Multi-scene aggregated optimization scheduling method and system for multi-agent differentiated demand

Abstract

The disclosure provides a multi-scene aggregation optimization scheduling method and system for multi-agent differentiated needs, relates to the technical field of power system optimization scheduling, and comprises the following steps: uniformly modeling multiple types of distributed resources according to the regulation and control characteristics of virtual power plants on resources; analyzing differentiated regulation and control requirements from three dimensions of power grid scheduling side, aggregator side and user side, and establishing a multi-level multi-time scale demand model; based on the multi-time scale demand model and the distributed resource model, a hierarchical multi-scene aggregation optimization model is constructed to perform multi-scene aggregation optimization scheduling; and the disclosure realizes comprehensive characterization of multi-agent multi-time scale demand and improves the grid regulation and control capability of the virtual power plant.

Description

Technical Field

[0001] This disclosure relates to the field of power system optimization and dispatching technology, specifically to a multi-scenario aggregated optimization and dispatching method and system for multi-entity differentiated needs. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] Virtual power plants, as an innovative power system operation organization model, effectively aggregate geographically dispersed distributed energy resources into a unified dispatchable and manageable whole through advanced information and communication technologies and energy management systems. However, the diverse resource types and significant differences in the response characteristics of various resources within a virtual power plant, coupled with the interplay of interests among the grid dispatching side, aggregator side, and user side, present numerous challenges to the optimal dispatch of virtual power plants.

[0004] Existing scheduling optimization methods include multi-energy virtual power plant optimization scheduling methods based on random forest algorithms and artificial neural networks, virtual power plant optimization scheduling methods based on multi-timescale collaborative decision-making, and decision-making methods based on a two-layer master-slave game structure between distribution network operators and virtual power plants. However, these existing methods still have the following limitations: (1) The characteristics of various types of resources are very different, and there is a lack of a unified and refined classification modeling framework based on the dimension of "regulation characteristics". The key constraints such as the dynamics of energy storage SOC, the randomness of electric vehicle travel, and the equivalent thermodynamic process of temperature control load are not adequately characterized. (2) The different needs of multiple entities such as the power grid dispatcher, aggregator and user are not fully considered. The modeling is mostly based on a single objective function or a master-slave game between two types of entities, and the coordination optimization faces the game dilemma. (3) The layering logic dimension is singular, or it is only vertically layered according to the "time scale granularity", or it is only layered according to the master-slave game, without horizontal aggregation layering according to the "physical characteristics of application scenarios" such as industrial parks, commercial buildings, residential communities, and electric vehicle charging stations; (4) The scheduling needs of multi-timescale markets (day-ahead, real-time, auxiliary services, deviation assessment) are intertwined, the coordination mechanism is imperfect, and the market participation channels are limited; (5) Traditional centralized optimization methods have privacy leakage risks and communication bottlenecks, and their distributed collaborative optimization capabilities are insufficient; (6) There is a lack of a fair and reasonable profit distribution mechanism that reflects the marginal contributions of each entity. Summary of the Invention

[0005] To address the aforementioned issues, this disclosure proposes a multi-scenario aggregation optimization scheduling method and system oriented towards the differentiated needs of multiple stakeholders. By establishing a unified and refined modeling framework for two types of resources—adjustable and semi-adjustable—it constructs a multi-stakeholder differentiated regulation demand model from multiple dimensions, builds a hierarchical aggregation optimization model covering four typical scenarios, and employs the ADMM distributed optimization algorithm to achieve parallel solution under privacy protection for each scenario. This effectively solves key issues in the multi-scenario aggregation optimization scheduling of virtual power plants, such as the uniformity of resource modeling, multi-stakeholder coordination, privacy protection, and allocation fairness, thereby improving the operational efficiency of virtual power plants and the grid regulation capability.

[0006] According to some embodiments, the present disclosure adopts the following technical solutions: Multi-scenario aggregation and optimization scheduling methods for addressing the differentiated needs of multiple stakeholders include: Based on the resource regulation characteristics of virtual power plants, a unified model is constructed for multiple types of distributed resources. We analyze differentiated regulation demand from three dimensions: power grid dispatching, aggregator, and user, and establish a multi-level, multi-time-scale demand model. Based on a multi-timescale demand model and a distributed resource model, a hierarchical multi-scenario aggregation optimization model is constructed to perform multi-scenario aggregation optimization scheduling. The hierarchical multi-scenario aggregation optimization model has an upper layer that is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategy for each scenario. The lower layer is the optimization layer within different scenarios, which uses the alternating direction multiplier method to achieve parallel solution under privacy protection for each scenario, and uses the Shapley value method to fairly allocate the total revenue of the virtual power plant obtained by the solution to multiple subjects.

[0007] As one embodiment, the unified modeling of multiple types of distributed resources based on the resource regulation characteristics of virtual power plants includes: Based on the resource regulation characteristics of virtual power plants, aggregated distributed resources are divided into two categories: adjustable resources and semi-adjustable resources, and refined mathematical models are established for each category. The adjustable resources include energy storage systems, controllable loads, and vehicle-to-grid (V2G) electric vehicles, which can flexibly adjust their output or power consumption according to dispatch instructions; the semi-adjustable resources include distributed photovoltaic resources and wind power resources, whose output mainly depends on natural conditions and can only be adjusted for wind and solar curtailment within a limited range.

[0008] As one embodiment, the analysis of differentiated regulation demand from three dimensions—grid dispatching, aggregator, and user—and the establishment of a multi-layered, multi-time-scale demand model includes: The power grid dispatching side focuses on the safe and stable operation of the system, requiring virtual power plants to provide reliable regulation services with fast response speed and small execution deviation; the aggregator side pursues the maximization of operational efficiency and hopes to obtain the maximum benefits by participating in multiple markets; the user side expects to obtain reasonable returns while ensuring electricity demand and comfort. Establish a multi-timescale demand model that includes four aspects: day-ahead market revenue, real-time market deviation settlement, ancillary service market revenue, and deviation assessment constraints.

[0009] As one embodiment, the hierarchical multi-scenario aggregation optimization model adopts a hierarchical architecture design, forming an architecture of "4 types of physical scenarios and 1 operator entity"; The upper layer is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategies for each scenario. The lower layer is the internal optimization layer for different scenarios, including industrial park scenarios, commercial building scenarios, residential community scenarios, and electric vehicle charging station scenarios. The operator acts as the coordination hub between the upper and lower layers, and each scenario only needs to report the aggregated power response capability and marginal cost to the upper layer.

[0010] As one embodiment, the method of alternating direction multipliers is used to achieve parallel solution under privacy protection in various scenarios. The solution iteration format includes: Step 1: Update local variables; Step 2: Update global variables; Step 3: Lagrange multiplier update; Step 4: Algorithm convergence. The convergence criteria include the original residual convergence criterion and the dual residual convergence criterion. The algorithm converges when both the original residual and the dual residual satisfy the conditions.

[0011] As one embodiment, the method of using the Shapley value method to fairly distribute the total revenue of the virtual power plant obtained through multi-stakeholder allocation includes: Industrial parks, commercial buildings, residential communities, electric vehicle charging stations, and virtual power plant operators are considered as five game participants. All possible orders in which each participant can join the alliance are enumerated, the marginal contribution of each participant in each order is calculated, and the final distribution of revenue for each participant is obtained by weighting the distribution of revenue according to the weight coefficients.

[0012] As one example, the Shapley value method satisfies four fairness criteria: effectiveness, symmetry, virtuality, and additivity, ensuring fair and reasonable distribution of benefits and incentivizing all stakeholders to actively participate in the long-term operation of the virtual power plant. The specific calculation is as follows:

[0013] in, For the first i Shapley values ​​for each participating entity;N For the set of all subjects; S Not included i any subset of; For set S The number of elements; The number of all subjects; For the alliance S The value function, i.e., without i Alliance benefits upon participation; for i Benefits of joining the alliance; for i The marginal contribution; for i The number of ways the main body can be arranged; for i The number of ways the main body can be arranged; The total number of permutations of all subjects; The weighting coefficients are, i.e. i There was exactly one before it when it was added. The probability of this situation occurring for an individual subject.

[0014] As one example, the upper layer is the overall optimization layer of the virtual power plant, with the objective of maximizing the total revenue of the virtual power plant. The objective function is:

[0015] in, Total revenue for the virtual power plant; For the day's market trading profits; The benefits or costs of deviation assessment; For ancillary service revenue; Operating costs.

[0016] According to some embodiments, the present disclosure adopts the following technical solutions: A multi-scenario aggregation and optimization scheduling system for addressing the differentiated needs of multiple stakeholders includes: The resource modeling module is used to uniformly model multiple types of distributed resources based on the resource regulation characteristics of virtual power plants. The demand analysis module is used to analyze differentiated regulation demand from three dimensions: the power grid dispatching side, the aggregator side, and the user side, and to establish a multi-level, multi-time-scale demand model. The aggregation optimization module is used to construct a hierarchical multi-scenario aggregation optimization model based on multi-timescale demand models and distributed resource models to perform multi-scenario aggregation optimization scheduling. The hierarchical multi-scenario aggregation optimization model has an upper layer that is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategy for each scenario. The lower layer is the optimization layer within different scenarios, which uses the alternating direction multiplier method to achieve parallel solution under privacy protection for each scenario, and uses the Shapley value method to fairly allocate the total revenue of the virtual power plant obtained by the solution to multiple subjects.

[0017] According to some embodiments, the present disclosure adopts the following technical solutions: A computer program product includes a computer program that, when executed by a processor, implements the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs.

[0018] According to some embodiments, the present disclosure adopts the following technical solutions: A non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs.

[0019] According to some embodiments, the present disclosure adopts the following technical solutions: An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs.

[0020] Compared with the prior art, the beneficial effects of this disclosure are as follows: This disclosed multi-scenario aggregation optimization scheduling method, designed for the differentiated needs of multiple stakeholders, establishes a unified and refined modeling framework for various types of distributed resources. Distributed resources are scientifically categorized into two main types: adjustable and semi-adjustable, encompassing five resource classes: distributed photovoltaic, wind power, energy storage systems, electric vehicle clusters, and controllable loads. The output characteristics, operational constraints, and cost functions of each resource class are meticulously characterized. The model is standardized in form, highly parameterized, and easily expandable and integrated with new resource types.

[0021] This disclosed multi-scenario aggregation optimization scheduling method for multi-entity differentiated needs achieves a comprehensive characterization of multi-entity and multi-timescale needs. It establishes differentiated regulation demand models from three dimensions: grid dispatching, aggregator, and user, covering four levels: day-ahead market, real-time market, ancillary service market, and deviation assessment, comprehensively reflecting the operational scenarios and constraints of virtual power plants in a diversified electricity market.

[0022] This disclosed multi-scenario aggregation optimization scheduling method, designed for the differentiated needs of multiple stakeholders, constructs a hierarchical multi-scenario aggregation optimization model that balances overall optimization with scenario autonomy. The upper layer implements unified scheduling, while each scenario at the lower layer performs autonomous optimization based on its own physical characteristics. This approach protects the operational privacy of each scenario while achieving optimal overall returns. Four scenarios—industrial parks, commercial buildings, residential communities, and electric vehicle charging stations—play differentiated regulatory roles at different times, forming temporal complementarity and jointly meeting the grid's regulation needs.

[0023] This disclosed multi-scenario aggregation optimization scheduling method for the differentiated needs of multiple entities is based on the ADMM distributed optimization algorithm. Each entity does not need to expose its internal operating data, but only needs to exchange aggregation power information, effectively protecting commercial privacy. Based on the linear convergence property of ADMM and combined with the reasonable setting of the penalty parameter in this method, the algorithm is expected to converge within about 30 iterations, and the optimization gap can be controlled at about 0.5%~1.0%.

[0024] The disclosed multi-scenario aggregation optimization scheduling method for the differentiated needs of multiple entities adopts the Shapley value method to allocate alliance benefits according to the marginal contribution of each entity, satisfying the four fairness criteria of effectiveness, symmetry, virtuality and additivity. Compared with the average allocation or capacity allocation method, it significantly improves the fairness of allocation and incentive effect, which is conducive to the long-term stable participation of multiple entities.

[0025] The disclosed multi-scenario aggregation optimization scheduling method for the differentiated needs of multiple stakeholders demonstrates strong practicality through expected performance analysis. In a typical virtual power plant daily operation scenario, based on the model mechanism and typical parameter values ​​established by this method, the expected tracking error RMSE of the regulation command can be controlled within the range of approximately 0.3~0.4 MW. The daily operation revenue structure is expected to present a reasonable distribution: approximately 46% for electricity trading revenue, approximately 29% for peak-shaving services, approximately 16% for frequency regulation services, and approximately 9% for reserve services. The allocation results based on Shapley values ​​are expected to show a structure of approximately 35% for industrial parks, approximately 24% for commercial buildings, approximately 16% for residential communities, approximately 13% for charging stations, and approximately 12% for operators. This objectively reflects the marginal contribution of each stakeholder and demonstrates significantly better fairness than traditional average allocation or capacity-based allocation methods. Attached Figure Description

[0026] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0027] Figure 1 This is a flowchart of a multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs, as described in an embodiment of this disclosure. Figure 2This is a schematic diagram of the distributed resource classification system for virtual power plants according to an embodiment of this disclosure; Figure 3 This is a hierarchical multi-scenario aggregation optimization architecture as described in this disclosure embodiment; Figure 4 This is a flowchart of the ADMM distributed optimization solution process according to an embodiment of the present disclosure; Figure 5 This is a schematic diagram of multi-entity revenue distribution based on Shapley values, according to an embodiment of this disclosure. Detailed Implementation

[0028] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0029] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0030] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0031] Example 1 One embodiment of this disclosure provides a multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs, the method steps of which include: Step 1: Based on the resource regulation characteristics of the virtual power plant, perform unified modeling for multiple types of distributed resources; Step 2: Analyze differentiated regulation demand from three dimensions: power grid dispatching side, aggregator side, and user side, and establish a multi-layered, multi-time-scale demand model; Step 3: Based on the multi-timescale demand model and the distributed resource model, construct a hierarchical multi-scenario aggregation optimization model to perform multi-scenario aggregation optimization scheduling; The hierarchical multi-scenario aggregation optimization model has an upper layer that is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategy for each scenario. The lower layer is the optimization layer within different scenarios, which uses the alternating direction multiplier method to achieve parallel solution under privacy protection for each scenario, and uses the Shapley value method to fairly allocate the total revenue of the virtual power plant obtained by the solution to multiple subjects.

[0032] As one embodiment, this disclosure presents a multi-scenario aggregation optimization scheduling method for multi-entity differentiated needs. By establishing a unified and refined modeling framework for adjustable / semi-adjustable resources, it constructs a multi-entity differentiated adjustment demand model from three dimensions: the grid dispatching side, the aggregator side, and the user side. It builds a hierarchical aggregation optimization model covering four typical scenarios: industrial parks, commercial buildings, residential communities, and electric vehicle charging stations. The ADMM distributed optimization algorithm is used to achieve parallel solution under privacy protection for each scenario, and the Shapley value method is used to achieve fair distribution of benefits among multiple entities. This effectively solves key problems in the multi-scenario aggregation optimization scheduling of virtual power plants, such as the uniformity of resource modeling, multi-entity coordination, privacy protection, and allocation fairness, thereby improving the operational efficiency of virtual power plants and the grid regulation capability. The specific implementation process is as follows: S1: Unified modeling of multiple types of distributed resources; Based on the resource regulation characteristics of virtual power plants, aggregated distributed resources are divided into two categories: adjustable resources and semi-adjustable resources, and refined mathematical models are established for each. Adjustable resources include energy storage system resources, controllable load resources, and electric vehicle V2G resources, which can flexibly adjust output or power consumption according to dispatch instructions. Semi-adjustable resources include distributed photovoltaic resources and wind power resources, whose output mainly depends on natural conditions and can only be regulated within a limited range. The specific modeling process is as follows: (1) Distributed photovoltaic resource modeling The output of distributed photovoltaic (PV) power generation systems exhibits significant diurnal periodicity and seasonality, while also showing strong volatility due to random factors such as cloud cover. The basic output model for PV modules is as follows: (1) in, for t The actual output power of the photovoltaic system at any given time, in MW; Rated power of photovoltaic modules under standard test conditions (STC), in MW; for t Solar irradiance at any given time, expressed in W / m². The irradiance is taken as 1000 W / m² under standard test conditions. The power temperature coefficient is typically taken as -0.35% / ℃ to -0.50% / ℃, and in this disclosure it is taken as -0.45% / ℃; for t Photovoltaic panel temperature at any given time, in °C; The standard test temperature is 25℃. The inverter conversion efficiency is typically taken as 0.95 to 0.98.

[0033] The relationship between photovoltaic panel temperature and ambient temperature and solar irradiance is as follows: (2) in, for t Ambient temperature at any given time, in °C; NOCT The nominal operating cell temperature is typically set between 45°C and 47°C.

[0034] Considering the uncertainty of photovoltaic power output, the K-means clustering method is used to extract typical power output scenarios, and the Wasserstein distance index is used to optimally discretize the probability distribution.

[0035] Furthermore, the extracted typical power output scenarios are represented by several power output curves, each with a corresponding probability of occurrence. The specific process includes: The original data consists of a large amount of historical daily photovoltaic power output data. The power output process for each day (considering 96 points) can be discretized into a 96-dimensional vector. After normalization using the installed capacity, the vector corresponding to each day is a sample point in the feature space.

[0036] The clustering process is implemented, using K-means to cluster hundreds or thousands of historical curves into K classes (5 classes in the example), and grouping the sunrise power curves with similar shapes together.

[0037] The specific output is a typical power output scenario. The cluster center of each category is a typical power output scenario curve. Physically, it roughly corresponds to different weather conditions. For example: (1) Sunny day, noon peak, full and symmetrical curve; (2) Cloudy day, low overall power output, accompanied by random fluctuations; (3) Overcast day, very low power output throughout the day, flat curve; (4) Sunny in the morning, overcast in the afternoon, asymmetrical curve; (5) Overcast in the morning, sunny in the afternoon, asymmetrical curve. The probability of each scenario is equal to the historical number of days that the scenario fell into that cluster divided by the total number of days. The sum of the probabilities is 1.

[0038] The purpose of this method is to transform the original continuous uncertainty into a combination of K representatives and their corresponding probabilities, turning stochastic optimization into a deterministic problem of calculating the expected value of five scenarios with weights.

[0039] (2) Wind power resource modeling The relationship between the fan output power and the wind speed is described by a piecewise function: (3) in, For wind speed v The output power of the wind turbine, in MW; Rated power of the wind turbine, in MW; vWind speed, in m / s; The cut-in wind speed is the minimum wind speed at which the wind turbine starts generating electricity, typically 3 to 4 m / s. The cut-off wind speed, which is the wind speed at which the fan is stopped to protect it, is generally 20~25 m / s; Rated wind speed, which is the wind speed at which the fan reaches its rated power, is generally 11~15 m / s.

[0040] The statistical characteristics of wind speed are represented by the Weibull distribution: (4) in, Let be the probability density function of wind speed; k This is a shape parameter, typically taken as 1.5 to 3.0; c This is a scale parameter, related to the average wind speed.

[0041] (3) The dynamic evolution equation of the SOC of the energy storage system is: (5) in, for t The state of charge at time t, with a value range of [0,1]; for t Charging power at any given moment, measured in MW; for t Discharge power at any given moment, in MW; For charging efficiency, a value of 0.92 to 0.96 is generally used; For optimal discharge efficiency, a value of 0.92 to 0.96 is typically used. Rated capacity of the energy storage system, in MWh; The time step is expressed in hours (h).

[0042] Energy storage system operating constraints include: 1) Upper and lower limits constraints for energy storage: (6) 2) Charge and discharge power constraints: (7) (8) 3) Charge-discharge mutual exclusion constraint: (9) 4) Daily cycle constraint: (10) in, and These are the lower and upper limits of SOC, respectively; and These are the maximum charging power and the maximum discharging power, respectively. and These are binary variables of 0 and 1 representing the charging and discharging states, respectively; T is the duration of a scheduling cycle (e.g., 24 hours). The allowable SOC deviation is generally taken as 0.05~0.10.

[0043] (4) Electric vehicle cluster resource modeling Probabilistic modeling of electric vehicle arrival and departure times: (11) (12) in, and These are the mean and standard deviation of the arrival times, respectively. and These represent the mean and standard deviation of departure time, respectively. Based on typical commuting patterns, private cars generally arrive at charging stations after get off work (…). ≈18:00, ≈1.5h), leaving before work the next day ( ≈7:30, ≈1.0h).

[0044] The initial SOC follows a truncated normal distribution: (13) Wherein, TN represents a truncated normal distribution; and These are the mean and standard deviation of the initial SOC, respectively, with a typical value being... =40%, =15%; and To cut off the boundary.

[0045] Charging demand model: (14) in, Charging requirements, in kWh; SOC at arrival The target SOC at the time of expected departure is generally set at 80% to 100%. Battery capacity, in kWh.

[0046] Electric vehicle cluster aggregated power model: (15) in, for tAggregated power of the electric vehicle cluster at any given time; For the first i The charging and discharging power of an electric vehicle (charging is positive, V2G discharging is negative). As an access status indicator variable, when the first i electric vehicles t When connected to the charging station ,otherwise .

[0047] (5) Controllable load resource modeling Controllable loads are classified into three categories: transferable loads, reduceable loads, and temperature-controlled loads.

[0048] 1) Transferable load constraints: (16) (17) (18) in, Let be the power of the load that can be transferred at time t; To meet the total electricity demand, it must be completed within the specified time window; and These represent the start and end times of the time window, respectively. This is for maximum power limitation.

[0049] 2) Load constraints can be reduced: (19) (20) (twenty one) in, for t Power reduction at any moment; For maximum power reduction; The 0-1 variables represent the state of reduction; and These are the minimum and maximum durations of a single response, respectively. The duration of a single continuous response; This is the maximum cumulative daily power reduction, used to protect users' production and operations.

[0050] 3) Temperature-controlled load ETP model: (twenty two) (twenty three) (twenty four) in, fort Indoor temperature at any given time, in °C; for t Outdoor temperature at any time, in °C; R is building thermal resistance, in °C / kW; C The building's heat capacity is expressed in kWh / ℃; τ = R × C is the thermal time constant. Air conditioner power, in kW; COP is the coefficient of performance for cooling / heating. To set the temperature; This is the temperature dead zone, generally taken as 1~2℃, within which users feel comfortable.

[0051] S2: Analysis of Differentiated Adjustment Needs of Multiple Subjects; This paper analyzes differentiated regulation demands from three dimensions: grid dispatching, aggregator, and user. The grid dispatching side focuses on the safe and stable operation of the system, requiring virtual power plants to provide reliable regulation services with fast response times and minimal execution deviations. Aggregators aim to maximize operational efficiency, hoping to obtain maximum revenue through participation in multiple markets. Users expect to obtain reasonable returns while ensuring electricity demand and comfort. A multi-timescale demand model covering the following four levels is established: (1) Day-ahead market return model: (25) in, This represents the market earnings and gross profit as of the current day. for t The day-ahead market electricity price at any given time, in yuan / MWh; for t The reported output value at any given time indicates that electricity is sold, while a negative value indicates that electricity is purchased.

[0052] (2) Real-time market deviation settlement model: (26) (27) in, for t Power deviation at any given moment; To contribute practically; This is the deviation settlement price, i.e., the real-time market clearing electricity price.

[0053] (3) Revenue model for ancillary services market: (28) in, , , They are respectively t Compensation prices for peak shaving, frequency regulation, and backup services; , , These represent the corresponding service capacities.

[0054] (4) Deviation assessment constraints: (29) in, The allowable deviation percentage is generally ±3% to 5%.

[0055] S3: Construct a hierarchical multi-scenario aggregation optimization model; The hierarchical multi-scenario aggregation optimization model adopts a hierarchical architecture design, forming an architecture of "4 types of physical scenarios + 1 operator entity".

[0056] (a) Upper-level optimization model The upper layer is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategies for each scenario. The objective function is: (30) in, Total revenue for the virtual power plant; For the day's market trading profits; The benefits or costs of deviation assessment; For ancillary service revenue; Operating costs.

[0057] The power balance constraint is: (31) in, For the first i One distributed power source output; For the first j Power consumption of each load; This refers to the energy storage charging and discharging power (discharging is positive, charging is negative). The charging and discharging power of the electric vehicle cluster (charging is positive, V2G discharging is negative). This represents the power exchanged with the power grid (negative for purchasing electricity, positive for selling electricity).

[0058] (ii) Lower-level optimization model The lower layer consists of internal optimization layers for different scenarios, including industrial park scenarios, commercial building scenarios, residential community scenarios, and electric vehicle charging station scenarios. The operator acts as the coordination hub between the upper and lower layers, and each scenario only needs to report the aggregated power response capability and marginal cost to the upper layer.

[0059] (1) Industrial park scene The objective is to minimize the combined costs of electricity and demand response compensation. (32) in, Electricity costs for the industrial park; for t Electricity price at any time; for t Real-time power consumption; To compensate for revenue in response to demand.

[0060] (2) Commercial building scene The goal is to minimize overall energy costs: (33) in, The overall energy cost for commercial buildings; for t Electricity price at any time; for t Total power consumption of the building at any time; For the operating costs of the air conditioning system; The operating cost of the lighting system is constrained by indoor temperature comfort. And lighting brightness constraints.

[0061] (3) Residential community scene The objective is to minimize overall cost: (34) in, The overall cost for residential communities; for t Total electricity consumption in the community at any given time; The penalty coefficient for user dissatisfaction; for t Indicators of user dissatisfaction caused by load adjustments at all times; For revenue from surplus electricity generated by residential solar PV systems and fed into the grid, a participation rate parameter is introduced. Depicting differences in user response intentions This indicates that the user is not involved in the adjustment process at all. This indicates that the user is fully cooperating with the adjustment instructions.

[0062] (4) Electric vehicle charging station scenario The goal is to minimize operating costs: (35) in, The operating costs of charging stations; For the first i electric vehicles t The charging power at any given moment; This refers to the V2G discharge power. The electricity price is used to compensate for V2G discharge. Battery degradation costs are related to the depth and frequency of charge / discharge. Off-site SOC constraints. Ensure that each vehicle reaches the user's expected SOC level when leaving, and ensure that the total power constraint of the charging station does not exceed the transformer capacity limit.

[0063] (5) Operator role As the coordination hub between the upper and lower layers, operators create value through market bidding strategies and task allocation adjustments across various scenarios. Their revenue sources include market transaction price differences, ancillary service revenue sharing, and deviation management revenue. Each scenario only needs to report aggregated power response capabilities and marginal costs to the upper layer, without exposing detailed internal equipment-level operational data.

[0064] S4: Distributed Optimization Solution The Alternating Direction Multiplier Method (ADMM) is used to solve the distributed optimization problem. The iterative format of ADMM is as follows: (1) Local variable update (parallel solution for each entity): (36) (2) Global variable update (coordination center solution): (37) (3) Lagrange multiplier update: (38) in, For the first i z represents the local decision-making variables of each subject; z represents the global coordination variable. These are Lagrange multiplier vectors; This is the penalty parameter, typically ranging from 0.1 to 10; For the first i Local objective functions for each main scene; The objective function is the global objective function. A , B , c Constraint matrices and vectors; k This represents the number of iterations.

[0065] (4) Convergence criterion: 1) Original residual convergence criterion: , (39) 2) Dual residual convergence criterion: , (40) The algorithm converges when both the original residual and the dual residual satisfy the above conditions. and This is the preset convergence threshold.

[0066] S5: Using the Shapley value method for fair allocation among multiple subjects; Industrial parks, commercial buildings, residential communities, electric vehicle charging stations, and virtual power plant operators are treated as five game participants. The Shapley value method is used to fairly distribute the total revenue of the virtual power plant alliance. The Shapley value is calculated by enumerating all possible order in which each entity can join the alliance, calculating its marginal contribution in each order, and then weighting the contributions by a coefficient to obtain the final revenue allocation for each entity. This method satisfies four fairness criteria: efficiency, symmetry, virtuality, and additivity, ensuring a fair and reasonable distribution of revenue and incentivizing all entities to actively participate in the long-term operation of the virtual power plant.

[0067] (41) in, For the first i Shapley value (distribution of income) for each participating entity. N For the set of all subjects; S Not included i any subset of; For set S The number of elements; The number of all subjects; For the alliance S The value function, i.e., without i Alliance benefits upon participation; for i Benefits of joining the alliance; for i The marginal contribution; for i The number of ways the main body can be arranged; for i The number of ways the main body can be arranged; The total number of possible arrangements of all subjects; The weighting coefficients are, i.e. i There was exactly one before it when it was added. The probability of this situation occurring for an individual subject.

[0068] The Shapley value satisfies four fairness criteria: validity, symmetry, virtuality, and additivity, ensuring a fair and reasonable distribution of benefits.

[0069] Example 2 One embodiment of this disclosure provides a multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs. The implementation method and expected effects of the technical solution of this disclosure are explained in conjunction with a typical application scenario.

[0070] (1) Implementation scenario setting This paper analyzes the expected effects of a typical daily operation scenario of a virtual power plant in a certain region. The virtual power plant aggregates the following resources: approximately 8MW of distributed photovoltaic capacity and approximately 5MW of distributed wind power capacity; an energy storage system with a rated power of 3MW and a rated capacity of 9MWh, with a charge / discharge efficiency of 0.95 and an initial SOC value of 0.5; an electric vehicle charging station containing approximately 200 electric vehicles with an average battery capacity of 60kWh; and controllable loads including interruptible loads in industrial parks, air conditioning loads in commercial buildings, and transferable loads in residential communities.

[0071] (2) Implementation process Step 1: Establish refined models and calibrate parameters for photovoltaic, wind power, energy storage, electric vehicles, and controllable loads respectively, following step S1 of Example 1. Among them, photovoltaic output is generated into 5 typical scenarios using K-means clustering, and wind power is sampled using Weibull distribution parameters k=2.0 and c=8.5.

[0072] Step 2: Construct a multi-entity differentiated demand model according to step S2, and input the day-ahead market 24-point electricity price curve (peak price 1100 yuan / MWh, flat price 680 yuan / MWh, valley price 320 yuan / MWh) and the compensation prices for three types of ancillary services: peak shaving, frequency regulation, and standby.

[0073] Step 3: Construct a hierarchical multi-scenario aggregation optimization model according to step S3. The upper-level objective is to maximize the total revenue of the virtual power plant, and the lower-level four types of scenarios establish internal optimization models and set scenario boundary constraints respectively.

[0074] Step 4: Following step S4, use the ADMM algorithm for distributed solution, with the penalty parameter ρ set to 1.0, and the convergence thresholds for both the original residual and the dual residual set to 10. - ³.

[0075] Step 5: Following step S5, the Shapley value method is used to fairly distribute the total revenue of the virtual power plant obtained from the solution to the five categories of entities.

[0076] (3) Analysis of expected results Expected Scenario Regulation Contribution: Based on the resource model established by this method and the electricity consumption characteristics of each scenario, it is expected that industrial parks can contribute about 3 to 4.5 MW of regulation output during morning and evening peak hours, commercial buildings can contribute about 2 to 3 MW during the day, residential communities can contribute about 1.5 to 2.5 MW at night, and electric vehicle charging stations can contribute about 2 to 3 MW at noon and night. The four types of scenarios will form significant temporal complementarity at different times, thereby forming a continuous aggregate regulation capability throughout the entire time period.

[0077] Expected tracking accuracy: Based on the ADMM distributed coordination mechanism and refined modeling of each resource adopted in this method, the tracking error RMSE between the actual response output of the virtual power plant and the grid regulation command is expected to be controlled within the range of approximately 0.3~0.4MW, which can meet the deviation assessment requirements.

[0078] Expected economic benefits: Under the above typical market environment and parameter values, the expected daily operating revenue structure is expected to show the following distribution: electricity trading revenue accounts for about 46%, peak shaving service revenue accounts for about 29%, frequency regulation service revenue accounts for about 16%, and standby service revenue accounts for about 9%. After deducting operating costs, the expected net income to total gross income ratio is about 67%, which reflects the potential advantage of this method in multi-market synergistic benefits.

[0079] Expected allocation results: Based on theoretical calculations of the marginal contributions of each entity, the final allocation results based on Shapley values ​​are expected to present the following structure: industrial parks receive approximately 35%, commercial buildings approximately 24%, residential communities approximately 16%, charging stations approximately 13%, and operators approximately 12%. This allocation structure matches the adjustable resource scale of each entity and its contribution to the alliance's revenue.

[0080] Expected solution efficiency: Based on the linear convergence property of the ADMM algorithm and combined with the reasonable setting of the penalty parameter in this method, the algorithm is expected to converge within about 30 iterations, and the optimization gap can be controlled at about 0.5%~1.0%. Compared with centralized solution, this invention has expected improvements in privacy protection, scalability and solution efficiency.

[0081] Example 3 One embodiment of this disclosure provides a multi-scenario aggregation optimization scheduling system for multi-subject differentiated needs, including: The resource modeling module is used to uniformly model multiple types of distributed resources based on the resource regulation characteristics of virtual power plants. The demand analysis module is used to analyze differentiated regulation demand from three dimensions: the power grid dispatching side, the aggregator side, and the user side, and to establish a multi-level, multi-time-scale demand model. The aggregation optimization module is used to construct a hierarchical multi-scenario aggregation optimization model based on multi-timescale demand models and distributed resource models to perform multi-scenario aggregation optimization scheduling. The hierarchical multi-scenario aggregation optimization model has an upper layer that is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategy for each scenario. The lower layer is the optimization layer within different scenarios, which uses the alternating direction multiplier method to achieve parallel solution under privacy protection for each scenario, and uses the Shapley value method to fairly allocate the total revenue of the virtual power plant obtained by the solution to multiple subjects.

[0082] Example 4 One embodiment of this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs.

[0083] Example 5 One embodiment of this disclosure provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, they implement the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs.

[0084] Example 6 One embodiment of this disclosure provides an electronic device, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs.

[0085] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0086] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0087] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A multi-scenario aggregation optimization scheduling method for multi-entity differentiated needs, characterized in that, include: Based on the resource regulation characteristics of virtual power plants, a unified model is constructed for multiple types of distributed resources. We analyze differentiated regulation demand from three dimensions: power grid dispatching, aggregator, and user, and establish a multi-level, multi-time-scale demand model. Based on a multi-timescale demand model and a distributed resource model, a hierarchical multi-scenario aggregation optimization model is constructed to perform multi-scenario aggregation optimization scheduling. Among them, the upper layer of the hierarchical multi-scenario aggregation optimization model is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategy for each scenario. The lower layer is an internal optimization layer for different scenarios. The alternating direction multiplier method is used to achieve parallel solution under privacy protection for each scenario, and the Shapley value method is used to fairly distribute the total revenue of the virtual power plant obtained by the solution to multiple parties.

2. The multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in claim 1, characterized in that, The method of unified modeling of multiple types of distributed resources based on the resource regulation characteristics of virtual power plants includes: Based on the resource regulation characteristics of virtual power plants, aggregated distributed resources are divided into two categories: adjustable resources and semi-adjustable resources, and refined mathematical models are established for each category. The adjustable resources include energy storage systems, controllable loads, and vehicle-to-grid (V2G) electric vehicles, which can flexibly adjust their output or power consumption according to dispatch instructions; the semi-adjustable resources include distributed photovoltaic resources and wind power resources, whose output mainly depends on natural conditions and can only be adjusted for wind and solar curtailment within a limited range.

3. The multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in claim 1, characterized in that, The analysis of differentiated regulation demand from three dimensions—grid dispatching, aggregator, and user—establishes a multi-layered, multi-time-scale demand model, including: The power grid dispatching side focuses on the safe and stable operation of the system, requiring virtual power plants to provide reliable regulation services with fast response speed and small execution deviation; the aggregator side pursues the maximization of operational efficiency and hopes to obtain the maximum benefits by participating in multiple markets; the user side expects to obtain reasonable returns while ensuring electricity demand and comfort. Establish a multi-timescale demand model that includes four aspects: day-ahead market revenue, real-time market deviation settlement, ancillary service market revenue, and deviation assessment constraints.

4. The multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in claim 1, characterized in that, The hierarchical multi-scenario aggregation optimization model adopts a hierarchical architecture design, forming an architecture of "4 types of physical scenarios and 1 operator entity"; The upper layer is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategies for each scenario. The lower layer consists of internal optimization layers for different scenarios, including industrial park scenarios, commercial building scenarios, residential community scenarios, and electric vehicle charging station scenarios. The operator acts as the coordination hub between the upper and lower layers, and each scenario only needs to report the aggregated power response capability and marginal cost to the upper layer.

5. The multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in claim 1, characterized in that, The method of alternating direction multipliers is used to achieve parallel solution under privacy protection in various scenarios. The solution iteration format includes: Step 1: Update local variables; Step 2: Update global variables; Step 3: Lagrange multiplier update; Step 4: Algorithm convergence. The convergence criteria include the original residual convergence criterion and the dual residual convergence criterion. The algorithm converges when both the original residual and the dual residual satisfy the conditions.

6. The multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in claim 1, characterized in that, The method of using the Shapley value method to fairly distribute the total revenue of the virtual power plant obtained through multi-stakeholder allocation includes: Industrial parks, commercial buildings, residential communities, electric vehicle charging stations, and virtual power plant operators are considered as five game participants. All possible orders in which each participant can join the alliance are enumerated, the marginal contribution of each participant in each order is calculated, and the final distribution of revenue for each participant is obtained by weighting the distribution of revenue according to the weight coefficients.

7. The multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in claim 6, characterized in that, The Shapley value method satisfies four fairness criteria: effectiveness, symmetry, virtuality, and additivity. This ensures fair and reasonable revenue distribution and incentivizes all stakeholders to actively participate in the long-term operation of the virtual power plant. The specific calculation is as follows: in, For the first i Shapley values ​​for each participating entity; N For the set of all subjects; S Not included i any subset of; For set S The number of elements; The number of all subjects; For the alliance S The value function, i.e., without i Alliance benefits upon participation; for i Benefits of joining the alliance; for i The marginal contribution; for i The number of ways the main body can be arranged; for i The number of ways the main body can be arranged; The total number of permutations of all subjects; The weighting coefficients are, i.e. i There was exactly one before it when it was added. The probability of this situation occurring for an individual subject.

8. The multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in claim 4, characterized in that, The upper layer is the overall optimization layer of the virtual power plant, with the objective of maximizing the total revenue of the virtual power plant. The objective function is: in, Total revenue for the virtual power plant; For the day's market trading profits; The benefits or costs of deviation assessment; For ancillary service revenue; Operating costs.

9. A multi-scenario aggregation and optimization scheduling system for multi-entity differentiated needs, characterized in that: include: The resource modeling module is used to uniformly model multiple types of distributed resources based on the resource regulation characteristics of virtual power plants. The demand analysis module is used to analyze differentiated regulation demand from three dimensions: the power grid dispatching side, the aggregator side, and the user side, and to establish a multi-level, multi-time-scale demand model. The aggregation optimization module is used to construct a hierarchical multi-scenario aggregation optimization model based on multi-timescale demand models and distributed resource models to perform multi-scenario aggregation optimization scheduling. Among them, the upper layer of the hierarchical multi-scenario aggregation optimization model is the overall optimization layer of the virtual power plant, which aims to maximize the total revenue of the virtual power plant and determines the power allocation and market bidding strategy for each scenario. The lower layer is an internal optimization layer for different scenarios. The alternating direction multiplier method is used to achieve parallel solution under privacy protection for each scenario, and the Shapley value method is used to fairly distribute the total revenue of the virtual power plant obtained by the solution to multiple parties.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in any one of claims 1-8.

11. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in any one of claims 1-8.

12. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the multi-scenario aggregation optimization scheduling method for multi-subject differentiated needs as described in any one of claims 1-8.

Images