clustering results converge, that is, when the clustering results no longer change during the iteration process, the iteration is terminated.

[0125] Step 3: Trigger autonomous control:

[0126] Triggering autonomous control is a key feature of the coordinated management system for distribution substations. It allows for rapid response to localized anomalies within the substation, such as voltage limits and excessive harmonics, thereby ensuring power quality and minimizing impacts on users. By delegating control to subunits, the burden on the central controller is reduced, improving the system's responsiveness and flexibility.

[0127] Scenario 1: Voltage exceeds the limit When the voltage of a node in the distribution area exceeds the qualified range (95%-105% of the nominal voltage), the system will trigger the energy storage device of the sub-unit to provide voltage support. The energy storage of the sub-unit is called for voltage support and the power is adjusted to:

[0128]

[0129] Where, Indicates the adjustment power; K V =0.1pu / MVar (droop coefficient, in line with IEEE1547 standard); Indicates the rated capacity of energy storage (kVA); V nom =400V (nominal voltage of the station); V i represents the actual voltage value of node i at time t; V ref Indicates the voltage reference value. When the actual voltage V i Deviation from V ref When the voltage reaches the reference value, the energy storage regulation is triggered.

[0130] Scenario 2: Harmonics Exceeding Standards When the total harmonic distortion (THD) of a node in the distribution area exceeds 5%, the system will trigger the adjacent sub-units to perform harmonic compensation.

[0131] Triggering harmonic compensation of adjacent sub-units, the compensation current is:

[0132]

[0133] Where, represents the mutual impedance of nodes m and n at the hth harmonic (unit: Ω); N(m) represents the set of partitions electrically adjacent to subunit m; Indicates compensation current; represents the hth harmonic current component; N(m) represents the set of subunits electrically adjacent to subunit m (i.e., other subunits that have direct line connections to subunit m); THD k represents the total harmonic distortion rate of subunit k; I h Indicates the harmonic current content; I I Indicates the effective value of fundamental current.

[0134] It should be noted that the dynamic autonomous partition governance unit 2 also includes:

[0135] Real-time monitoring of voltage deviation rate within sub-units and harmonic distortion THD k (t), when δV k (t)>5% or THD k When (t)>5%, the power mutual assistance strategy between adjacent sub-units is automatically triggered.

[0136] The multi-time-scale collaborative optimization unit 3 is used to build a multi-time-scale collaborative mechanism at the second, minute, and hour levels to coordinate the control objectives of the distribution substation area at different time scales.

[0137] Among them, the multi-time scale collaborative optimization unit 3 includes:

[0138] Second-level control module, used to suppress voltage flicker and frequency fluctuation in the distribution area based on inverter reactive power regulation and energy storage reactive power regulation;

[0139] The minute-level scheduling module is used to optimize the operation strategy of energy storage and adjustable load in the distribution station area on a preset minute time scale with the goal of short-term economy;

[0140] The hourly planning module is used to formulate the day-ahead resource allocation strategy for the distribution substation based on the load forecast and photovoltaic forecast for the distribution substation in the future time period.

[0141] The hourly planning module formulates the day-ahead resource allocation strategy for the distribution substation based on the load forecast and PV forecast for the distribution substation in the future time period. The module includes the following:

[0142] Based on historical operating data, the long short-term memory network model is used to predict the load curve of the distribution substation in the future time period;

[0143] Based on cloud cover and irradiance data, a convolutional neural network is combined with a long short-term memory network model to predict the photovoltaic output curve of the distribution station area in the future time period.

[0144] An economic dispatch model is constructed based on the load curve and photovoltaic output curve of the distribution substation in the future time period. With the goal of minimizing the sum of the grid's electricity purchase cost and energy storage cycle loss, the economic dispatch model is solved in combination with predefined constraints to output the day-ahead resource allocation strategy for the distribution substation.

[0145] It should be noted that, the following will further illustrate, in conjunction with specific implementation methods, how the multi-time-scale collaborative optimization unit 3 constructs a multi-time-scale collaborative mechanism at the second, minute, and hour levels, and coordinates the control targets of the distribution station area at different time scales.

[0146] Among them, the multi-time scale collaborative optimization unit 3 is applicable to the following scenarios:

[0147] Distribution substations with a photovoltaic penetration rate of 10% to 80%;

[0148] The energy storage configuration capacity is 10% to 30% of the peak load of the substation;

[0149] Adjustable load ratio ≥15%.

[0150] Step 1: Second-level control (≤1 minute response): Second-level control is the fastest-responding control level in the distribution substation collaborative management system. It can quickly suppress voltage flicker and frequency fluctuations, ensure power quality, and avoid impacts on users.

[0151] Control objective: Suppress voltage flicker and frequency fluctuation.

[0152] Control strategy:

[0153] Inverter reactive power regulation: PV inverters provide reactive power support, suppressing voltage fluctuations by regulating the reactive power output of the inverter. When the node voltage is below the set value, the inverter outputs reactive power, raising the node voltage. Conversely, when the node voltage is above the set value, the inverter absorbs reactive power, lowering the node voltage.

[0154] Energy storage for reactive power regulation: Some energy storage devices, such as those using power converters, also have the ability to provide reactive power. These devices can contribute to voltage regulation by adjusting the phase of their output current to output or absorb reactive power.

[0155] Inverter reactive power regulation formula:

[0156]

[0157] Where Q PV,i (t) represents the inverter reactive power regulation at time t; Q max,i Indicates the maximum reactive capacity of the inverter (set according to manufacturer's specifications); ΔV i (t) = V i (t)-V ref ;ΔV i (t) represents the real-time voltage deviation of node i; V i (t) represents the actual voltage measurement value of node i at time t; V ref Indicates the voltage reference value (target voltage); sgn(ΔV i (t)) represents the sign function of voltage deviation, which is used to determine the output direction of reactive power; V th Indicates the voltage regulation threshold, specifically 0.05pu (voltage regulation threshold); Indicates the calculated voltage deviation rate.

[0158] Energy storage reactive power regulation: Energy storage devices with reactive power regulation capabilities can output or absorb reactive power according to system requirements. The specific regulation method is similar to that of inverters.

[0159] Step 2: Minute-level scheduling (5-minute cycle): On a 5-minute time scale, with economy as the goal, optimize the operating strategies of energy storage and adjustable loads to reduce the grid's electricity purchase costs and energy storage cycle losses.

[0160] Rolling optimization model: Starting from the current moment, it optimizes the scheduling of energy storage and load in the next hour.

[0161] Objective function: Minimize the sum of grid electricity purchase cost and energy storage cycle loss.

[0162] The constraints are: dynamic change constraints of energy storage state of charge (SOC), grid power balance constraints, and line capacity limitations. Their expressions include:

[0163]

[0164] constraint

[0165] Where, α = 0.7, β = 0.3 (weight coefficients, determined by sensitivity analysis); η ch =0.95,η dis =0.90 (charge and discharge efficiency, refer to lithium battery characteristics); T = 12 (optimization window is 1 hour, Δt = 5 minutes); P grid (τ) represents the actual grid power; represents the planned grid power; SOC(τ) represents the state of charge of the energy storage device at time τ; P load,i (τ) represents the power demand of load i at time τ; P PV,i (τ) represents the actual output of photovoltaic equipment i at time τ; P ESS,i (τ) represents the charge and discharge power of energy storage i at time τ; Δt represents the scheduling time step; Indicates the rated capacity of the energy storage device; Indicates the lower limit of the discharge power of the energy storage device.

[0166] Step 3: Hourly planning (24-hour cycle):

[0167] Prediction Model:

[0168] Load Forecasting: An LSTM neural network model is used to predict the load profile for the next 24 hours based on historical load data, temperature, date type, and other information. The Long Short-Term Memory (LSTM) network is a special type of recurrent neural network (RNN) that effectively processes sequential data and captures long-term dependencies within time series. LSTM performs well in load forecasting because it learns how load changes over time and predicts future load trends. The LSTM model accurately predicts load profiles and captures load trends, providing reliable load forecast data for distribution substation operations.

[0169] Photovoltaic forecast: A CNN-LSTM hybrid neural network model is used to predict the PV output curve for the next 24 hours based on information such as cloud cover and irradiance. Convolutional neural networks (CNNs) effectively extract image features and apply them to time series forecasting. The CNN-LSTM hybrid model combines the advantages of CNN and LSTM to simultaneously extract both spatial and temporal features of PV output, thereby improving forecast accuracy. The CNN-LSTM model more accurately predicts PV output curves and captures the temporal and spatial variations of PV output, providing more reliable PV forecast data for distribution station operations.

[0170] Economic dispatch model:

[0171]

[0172] constraint

[0173] Where, Indicates the predicted output of the photovoltaic equipment; Indicates the predicted power of the load; C grid (t) represents the time-of-use electricity price of the power grid at time t; P grid (t) represents the actual grid power at time t; C ESS Indicates the rated capacity of the energy storage device; Indicates the upper limit of the maximum allowable charging power of energy storage; Indicates the upper limit of the maximum allowable discharge power of energy storage; P PV (t) represents the output of photovoltaic equipment; P ESS (t) represents the actual charging and discharging power of the energy storage at time t; L j represents the set of nodes connected by line j; P i (t) represents the power of node i at time t (kW); S j represents the maximum transmission capacity of line j (kVA); P load (t) represents the load power.

[0174] Objective function: Minimize the sum of grid electricity purchase cost and energy storage cycle loss.

[0175] Constraints: Condition 1 represents the power balance relationship of the distribution substation at time t, that is, the sum of the power purchased by the grid, the photovoltaic output, and the energy storage output must equal the load demand of the substation; Condition 2 indicates that the sum of the power of all nodes on line j at time t cannot exceed the maximum transmission capacity of line j.

[0176] The edge-cloud collaborative decision-making unit 4 is used to complement the edge-side real-time control strategy with the cloud-side global strategy to achieve maintenance of edge devices.

[0177] The edge-cloud collaborative decision-making unit 4 complements the edge-side real-time control strategy with the cloud-side global strategy to achieve maintenance of edge devices, including:

[0178] Simplify the dimensions of the system matrix, input matrix, and output matrix, construct state space equations to perform lightweight calculations on edge devices, and output state estimates and control instructions for edge devices;

[0179] Among them, the system matrix is ​​used to analyze the natural evolution of the internal state of the subunit;

[0180] The input matrix is ​​used to control the effect of the state space equation input on the state variables;

[0181] The output matrix is ​​used to analyze the mapping relationship between state variables and observable outputs;

[0182] Obtain the historical operating data of edge devices, and comprehensively calculate the health of edge devices in combination with the state estimation and control instructions of edge devices, and generate maintenance strategies for edge devices based on the health calculation results.

[0183] It should be noted that the specific embodiments described below further illustrate how the edge-cloud collaborative decision-making unit 4 complements the edge-side real-time control strategy with the cloud-side global strategy to achieve maintenance of edge devices.

[0184] Step 1: Edge side lightweight control:

[0185] This system enables real-time control and rapid response at the edge, using simplified state-space equations for lightweight computation to meet the real-time requirements of edge devices. Outputs include edge device state estimates (such as voltage, current, and SOC) and control commands (such as charge and discharge commands for energy storage devices and reactive power regulation commands for photovoltaic inverters).

[0186] Model: Simplified state space equation (update period 1 second). The lightweight model is simplified by A k 、B k 、C kThe dimension of the state space is reduced (for example, only key state variables are retained), the computational complexity is reduced, and the real-time requirements of edge devices are met. The state space equation is:

[0187]

[0188] Where A k represents the system matrix; B k represents the input matrix; C k represents the output matrix; x k represents the state vector; u k represents the control input; y k (t) represents the observable output vector of subunit k at time t.

[0189] in:

[0190] x k =[V k ,I k ,SOC k ] T ;

[0191] u k =[P ESS,k ,Q PV,k ,] T ;

[0192] A k It is the system matrix, which describes the natural evolution law of the internal state of the subunit, that is, the dynamic change of the system state without external control.

[0193] Specific meaning:

[0194] Main diagonal elements: self-attenuation or inertia of state variables (such as natural voltage drop due to line impedance and self-discharge of stored energy).

[0195] Non-diagonal elements: coupling relationships between state variables (such as the impact of voltage changes on current, and the feedback of SOC changes on voltage).

[0196] Example:

[0197]

[0198] A 11 =0.98: The voltage decreases by 2% per hour due to line losses.

[0199] A 12 =-0.02: Coupling coefficient where the voltage drops due to an increase in current.

[0200] A 33 =0.99: Energy storage SOC self-discharges 1% per hour.

[0201] Bk is the input matrix, which represents the direct effect of the control input on the state variables, that is, how the adjustment command changes the system state.

[0202] Physical meaning: It represents the direct impact of the control input on the state variable, that is, how the adjustment instruction changes the system state.

[0203] Specific meaning:

[0204] Each element B ij : Gain of the j-th control input on the i-th state variable.

[0205] Example:

[0206]

[0207] B 11 =0.1: For every 1kW increase in energy storage power, the voltage increases by 0.1V.

[0208] B 22 =0.05: For every 1kVar increase in PV reactive power, the current decreases by 0.05A.

[0209] B 31 =-0.2: 1kW discharge of energy storage results in a 0.2% decrease in SOC.

[0210] C k It is the output matrix that defines the mapping relationship between state variables and observable outputs, that is, which state variables are actually measured or used for control.

[0211] Specific meaning:

[0212] The main diagonal elements are 1, which means the corresponding state (such as voltage, SOC) is directly observed.

[0213] Other elements may represent indirect observations (such as inferring power from current).

[0214] Example:

[0215]

[0216] First row: Direct output voltage V k .

[0217] Second row: Directly output energy storage SOC k , current I k Not observed.

[0218] Step 2: Cloud health prediction:

[0219] This enables global decision-making and long-term planning in the cloud. By analyzing historical operating data of edge devices, it predicts the health and remaining life of the devices and generates maintenance strategies based on the predictions. Inputs include historical operating data (such as temperature, current, and voltage) and estimated state data of edge devices.

[0220] Health calculation:

[0221]

[0222] Where H i (t) represents the health of the edge device; S i,k represents the kth health indicator of equipment i (such as transformer oil temperature, cable insulation resistance); Indicates the baseline value of the health indicator of the equipment when it leaves the factory; w k represents the weight coefficient of the kth health indicator, which is determined using the weight entropy method; σ(z) represents the Sigmoid activation function (health correction gate), with a value range of [0,1], which maps any input to a probability interval. When the device ages (z→+∞), σ(z)→1, amplifying the health attenuation effect; S i,k (t) represents the actual value of the kth health indicator of device i at time t.

[0223] Maintenance strategy: If H i (t)<0.7, generate maintenance instructions and adjust the priority of elastic resources.

[0224] It should be noted that the edge-cloud collaborative decision-making unit 4 also includes: a health prediction model based on the historical operation data of the equipment to calculate the remaining life of the equipment L i (t) = L new ·e -α·Hi(t) , where H i (t) represents the health index, α represents the aging rate coefficient, e represents the base of the exponential function (≈2.718), L new Indicates the design life of the equipment when it is new. i (t) <L th Generate early warning signals.

[0225] The elastic resource pool unit 5 is used to calculate the elastic contribution of resources and set priorities, and call resources according to the priorities to meet the load demand.

[0226] The elastic resource pool unit 5 calculates the elastic contribution of resources and sets priorities, and calls resources according to the priorities to meet the load demand, including:

[0227] An elastic resource pool is constructed based on energy storage and adjustable load resources, the elastic contribution of resources in the elastic resource pool is calculated, and the elastic contribution of resources is sorted from high to low to obtain the priority order.

[0228] Detect the connection status between the distribution substation and the main grid. If the distribution substation is disconnected from the main grid and in tinkering mode, call the island mode resources in order of priority to meet the critical load requirements;

[0229] Construct constraints for resource power limits, energy storage state of charge safety ranges, resource response time constraints, and line mutual aid capacity limits.

[0230] The power balance after calling resources is verified based on the constraint conditions. If the verification fails, the remaining resources are re-called until the preset conditions are met.

[0231] It should be noted that, in conjunction with specific implementations, the elastic resource pool unit 5 will be further described below in calculating the elastic contribution of resources, setting priorities, and calling resources according to priorities to meet load requirements.

[0232] Elasticity contribution assessment:

[0233] Calculation formula:

[0234]

[0235] Where, E flex,i represents the adjustable energy of resource i; f(SOC i ,N cycle,i ) represents the aging factor; Indicates regulatory potential; Indicates economy; T resp,i represents the response time of resource i (unit: seconds); α and β represent weight coefficients; λ represents the penalty coefficient, η i (t) represents the resource elasticity contribution of resource i at time t, where:

[0236]

[0237] f(SOC i ,N cycle,i )=α·(1-SOC i )+β·N cycle,i ;

[0238] Where, f(SOC i ,N cycle,i ) represents the aging factor; represents the minimum allowed power lower limit of resource i; represents the maximum allowable power limit of resource i; Ncycle,i represents the historical cumulative number of charge and discharge cycles of the i-th energy storage device.

[0239] Step 2: Calling resources in island mode:

[0240] Trigger condition: Detection of disconnection between the substation and the main grid (V grid = 0 and lasts for 10 seconds)

[0241] Calling strategy:

[0242] ① Contribution calculation:

[0243] According to the elastic contribution index η i (t) Sort resources from high to low:

[0244]

[0245] Where η i (t) represents the elastic contribution; E i Indicates adjustable energy (such as energy storage capacity); c i Indicates the adjustment cost (yuan / kWh); Age i =f(SOC i ,N cycle,i ) represents the aging factor (λ = 0.1 is the penalty coefficient); γ represents the photovoltaic resource fluctuation compensation coefficient.

[0246] ② Priority sorting:

[0247] Press η i (t) Sort resources from high to low (e.g., energy storage > adjustable load > photovoltaic).

[0248] ③ Resource call:

[0249] Call resources in sequence until the following conditions are met:

[0250]

[0251] Where, L crit,j (t) represents the critical load set such as hospitals and communication base stations.

[0252] ④Power balance verification:

[0253] The verification conditions must meet the following constraints:

[0254]

[0255] If verification fails, return to step ③ and call the remaining resources again.

[0256] It should be noted that the sandbox test environment in the elastic resource pool unit 5 includes:

[0257] The state equation of the digital twin of the substation is expressed as dx / dt=Ax+Bu+w;

[0258] The fault injection interface supports simulation of abnormal operating conditions such as voltage sag (ΔV∈[-0.3,0]pu) and harmonic injection (THD∈[5%,20%]).

[0259] In addition, the present invention further includes a treatment effect evaluation unit 6, which generates a verification report containing the following indicators based on the treatment effect evaluation unit 6:

[0260] Voltage qualification rate:

[0261]

[0262] Eligible scope:

[0263] 0.95V nom ≤V(t)≤1.05V nom ;

[0264] Where V nom =400V.

[0265] Resource utilization:

[0266]

[0267] Where, Indicates the actual output energy of the resource; Indicates the theoretical maximum output energy at the rated capacity of the resource.

[0268] Economic indicators:

[0269]

[0270] Where, Indicates the time-based electricity price of the power grid; represents the energy storage cycle loss cost; c grid (t) represents the time-of-use electricity price of the power grid; λ i Indicates the energy storage aging coefficient (refer to battery manufacturer data).

[0271] Harmonic control rate:

[0272]

[0273] In the formula, the calculation basis is the total harmonic distortion (THD), h≥2 is the harmonic component; I h (t) represents the hth harmonic current component (unit: A); Indicates the total distortion energy of harmonic components (h≥2); Indicates the total current energy including the fundamental wave (h=1).

[0274] Harmonic control rate:

[0275]

[0276] Where, THD comp,i Indicates the harmonic compensation capability of device i.

[0277] According to another embodiment of the present invention, Figure 6 As shown, a collaborative governance method for distribution substations based on a virtual power plant is also provided, which includes:

[0278] S1, collects real-time operation data of nodes in the distribution area and virtualizes resource attributes;

[0279] S2. Divide the distribution area node into several sub-units and correct the abnormal state in the sub-unit based on the power mutual assistance strategy between adjacent sub-units;

[0280] S3. Build a multi-time-scale coordination mechanism at the second, minute, and hour levels to coordinate the control objectives of the distribution substation area at different time scales;

[0281] S4. Complement edge-side real-time control with cloud-based global policies to achieve maintenance of edge devices;

[0282] S5. Calculate the elastic contribution of resources and set priorities, and call resources according to the priorities to meet the load demand.

[0283] In summary, if Figure 1 As shown, the system encompasses six major components: data collection, dynamic partitioning, multi-timescale optimization, edge-cloud collaboration, elastic resource scheduling, and governance evaluation. Data collection unit 1 acquires real-time operational data from substation nodes and virtualizes resource attributes, providing basic input for subsequent modules. Dynamic autonomous partitioning governance unit 2 divides autonomous subunits through cluster analysis, triggering mutual assistance strategies to address local anomalies. Multi-timescale collaborative optimization unit 3 achieves layered, second-level rapid response, minute-level economic scheduling, and hour-level predictive planning. Edge-cloud collaborative decision-making unit 4 complements global cloud-based policy generation through real-time edge-side control. Elastic resource pool unit 5 prioritizes resource allocation to ensure resilience. Governance effectiveness evaluation unit 6 quantifies system performance. These components are connected through data flows, forming a closed loop of "perception-analysis-decision-execution-verification." The feedback mechanism of edge-cloud collaborative decision-making unit 4 dynamically optimizes data collection and partitioning strategies, enhancing system adaptability.

[0284] like Figure 2As shown in the figure, a five-step serial process for dynamic partitioning is presented. First, real-time data from the substation (such as load, photovoltaic output, voltage, etc.) is collected. After standardization, the power imbalance and energy storage regulation potential are analyzed through an improved clustering algorithm to divide the optimized subunits. The operating status of the subunits is then monitored. If the voltage deviation or harmonics exceed the standard, the power mutual assistance strategy between adjacent subunits is triggered. The key logic is that the mutual assistance result in the fifth step will reversely correct the priority of data collection (such as increasing the sampling frequency in abnormal areas), forming a closed-loop iterative optimization to ensure that the partitioning strategy is dynamically adjusted with the substation status.

[0285] like Figure 3 As shown, this demonstrates the synergistic relationship between second-, minute-, and hour-level optimization. Second-level control suppresses voltage flicker with a second-level response; minute-level scheduling optimizes the short-term economic efficiency of energy storage and load every five minutes; and hour-level planning formulates a day-ahead resource allocation strategy based on load and PV forecasts. These three levels are interconnected through a bidirectional data flow: second-level real-time data provides input for upper-level optimization, while hourly forecasts constrain minute-level scheduling targets, forming a closed-loop optimization loop across timescales that balances rapid response with long-term planning.

[0286] like Figure 4 The figure below depicts the division of labor and collaboration between the edge and cloud. On the edge, a lightweight state estimator (updates local device status in real time) and a local control actuator (responds quickly to commands) run in parallel to ensure low-latency control. The cloud first simulates complex scenarios using a high-precision simulation engine, then performs health prediction (assessing device lifespan) and policy generation (developing a global optimization plan) in parallel. Edge data upload drives cloud analysis, and cloud-generated policies are distributed to the edge for execution. This creates an efficient "local perception-cloud decision-making-local execution" collaboration, leveraging cloud computing power while preserving edge real-time performance.

[0287] like Figure 5 As shown, the system first detects a grid disconnect and confirms the island state. It then calculates the contribution of resources like energy storage and adjustable loads, prioritizes them, and deploys them sequentially to meet the needs of critical loads like hospitals and base stations. After deployment, it verifies power balance. If not, it redeploys remaining resources until the conditions are met. This conditional loop ensures reliable and efficient resource scheduling, avoiding over-deployment and resource waste. Contribution ranking quantifies resource value, ensuring the resilience of power supply to critical loads.

[0288] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The distribution area collaborative management system based on virtual power plants is characterized by: The system includes: Data collection unit, used to collect real-time operation data of nodes in the distribution area and virtualize resource attributes; Dynamic autonomous partition management unit, used to divide the distribution area node into several sub-units and correct abnormal conditions within the sub-unit based on the power mutual assistance strategy between adjacent sub-units; Multi-time scale collaborative optimization unit, used to build multi-time scale collaborative mechanisms at the second, minute, and hour levels, and coordinate the control objectives of distribution substations at different time scales; Edge-cloud collaborative decision-making unit, which complements edge-side real-time control strategies with cloud-based global strategies to achieve maintenance of edge devices; The elastic resource pool unit is used to calculate the elastic contribution of resources and set priorities, and call resources according to the priorities to meet load requirements.

2. The distribution substation collaborative management system based on virtual power plant according to claim 1 is characterized in that: The dynamic autonomous partition management unit divides the distribution area node into a plurality of sub-units and corrects the abnormal state in the sub-unit based on the power mutual assistance strategy between adjacent sub-units, including: The improved K-means clustering algorithm is used to divide the distribution area nodes with the same similar characteristics into the same group to obtain several sub-units; Identify local abnormal scenarios of the subunit, where the local abnormal scenarios include voltage exceeding limit scenarios and harmonic exceeding standard scenarios; Triggering the autonomous control mechanism to correct abnormal scenarios, including: Based on the voltage over-limit scenario, if the voltage of the distribution area node exceeds the preset range, the energy storage device of the sub-unit is triggered to provide voltage support; Based on the harmonic exceeding standard scenario, if the total harmonic distortion rate of the distribution area node exceeds the preset range, the adjacent sub-units will be triggered to perform harmonic compensation.

3. The distribution substation collaborative management system based on virtual power plant according to claim 2 is characterized in that: The improved K-means clustering algorithm is used to divide the distribution area nodes with the same similar characteristics into the same group, and several sub-units are obtained, including: Based on historical operating data, the distribution substation node with the largest load fluctuation is selected as the initial cluster center, and the weight coefficients of power imbalance and energy storage regulation potential are set; Iteratively calculate the distance between each distribution area node and the cluster center, and assign the distribution area node to the subunit with the nearest cluster center; The power imbalance and energy storage regulation potential of each sub-unit are calculated based on the clustering results, and an objective function is constructed with the goal of minimizing the power imbalance and maximizing the energy storage regulation potential. The objective function is solved using a local search strategy to obtain the optimal solution, the clustering results are optimized based on the optimal solution, and the iteration is stopped when the clustering results converge or the maximum number of iterations is met.

4. The distribution substation collaborative management system based on virtual power plant according to claim 1 is characterized in that: The multi-timescale collaborative optimization unit includes: Second-level control module, used to suppress voltage flicker and frequency fluctuation in the distribution area based on inverter reactive power regulation and energy storage reactive power regulation; The minute-level scheduling module is used to optimize the operation strategy of energy storage and adjustable load in the distribution station area on a preset minute time scale with the goal of short-term economy; The hourly planning module is used to formulate the day-ahead resource allocation strategy for the distribution substation based on the load forecast and photovoltaic forecast for the distribution substation in the future time period.

5. The distribution substation collaborative management system based on virtual power plant according to claim 4 is characterized in that: The hourly planning model includes the following steps when formulating a day-ahead resource allocation strategy for a distribution substation based on the load forecast and PV forecast for the distribution substation in the future time period: Based on historical operating data, the long short-term memory network model is used to predict the load curve of the distribution substation in the future time period; Based on cloud cover and irradiance data, a convolutional neural network is combined with a long short-term memory network model to predict the photovoltaic output curve of the distribution station area in the future time period. An economic dispatch model is constructed based on the load curve and photovoltaic output curve of the distribution substation in the future time period. With the goal of minimizing the sum of the grid's electricity purchase cost and energy storage cycle loss, the economic dispatch model is solved in combination with predefined constraints to output the day-ahead resource allocation strategy for the distribution substation.

6. The distribution substation collaborative management system based on virtual power plant according to claim 5 is characterized in that: The expression of the economic dispatch model includes: Where, P grid (t) represents the actual grid power at time t; L j represents the set of nodes connected by line j; P i (t) represents the power of node i at time t; S j represents the maximum transmission capacity of line j; C grid (t) represents the time-of-use electricity price of the power grid at time t; C ESS Indicates the rated capacity of the energy storage device; Indicates the upper limit of the maximum allowable charging power of energy storage; Indicates the upper limit of the maximum allowable discharge power of energy storage; P PV (t) represents the output of photovoltaic equipment; P ESS (t) represents the actual charging and discharging power of the energy storage at time t; P load (t) represents the load power.

7. The distribution substation collaborative management system based on virtual power plant according to claim 1 is characterized in that: The edge-cloud collaborative decision-making unit includes the following steps when complementing the edge-side real-time control strategy with the cloud-side global strategy to achieve maintenance of edge devices: Simplify the dimensions of the system matrix, input matrix, and output matrix, construct state space equations to perform lightweight calculations on edge devices, and output state estimates and control instructions for edge devices; Wherein, the system matrix is ​​used to analyze the natural evolution law of the internal state of the subunit; The input matrix is ​​used to control the influence of the state space equation input on the state variables; The output matrix is ​​used to analyze the mapping relationship from state variables to observable outputs; Obtain the historical operating data of edge devices, and comprehensively calculate the health of edge devices in combination with the state estimation and control instructions of edge devices, and generate maintenance strategies for edge devices based on the health calculation results.

8. The distribution substation collaborative management system based on virtual power plant according to claim 1 is characterized in that: The elastic resource pool unit calculates the elastic contribution of resources and sets priorities, and calls resources according to the priorities to meet the load demand, including: Build an elastic resource pool based on energy storage and adjustable load resources, calculate the elastic contribution of resources in the elastic resource pool, and sort the elastic contribution of resources from high to low to obtain the priority order; Detect the connection status between the distribution substation and the main grid. If the distribution substation is disconnected from the main grid and in tinkering mode, call the island mode resources in order of priority to meet the critical load requirements; Construct constraints for resource power limits, energy storage state of charge safety ranges, resource response time constraints, and line mutual aid capacity limits. The power balance after calling resources is verified based on the constraint conditions. If the verification fails, the remaining resources are re-called until the preset conditions are met.

9. The distribution substation collaborative management system based on virtual power plant according to claim 8 is characterized in that: The calculation formula of the resource elasticity contribution is: Where η i (t) represents the resource elasticity contribution of resource i at time t; E flex,i represents the adjustable energy of resource i; Indicates regulatory potential; Indicates economy; T resp,i represents the response time of resource i; α and β are weight coefficients; λ represents the penalty coefficient; f(SOC i ,N cycle,i ) represents the aging factor.

10. A method for collaborative management of distribution substations based on a virtual power plant, using a collaborative management system for distribution substations based on a virtual power plant according to any one of claims 1 to 9, characterized in that: The method includes: Collect real-time operating data of nodes in the distribution area and virtualize resource attributes; The distribution area node is divided into several sub-units, and the abnormal state in the sub-unit is corrected based on the power mutual assistance strategy between adjacent sub-units; Build a multi-time scale coordination mechanism at the second, minute, and hour levels to coordinate the control targets of distribution substations at different time scales; Complement edge-side real-time control with cloud-based global policies to achieve maintenance of edge devices; Calculate the elastic contribution of resources and set priorities, and call resources according to the priorities to meet load requirements.

Images