Virtual power plant aggregation cost quantification method considering multivariate demand response fuzzy dynamic willingness

Abstract

The invention provides a virtual power plant aggregation cost quantification method considering multiple demand response fuzzy dynamic willingness, and belongs to the field of distributed resource aggregation regulation and control. Firstly, a constructed flexible load user dynamic response willingness model quantifies the participation level of flexible loads such as an electric vehicle and an air conditioner cluster under multi-factor driving. And secondly, the constructed wind power / photovoltaic typical scene set can be used for risk quantification considering uncertainty in a virtual power plant aggregation cost calculation risk model. Thirdly, a constructed virtual power plant aggregation cost calculation risk model objective function comprehensively considers resource calling cost, network loss, adjustment deviation penalty and system tail risk loss; and finally, establishing constraint conditions of the virtual power plant aggregation cost calculation risk model, and realizing evaluation of schedulable margins of an electric vehicle cluster and an air conditioner cluster. According to the method, credible input can be provided for the adjustable resource boundary of the virtual power plant, and quantitative representation of the relationship between the aggregation main body adjustment service amount and the cost is realized.

Description

Technical Field

[0001] The present invention belongs to the field of distributed resource aggregation control, and relates to a virtual power plant aggregation cost quantification method taking into account multiple demand response fuzzy dynamic willingness. Background Art

[0002] The traditional energy model is facing challenges. my country is accelerating the construction of a new power system, which is characterized by a high proportion of renewable energy access. Improving system flexibility and regulation capabilities has become one of the core challenges of current power dispatching and operation. In recent years, the proportion of new energy output such as wind power and photovoltaic power has continued to rise, but its intermittent and volatile nature poses a significant threat to the stable operation of the power grid; at the same time, traditional thermal power units are gradually withdrawing from the peak-shaving market, and the gap in standby regulation capabilities is becoming increasingly apparent. In this context, tapping and activating the peak-shaving potential of distributed resources has become an important path to building a flexible, safe and efficient power system.

[0003] Among distributed energy resources (DER), flexible loads represented by electric vehicles, air conditioning loads, and industrial loads play an increasingly important role in improving the flexibility of power systems due to their certain regulation capabilities, wide distribution, and controllability. In recent years, with the popularization of smart appliances and the increase in the penetration rate of electric vehicles, the number of flexible load resources in my country has grown rapidly, with considerable potential for regulation participation. However, due to their small response granularity, strong heterogeneity of user behavior, and frequent dynamic changes in operating status, it is difficult to achieve effective participation in direct regulation and peak-shaving services under the traditional power market system. Therefore, it is necessary to use aggregation mechanisms to achieve resource integration and improve their regulatory value to the power grid. Virtual Power Plant (VPP) is the key technical path proposed under this demand. Virtual Power Plant virtualizes multiple distributed resources through cyber-physical systems, and realizes unified modeling, optimized scheduling, and market interaction without changing the original physical structure.

[0004] Currently, virtual power plant technology has been widely applied in pilot projects in Europe, North America and some demonstration areas in my country. Represented by the three-level control architecture, virtual power plants not only have the ability to access multi-source heterogeneous resources, but also realize aggregate scheduling linkage in multiple stages such as day-ahead, intraday and real-time. In terms of operation objectives, virtual power plant research focuses on cost optimization, revenue maximization and power balance, but there are still practical challenges such as insufficient modeling of user behavior and imperfect risk control mechanisms.

[0005] Especially under the trend of widespread access of flexible loads to the virtual power plant system, existing research has carried out modeling work on the operation optimization of virtual power plants in multiple scenarios such as ancillary services, day-ahead market bidding, peak shaving and valley filling. Although the above research has considered aspects such as source-load coordination and economic objectives, the modeling of user behavior is often simplified, and the subjective response differences of flexible loads cannot be fully captured. How to accurately evaluate its actual adjustable capacity has become the key to the credibility and implementation effect of the model. As a typical user-side resource, the response decision of flexible loads is significantly affected by subjective factors such as price signals, comfort, and preferences. If it is regarded as a deterministic regulation resource and incorporated into the virtual power plant model, it may lead to dispatching execution deviations, affecting the cost expectation and response stability of the virtual power plant's participation in the market; on the other hand, in addition to the uncertainty driven by user behavior, the output fluctuations of distributed renewable energy itself also constitute an important external risk source for the implementation of the virtual power plant plan. Affected by meteorological conditions, the wind and light power has strong temporal correlation and uncertainty. If these uncertainty sources are ignored in the virtual power plant dispatching process, the dispatching plan will lack robustness, and the cost control will face the problem of rising tail risk.

[0006] At present, there is an urgent need to construct a method for quantifying the aggregation cost of virtual power plants that takes into account the fuzzy dynamic willingness of multiple demand responses. This method can quantify the participation levels of flexible loads such as electric vehicles and air-conditioning clusters under the drive of multiple factors, and establish a risk decision model for the aggregation cost of virtual power plants that includes call costs, network losses, deviation penalty values, and risk indicators, so as to realize the quantitative characterization of the relationship between the regulation service volume and cost of the aggregation entity. Summary of the Invention

[0007] Aiming at the problems of the existing technology, the present invention provides a method for quantifying the aggregation cost of virtual power plants that takes into account the fuzzy dynamic willingness of multiple demand responses. This method constructs a modeling framework for the response willingness of flexible loads, and uses the TSK fuzzy inference system to depict the behavior of electric vehicle and air-conditioning users; secondly, the present invention introduces the CVaR risk measurement method and constructs a risk model for calculating the aggregation cost of virtual power plants that integrates response willingness, uncertainty, network constraints, and risk control.

[0008] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0009] A method for quantifying the aggregation cost of virtual power plants that takes into account the fuzzy dynamic willingness of multiple demand responses, comprising the following steps:

[0010] S1. Construct a dynamic response willingness model for flexible load users, including the following steps:

[0011] S1.1 In the process of the virtual power plant providing peak shaving auxiliary services, the response behavior of flexible loads is not completely determined by system control parameters, but is significantly affected by subjective factors such as user willingness, comfort level, and incentive level. Therefore, its response ability has significant uncertainty and individual differences. If it is directly incorporated into the virtual power plant aggregation scheduling model as a deterministic resource, it may lead to actual execution deviations, affecting the regulation effect and cost expectation of the virtual power plant. Therefore, it is necessary to construct a dynamic response willingness model for flexible load users and couple it with the aggregation scheduling process.

[0012] The present invention uses the TSK fuzzy inference system proposed by Takagi, Sugeno, and Kang to model the response willingness of flexible load users and constructs a dynamic response willingness model for flexible load users. This TSK fuzzy inference system has strong non-linear modeling ability and good interpretability, and is suitable for dealing with behavior modeling problems with multiple inputs, fuzzy cognition, and continuous output quantities. Compared with traditional fuzzy systems, the TSK fuzzy inference system has more advantages in terms of clear structure and computational efficiency, and is convenient to be embedded into the scheduling optimization framework for linkage calculation.

[0013] The TSK fuzzy inference system consists of multiple fuzzy rules, and each fuzzy rule adopts the general expression form such as the k-th rule R k as follows:

[0014]

[0015] where x1, x2,..., x n are input variables, and the input variables can be price, electricity demand, temperature difference, etc. represents the fuzzy set corresponding to the n-th input variable in the k-th rule, and y k is the output function of the k-th rule, generally a linear combination of inputs.

[0016] For a given input x = (x1, x2,..., x n ), the membership degree of each fuzzy rule is:

[0017]

[0018] Finally, the output λ of the dynamic response willingness model of flexible load users is the weighted average of the outputs of each fuzzy rule as the response willingness:

[0019]

[0020] where K is the total number of rules, and y k (x) is the function of rule k.

[0021] In the present invention, the response willingness λ ∈ [0, 1] of flexible load users represents the degree of their participation in load regulation under current incentive conditions. By setting appropriate input variables and fuzzy rules, the response levels of group or individual users can be dynamically estimated and used to construct dynamic scheduling constraints or adjustable margin ranges.

[0022] S1.2 Based on the TSK fuzzy inference system and the flexible load user dynamic response willingness model established in S1.1, a flexible load response willingness model driven by multiple factors is established. Affected by multiple factor interactions, such as incentive electricity price levels, environmental conditions, and user behavior preferences, etc. Such decision-making behaviors have significant non-linear and fuzzy characteristics and are difficult to characterize by traditional linear modeling methods. Therefore, in the framework of the TSK fuzzy inference system, the present invention constructs a flexible load response willingness model driven by multiple factors to estimate the subjective participation tendency of flexible load users and provide a basis for subsequent adjustable capacity evaluation and scheduling decisions.

[0023] For two types of typical flexible loads, electric vehicles and air conditioners, key factor sets affecting their response willingness are constructed respectively:

[0024] For electric vehicle users, input variables are considered: State of Charge (SOC); incentive price level (Price). After normalization of the input variables, they are fuzzily mapped to variables: "LOW", "MID", "HIGH", and corresponding triangular membership functions are set for description.

[0025] For air conditioner users, input variables are considered: State of Charge (analogizing the air conditioner cluster as a virtual energy storage); incentive electricity price level (Price). After normalization of the input variables, they are fuzzily mapped to variables: "LOW", "MID", "HIGH", and corresponding triangular membership functions are set for description.

[0026] In the fuzzy rule layer, an "IF-THEN" type rule set is established through expert experience and behavior analysis, as shown in formula (4):

[0027]

[0028] The output function of each rule adopts a linear form and is written as:

[0029]

[0030] In the formula, f k (x) is the output function of the k-th rule; a k is the linear coefficient vector of the input variables; b k is the rule bias term; x is the vector composed of input variables.

[0031] The response willingness of flexible load users is calculated according to formulas (2)-(3). In the actual demand response process, there must be a certain degree of uncertainty in the response willingness. The triangular membership function is used to describe the uncertainty degree of the response willingness of flexible load users, and the fuzzy expression of the response willingness is as follows:

[0032]

[0033] In the formula: is the fuzzy expression of the response willingness of flexible load users; λ1, λ2, and λ3 are the triangular membership parameters of the response willingness, r1 and r3 are proportionality coefficients, and 0 < r1 < 1, r3 > 1. λ1 is the willingness membership parameter under the proportionality coefficient r1; λ2 is the willingness membership parameter when the proportionality coefficient is 1; λ3 is the willingness membership parameter under the proportionality coefficient r3. The proportionality coefficients are obtained by means of an expert system.

[0034] S2. Use the Monte Carlo simulation method to generate the annual output sample scenarios of wind power / photovoltaic, and extract the representative typical scenario set through the K-means clustering algorithm. In order to consider the correlation between the uncertainties of wind power and photovoltaic, the present invention adopts a cross-combination strategy to construct the Cartesian product of different wind power and photovoltaic typical scenarios to generate a calculation typical scenario set. The obtained typical scenario set is used as the input to provide basic data for the risk quantification considering uncertainties in the subsequent virtual power plant aggregation cost calculation risk model.

[0035] S3. Construct the objective function of the virtual power plant aggregation cost calculation risk model. According to the basic data provided by the typical scenario set obtained in S2, comprehensively consider the resource call cost, network loss, regulation deviation penalty, and system tail risk loss factors. The steps are as follows:

[0036] S3.1 On the basis of determining the uncertainty scenarios, in order to realize the reliable scheduling of the virtual power plant, the present invention constructs a multi-objective cost optimization function, comprehensively considering the resource call cost, network loss, regulation deviation penalty, and system tail risk loss factors. The multi-objective cost optimization function is as follows:

[0037]

[0038] In the formula, S is the set of scenarios, π s is the probability of the typical scenario set; T is the set of scheduling periods; is the call cost of DER; is the network loss cost; is the regulation power deviation penalty; L is the risk preference coefficient, used to measure the relationship between the aggregation cost and the risk of cost change, L ∈ [0, 1], and the larger its value, the more risk-averse the decision maker is; CVaR βIt is the CVaR value when the confidence level is β.

[0039] S3.2 According to the multi-objective cost optimization function in S3.1, construct the regulation costs of each type of DER equipment such as diesel generators, energy storage systems, and wind power / solar power. Calculated based on their unit call costs and call powers, their call costs are:

[0040]

[0041] In the formula, cm is the cost per unit regulation volume of the m-th DER, and M is the total number of DERs; ΔP t m is the regulation power of the m-th DER responding to the virtual power plant, and ΔP t m ={ΔP t DG , ΔP t ES , ΔP t RE , ΔP t EV , ΔP t AC}, ΔP t DG , ΔP t ES , ΔP t RE , ΔP t EV and ΔP t AC are the regulation powers of the generator, energy storage device, wind power / solar power, electric vehicle group, and air conditioner cluster at time t, respectively.

[0042] Among them, flexible loads participate in the regulation of the virtual power plant through incentive-based demand response, and their regulation cost is the incentive cost:

[0043]

[0044] In the formula, is the regulation load ladder price corresponding to the n-th level; is the regulation load ladder price corresponding to the (n + 1)-th level; D n is the interval of the n-th level load adjustment volume section; N is the number of ladder compensation price levels; is the regulation load ladder price corresponding to the 1st level; ΔP t m is the regulation power of the m-th DER responding to the virtual power plant.

[0045] S3.3 According to the multi-objective cost optimization function in S3.1, construct the network loss cost:

[0046]

[0047] Among them, represents the network loss cost; I ij,t,s represents the current between nodes i and j; R ij,t,s represents the resistance of the line between nodes i and j.

[0048] S3.4 According to the multi-objective cost optimization function in S3.1, the regulation power deviation penalty is:

[0049]

[0050] In the formula, is the unit deviation cost, is the deviation power of the mth DER, and M represents the total number of DERs.

[0051] S3.5 According to the multi-objective cost optimization function in S3.1, due to the uncertainty of distributed resources and flexible loads, the traditional scheduling model cannot guarantee the matching reliability between the planned power and the actual capacity. To avoid the loss of cost changes caused by the fluctuations in the capabilities of distributed resources, the present invention introduces CVaR as a risk control tool in the virtual power plant aggregation scheduling model to measure the tail risk and strengthen the decision-making robustness. The calculation method of CVaR is:

[0052]

[0053] In the formula, VaR is the set VaR value, η s is the value by which the aggregated cost exceeds VaR under scenario s, and its value is positive; β is the confidence level; S represents scenario s; CVaR β represents the CVaR value under the β confidence level.

[0054] S4. Construct the constraint conditions of the virtual power plant aggregation cost calculation risk model, and construct the flexible load scheduling margin evaluation model according to the flexible load user dynamic response willingness model constructed in S1. To ensure that the feasible solutions of the objective function of the virtual power plant aggregation cost calculation risk model constructed in S3 satisfy the system operation constraints and equipment operation characteristics, the present invention constructs a constraint system from multiple levels including power boundaries, equipment limitations, network power flow, and risk measurement. The steps are as follows:

[0055] S4.1 When depicting the virtual power plant aggregation cost, it is necessary to clarify the calculation boundary. Since the regulation level of distributed resources has strong time-series characteristics, loop through each scheduling moment and separately find the maximum and minimum values of the connection node P g,t with the superior power grid to obtain the adjustable boundary for the virtual power plant aggregation to participate in peak shaving services, as shown in formulas (13) to (17):

[0056]

[0057] P g,t = ΔP mod,t + P base,t (17)

[0058] Wherein, is the upper bound of the interactive power of the connection node between the virtual power plant and the superior power grid; is the lower bound of the interactive power of the connection node between the virtual power plant and the superior power grid; P g,t is the interactive power of the connection node between the virtual power plant and the superior power grid at time t; P base,t is the baseline power of the virtual power plant at time t, which can be obtained by the day-ahead optimal dispatch calculation of the virtual power plant; ΔP mod,t is the peak shaving service volume provided by the virtual power plant for the superior power grid at time t; is the upper bound of the peak shaving service volume provided by the virtual power plant for the superior power grid at time t; is the lower bound of the peak shaving service volume provided by the virtual power plant for the superior power grid at time t; Quantile(S,q) is the q-quantile of the data set S.

[0059] S4.2 Construct the operating constraints of the diesel generator:

[0060]

[0061] Wherein, P t DG is the power of the generator at time t; is the power of the generator at time t-1 and λ are the upper and lower limit coefficients of the operating power of the generator respectively; is the upper limit of the generator ramp rate; ΔP t DG is the regulating power of the generator at time t; is the baseline operating power of the generator at time t, obtained from the day-ahead optimal dispatch calculation of the VPP.

[0062] S4.3 Construct the operating constraints of the energy storage system:

[0063] -P t ES,max ≤ P t ES ≤ P t ES,max (21)

[0064]

[0065] SOC ES (t0) = SOCES (T)(26)

[0066]

[0067] In the formula, P t ES is the charging and discharging power of the energy storage device in the t-th period. The power direction is defined as positive during discharging and negative during charging. is the charging and discharging power of the energy storage device in the (t - 1)-th period; ΔP t ES is the regulation power of the energy storage device in the t-th period; is the baseline operating power of the energy storage device in the t-th period, which can be obtained by the day-ahead optimal scheduling of the VPP; E t is the electricity of the energy storage device in the t-th period; η c and η d are the charging efficiency and discharging efficiency of the energy storage device; SOC ES (t) is the state of charge of the energy storage device in the t-th period; and are the upper and lower limits of the state of charge of the energy storage device; E N is the rated capacity of the energy storage device; σ t is a variable representing the conversion of the charging and discharging state of the energy storage. It is 1 when the charging and discharging state of the energy storage converts from t to t + 1, and 0 when the state remains unchanged. σ max is the maximum value of the conversion of the charging and discharging times of the energy storage within adjacent r periods, which is used to limit the frequent state conversion of the energy storage.

[0068] S4.4 Construct the operation constraints of wind power / photovoltaic:

[0069] (1 - ω m )P t RE,max ≤P t RE ≤P t RE,max (28)

[0070]

[0071] ΔP t RE =P t RE -P t RE,max (30)

[0072] In the formula, P t RE,max is the power upper limit of the renewable energy unit in the t-th period, which is the typical scenario set in S2; P t RE is the power of the renewable energy unit in the t-th period; is the power of the renewable energy unit at time period t-1; is the upper limit of the ramping rate of the renewable energy unit; ω m is the maximum curtailment of wind / solar rate that meets the requirements; ΔP t RE is the regulation power of the renewable energy unit at time period t.

[0073] S4.5 According to the flexible load user dynamic response willingness model constructed in S1, the constraints of the electric vehicle cluster can be obtained as follows:

[0074]

[0075] In the formula: SOC EV (t) is the state of charge of the electric vehicle cluster at time period t; η ch is the charging efficiency of the electric vehicle; P t EV is the charging power of the electric vehicle cluster at time period t; ΔP t EV is the regulation power of the electric vehicle cluster at time period t; K(t) is the probability of the vehicle being parked at time period t; M unit is the mechanical energy consumed per kilometer by the vehicle; η em is the efficiency of converting electrical energy into mechanical energy; v is the driving speed of the electric vehicle; B av is the average battery capacity of the vehicle group; and are the upper and lower limits of the dispatchable margin of the electric vehicle group at time period t, respectively; and are the maximum / minimum values of the aggregated charging power of the electric vehicle group at time period t, respectively; is the original power consumption of the electric vehicle cluster at time period t; p ch is the rated charging power of each electric vehicle.

[0076] S4.6 According to the flexible load user dynamic response willingness model constructed in S1, the constraints of the air conditioner cluster can be obtained as follows:

[0077]

[0078] In the formula: The air conditioner model can be equivalent to a virtual energy storage model, SOC AC (t) is the state of charge of the equivalent energy storage of the air conditioner at time period t; and are the adjustable upper and lower limits of the temperature at time period t, respectively; is the outdoor temperature at time period t; is the indoor temperature set value at time period t; R a is the equivalent thermal resistance of the room; η is the energy efficiency ratio of the air conditioner; C is the equivalent heat capacity of the room; P tAC is the power consumption of the air conditioner cluster during period t; N AC is the number of air conditioners; E(·) is the mathematical expectation of the variable; and are the upper and lower limits of the dispatchable margin of the air conditioner cluster during period t, respectively; is the original aggregated power of the air conditioner cluster before participating in demand response during period t; ΔP t AC is the regulation power of the air conditioner cluster during period t.

[0079] Therefore, according to the flexible load user dynamic response willingness model constructed in S1, as well as the constraints of the electric vehicle cluster in S4.5 and the air conditioner cluster constraints in S4.6, Equations (32)-(34) and Equations (41)-(43) are the constructed flexible load dispatch margin evaluation models.

[0080] S4.7 In the virtual power plant dispatch modeling, to accurately describe the power flow and voltage variation of each node in the distribution network, the present invention adopts the Distflow model applicable to the radial network structure. This Distflow model can directly establish the physical relationship between the branch power and the node voltage, and can better reflect the operating characteristics of the low-voltage distribution system. The Distflow model has good mathematical properties and can be transformed into a convex optimization problem through second-order cone relaxation (SOCP), thereby improving the solution efficiency while ensuring physical accuracy. Based on this, the present invention introduces the Distflow power flow constraint into the virtual power plant aggregation dispatch model, and its mathematical expression is as follows:

[0081]

[0082] In the formula, k ∈ {k|j→k} is the set of child nodes k with node j as the parent node, and i ∈ {i|i→j} is the parent node i with j as the child node; x ij and r ij are the reactance value and resistance value of line L ij ; P j,s,t and Q j,s,t are the active and reactive injection powers of node j under scenario s during period t, respectively; δ is the balance node flag, and its value is 1 when node j is the balance node, otherwise it is 0; P g,t and Q g,t are the active power and reactive power of the generator at the balance node during period t, respectively; P g,max and P g,min are the maximum and minimum values of the active power of the balance node generator, respectively, and Q g,max and Q g,min are the maximum and minimum values of the reactive power of the balance node generator, respectively; N j is the set of DER devices connected to node j; and are the active and reactive power loads at node j during period t; U j,s,t is the voltage of node j under scenario s during period t, and are the maximum / minimum values of the voltage of node i respectively; is the maximum value of the complex power flowing through line L ij .

[0083] S4.8 Construct the CVaR constraint as:

[0084]

[0085] where η s is the value by which the aggregated cost exceeds VaR under scenario s, and its value is positive; β is the confidence level; S represents scenario s; CVaR β represents the CVaR value at the β confidence level; π s is the probability of the set of typical scenarios; represents the network loss cost;

[0086] is the unit deviation cost; is the DER regulation cost.

[0087] Therefore, the flexible load user dynamic response willingness model constructed in S1 realizes the quantification of the participation levels of flexible loads such as electric vehicles and air-conditioning clusters under the drive of multiple factors; the set of typical scenarios of wind power / solar power constructed in S2 can be used for risk quantification considering uncertainty in the virtual power plant aggregated cost calculation risk model; the objective function of the virtual power plant aggregated cost calculation risk model constructed in S3 comprehensively considers the resource call cost, network loss, regulation deviation penalty, and system tail risk loss; the constraint conditions of the virtual power plant aggregated cost calculation risk model constructed in S4 construct a constraint system from multiple levels of power boundary, equipment limitation, network power flow, and risk measurement, ensuring that the feasible solution of the objective function of the virtual power plant aggregated cost calculation risk model constructed in S3 satisfies the system operation constraints and equipment operation characteristics; the flexible load schedulable margin evaluation model constructed in S4 realizes the evaluation of the schedulable margins of electric vehicle clusters and air-conditioning clusters.

[0088] The present invention has the following beneficial effects:

[0089] (1) The flexible load user dynamic response willingness model provided by the present invention constructs a flexible load user dynamic response willingness modeling method based on the TSK fuzzy inference system for the subjective response characteristics of flexible load users. Considering multi-dimensional influencing factors such as electricity price incentives, electric vehicle state of charge, and room temperature, it dynamically depicts the participation willingness levels of electric vehicles and air-conditioning loads. The flexible load user dynamic response willingness model can fully reflect individual heterogeneity and time-varying responsiveness, providing a reliable input for the adjustable resource boundary of the virtual power plant.

[0090] (2) The method for constructing the virtual power plant aggregation cost calculation risk model provided by the present invention introduces the CVaR risk control method and establishes a virtual power plant aggregation cost calculation risk model including distributed resource call cost, network loss cost, response deviation cost, and CVaR risk index, realizing the quantitative characterization of the relationship between the regulation service quantity and cost of the aggregation entity. Description of the Drawings

[0091] Figure 1 is the flowchart of the method of the present invention;

[0092] Figure 2 is the flowchart of the output of the wind power / solar photovoltaic scenario;

[0093] Figure 3 is the topological structure diagram of the 11-node distribution system;

[0094] Figure 4 is the typical scenario curve graph of the wind turbine generator connected to Node 3;

[0095] Figure 5 is the typical scenario curve graph of the wind turbine generator connected to Node 6;

[0096] Figure 6 is the typical scenario curve graph of the photovoltaic generator set connected to Node 9;

[0097] Figure 7 is the time-of-use electricity price curve graph of a certain place;

[0098] Figure 8 is the baseline power curve graph of distributed resources and connection nodes;

[0099] Figure 9 is the curve graph for determining the aggregation boundary using the confidence method;

[0100] Figure 10 is the curve graph of the virtual power plant aggregation boundary corresponding to different flexible load willingness values;

[0101] Figure 11 is the dynamic response willingness curve graph when the flexible load participates in the virtual power plant aggregation;

[0102] Figure 12 It is the effective frontier curve graph of CVaR in the aggregation cost calculation of the virtual power plant;

[0103] Figure 13 It is the curve graph of the aggregation regulation cost of the virtual power plant in a typical period. Specific implementation manners

[0104] The present invention will be further described below with reference to the accompanying drawings.

[0105] The method flow chart of the present invention is as Figure 1 shown, and mainly includes the following steps:

[0106] S1. Construct a dynamic response willingness model for flexible load users, including the following steps:

[0107] S1.1 Use the TSK fuzzy inference system proposed by Takagi, Sugeno and Kang to model the response willingness of flexible load users. This TSK fuzzy inference system has strong non-linear modeling ability and good interpretability, and is suitable for dealing with the behavior modeling problems of multi-input, fuzzy cognition and output continuous quantities. Compared with traditional fuzzy systems, the TSK fuzzy inference system has more advantages in terms of clear structure and calculation efficiency, and is convenient to be embedded into the scheduling optimization framework for linkage calculation.

[0108] The TSK fuzzy inference system consists of multiple fuzzy rules, and each fuzzy rule adopts the general expression form such as the k-th rule R k as shown in formula (1).

[0109] For the given input x = (x1, x2,..., x n ), the membership degree of each rule is:

[0110]

[0111] The output λ of the final model is the weighted average of the outputs of each rule as the response willingness:

[0112]

[0113] where, K is the total number of rules, where, K is the total number of rules, y k (x) is the function of rule k.

[0114] S1.2 Based on the TSK fuzzy inference system and the flexible load user dynamic response willingness model established in S1.1, a flexible load user response willingness model driven by multiple factors is established. Affected by multiple factor interactions, such as the incentive electricity price level, environmental conditions, and the behavior preferences of flexible load users. Such decision-making behaviors have significant nonlinear and fuzzy characteristics and are difficult to characterize through traditional linear modeling methods. Therefore, in the present invention, the response willingness λ ∈ [0, 1] of flexible load users represents the degree of their participation in load regulation under current incentive conditions. By setting appropriate input variables and fuzzy rules, the response levels of group or individual flexible load users can be dynamically estimated and used to construct dynamic scheduling constraints or adjustable margin ranges.

[0115] For two types of typical flexible loads, electric vehicles and air conditioners, key factor sets affecting their response willingness are constructed respectively:

[0116] For electric vehicle users, consider input variables: State of Charge (SOC); incentive price level (Price). After normalization processing of the input variables, they are fuzzily mapped to variables: "LOW", "MID", "HIGH", and corresponding triangular membership functions are set for description.

[0117] For air conditioner users, consider input variables: State of Charge (SOC, analogizing the air conditioner cluster to a virtual energy storage); incentive electricity price level (Price). After normalization processing of the input variables, they are fuzzily mapped to variables: "LOW", "MID", "HIGH", and corresponding triangular membership functions are set for description.

[0118] In the fuzzy rule layer, an "IF-THEN" type rule set is established through expert experience and behavior analysis:

[0119]

[0120] The output function of each rule adopts a linear form and is written as:

[0121]

[0122] In the formula, f k (x) is the output function of the k-th rule; a k is the linear coefficient vector of the input variables; b k is the rule bias term; x is the vector composed of the input variables.

[0123] Calculate the response willingness of flexible loads according to formulas (2)-(3). In the actual demand response process, there must be a certain degree of uncertainty in the response willingness. Use the triangular membership function to describe the uncertainty degree of the response willingness of flexible load users, then the fuzzy expression of the response willingness is shown in formula (6).

[0124] S2. Use the Monte Carlo simulation method to generate the annual output sample scenarios of wind power / photovoltaic, and extract the representative typical scenario set through the K-means clustering algorithm. In order to consider the correlation between the uncertainties of wind power and photovoltaic, the present invention adopts a cross-combination strategy to construct the Cartesian product of different typical scenarios of wind power and photovoltaic to generate a calculation typical scenario set. Figure 2 It is the flow chart of the output of wind power / photovoltaic scenarios in the present invention. Figure 3 It is the topological structure diagram of an 11-node distribution system. Among them, node 1 is the connection node with the upper-level power grid. Wind turbines of 1.5 MW and 1 MW are connected to nodes 3 and 6 respectively, and a 1 MW photovoltaic unit is connected to node 9. According to the above steps in the figure, the typical scenarios and their outputs of wind power / photovoltaic at node 3 are shown in Figure 4 The typical scenarios and their outputs of wind power / photovoltaic at node 6 are shown in Figure 5 The typical scenarios and their outputs of wind power / photovoltaic at node 9 are shown in Figure 6 . In order to consider the correlation between uncertainties, the typical scenarios are processed during the overall operation, and finally 64 typical scenario sets are defined for subsequent calculations.

[0125] S3. Construct the objective function of the virtual power plant aggregation cost calculation risk model. According to the basic data provided by the typical scenario set obtained in S2, comprehensively consider factors such as resource call cost, network loss, regulation deviation penalty, and system tail risk loss, including the following steps:

[0126] S3.1 On the basis of determining the uncertainty scenarios, in order to achieve the reliable scheduling of the virtual power plant, the present invention constructs a multi-objective cost optimization function, comprehensively considering factors such as resource call cost, network loss, regulation deviation penalty, and system tail risk loss. Its structure is as follows:

[0127]

[0128] In the formula, S is the set of scenarios, and π s is the probability of the typical scenario set; T is the set of scheduling time periods; is the call cost of DER; is the network loss cost; is the regulation power deviation penalty; L is the risk preference coefficient, which is used to measure the relationship between the aggregation cost and the risk of cost change. L ∈ [0, 1], and the larger its value, the more risk-averse the decision maker is; CVaR βIs the CVaR value when the confidence level is β.

[0129] S3.2 According to the multi-objective cost optimization function in S3.1, construct the regulation costs of each type of DER equipment, namely diesel generators, energy storage systems, and wind power / solar power. Calculate according to their unit call costs and call powers, and their call costs are:

[0130]

[0131] In the formula, c m Is the cost per unit regulation amount of the m-th DER, and M is the total number of DERs; ΔP t m Is the regulation power of the m-th DER responding to the virtual power plant, ΔP t m ={ΔP t DG , ΔP t ES , ΔP t RE , ΔP t EV , ΔP t AC}, ΔP t DG , ΔP t ES , ΔP t RE , ΔP t EV , and ΔP t AC Are the regulation powers of the generator, energy storage device, wind power / solar power, electric vehicle group, and air conditioner cluster at time t, respectively.

[0132] Among them, flexible loads participate in the regulation of the VPP through incentive-based demand response, and their regulation costs are incentive costs as shown in formula (9).

[0133] S3.3 According to the multi-objective cost optimization function in S3.1, construct the network loss cost:

[0134]

[0135] Among them, Represents the network loss cost; I ij,t,s Represents the current between nodes i and j; R ij,t,s Represents the resistance of the line between nodes i and j.

[0136] S3.4 According to the multi-objective cost optimization function in S3.1, the regulation power deviation penalty is:

[0137]

[0138] In the formula, is the unit deviation cost, is the deviation power of the m-th DER, and M represents the total number of DERs.

[0139] S3.5 According to the multi-objective cost optimization function in S3.1, due to the uncertainty of distributed resources and flexible loads, the traditional scheduling model cannot guarantee the matching reliability between the planned power and the actual capacity. To avoid the loss of cost changes caused by the fluctuation of distributed resource capabilities, the present invention introduces CVaR as a risk control tool in the virtual power plant aggregation scheduling model to measure the tail risk and strengthen the decision-making robustness. The calculation method of CVaR is as follows:

[0140]

[0141] In the formula, VaR is the set VaR value, and η s is the value by which the aggregated cost exceeds VaR under scenario s, and its value is positive; β is the confidence level; S represents scenario s; CVaR β represents the CVaR value under the β confidence level.

[0142] S4. Construct the constraint conditions of the virtual power plant aggregated cost calculation risk model, and construct a flexible load scheduling margin evaluation model according to the flexible load user dynamic response willingness model constructed in S1. To ensure that the feasible solutions of the objective function of the virtual power plant aggregated cost calculation risk model constructed in S3 satisfy the system operation constraints and equipment operation characteristics, the present invention constructs a constraint system from multiple levels such as power boundaries, equipment limitations, network power flow, and risk measurement, including the following steps:

[0143] S4.1 When characterizing the virtual power plant aggregated cost, it is necessary to clarify the calculation boundary. Since the regulation level of distributed resources has strong time-series characteristics, each scheduling moment is cycled, and the maximum and minimum values of the node P g,t connected to the superior power grid are respectively obtained to obtain the adjustable boundary for the virtual power plant to participate in peak shaving services, as shown in formulas (13) to (17).

[0144] S4.2 Construct the operating constraints of the diesel generator, as shown in formulas (18) to (20).

[0145] S4.3 Construct the operating constraints of the energy storage system, as shown in formulas (21) to (27).

[0146] S4.4 Construct the operating constraints of wind power / solar power, as shown in formulas (28) to (30).

[0147] S4.5 According to the flexible load user dynamic response willingness model constructed in S1, the constraints of the electric vehicle cluster can be obtained, as shown in Formulas (31) to (37).

[0148] S4.6 According to the flexible load user dynamic response willingness model constructed in S1, the constraints of the air conditioner cluster can be obtained, as shown in Formulas (38) to (45).

[0149] Therefore, based on the flexible load user dynamic response willingness model constructed in S1, and the constraints of the electric vehicle cluster in S4.5 and the air conditioner cluster in S4.6, Formulas (32)-(34) and Formulas (41)-(43) are the constructed flexible load scheduling margin evaluation models.

[0150] S4.7 In the virtual power plant scheduling modeling, to accurately describe the power flow and voltage change of each node in the distribution network, the present invention adopts the Distflow model applicable to the radial network structure. This model can directly establish the physical relationship between the branch power and the node voltage, and can better reflect the operation characteristics of the low-voltage distribution system. The Distflow model has good mathematical properties and can be transformed into a convex optimization problem through second-order cone relaxation (SOCP), so as to improve the solution efficiency while ensuring the physical accuracy. Based on this, the present invention introduces the Distflow power flow constraint into the virtual power plant aggregation scheduling model, and its expression is as shown in Formulas (46) to (52).

[0151] S4.8 Construct the CVaR constraint, as shown in Formulas (53) to (54).

[0152] The present invention takes Figure 3 the virtual power plant composed of the 11-node distribution system and the aggregation of multiple distributed resources in Figure 7 as an example to verify the effectiveness of the method of the present invention by providing peak shaving services to the superior power grid. Among them, Node 1 is the connection node with the superior power grid. Wind turbines of 1.5 MW and 1 MW are connected to Node 3 and Node 6 respectively, and a photovoltaic unit of 1 MW is connected to Node 9. The typical scenarios and their outputs of Node 3, Node 6 and Node 9 have been obtained according to S2. A diesel generator of 2 MW is connected to Node 7, a energy storage battery of 4 MVA is connected to Node 5, a cluster composed of 300 electric vehicles is connected to Node 4, and a cluster composed of 5000 air conditioners is connected to Node 10. Combining the local current time-of-use electricity price as Figure 8It is the baseline power curve graph of distributed resources and liaison nodes. Through the day-ahead optimal scheduling of the virtual power plant, the electric vehicle cluster mainly conducts centralized charging during the 1-7 time periods, and the power consumption of the air-conditioning cluster is in line with the changing trend of the outdoor temperature, mainly concentrated at noon. The aggregated power of the liaison node reflects the electricity consumption characteristics of this distributed resource area, showing a "double-peak" trend. The energy storage battery adjusts the regional power characteristics through time-of-use electricity price incentives, discharging during high-price periods and charging during low-price periods.

[0153] Based on the calculated baseline power of each distributed resource and liaison node, and combined with the 64 typical scenario sets constructed in the present invention, by cyclically solving the upper and lower limits of the aggregated power of the virtual power plant at each scheduling moment, the confidence method is used to determine the range of the aggregated power. Figure 9 It is the curve graph for determining the aggregated boundary using the confidence method. The uncertainties of renewable energy sources such as wind power and photovoltaic power significantly affect the aggregated boundary of the virtual power plant. Therefore, the cost fluctuation risks brought by uncertainties must be considered during the aggregation process of the virtual power plant.

[0154] Figure 10 It is the curve graph of the aggregated boundary of the virtual power plant corresponding to different flexible load willingness values, which shows the aggregated power range of the virtual power plant when the flexible load does not participate in peak shaving, participates in peak shaving with static response willingness, and participates in peak shaving with dynamic response willingness. Calculated according to a 95% confidence level, the aggregated peak shaving range is the largest with static response willingness, followed by dynamic response willingness, and the smallest when not participating in peak shaving. When the fixed response willingness is 1, the aggregated peak shaving range of the virtual power plant is too idealized to accurately reflect the response behavior of flexible load users in actual scheduling, thus possibly generating scheduling penalty costs.

[0155] Figure 11 It shows the dynamic response willingness of the flexible load in the aggregated scheduling of the virtual power plant. The response willingness of the air-conditioning cluster is relatively low during the 23-3 time periods and 8-14 time periods, which is in line with the outdoor temperature and user comfort requirements; the response willingness of the electric vehicle cluster is relatively high during the 0-8 time periods because the driving demand at night is low, and the incentive-based demand response can encourage users to actively participate in scheduling.

[0156] Figure 12It is the effective frontier curve of CVaR in the aggregated cost calculation of the virtual power plant. Taking t = 16 and ΔP = 1.6 MW as examples, CVaR is used to quantify the risk of the VPP aggregation cost. As the risk preference coefficient β increases, the CVaR value gradually decreases and the regulation cost gradually increases, indicating that the higher the risk aversion degree of the decision-maker, the more conservative the aggregation strategy, thus increasing the aggregation regulation cost. At the same time, the increase in the β value makes the decision-maker more sensitive to risks, and the change range of the aggregation regulation cost is larger. The two curves in the figure respectively represent the CVaR effective frontier curves under different confidence levels, showing that at the same risk level, a higher confidence level corresponds to a larger aggregation regulation cost.

[0157] Figure 13 It is the curve graph of the aggregated regulation cost of the virtual power plant in a typical period. The regulation ability of the virtual power plant to aggregate various distributed resources to provide peak shaving services has time series characteristics. Therefore, when characterizing the aggregation cost, by depicting the cost characteristics of each period, it provides an effective reference for the virtual power plant to participate in the market declaration capacity price. Taking the confidence level α = 0.8 and the risk preference coefficient β = 0.5 as examples, since the baseline power has been determined through day-ahead optimal scheduling, the cost functions of all periods pass through the origin, showing the regulation costs under different regulation amounts in 4 typical periods. As the regulation amount increases, the aggregated regulation cost of the virtual power plant rises, and the curve presents a "U" shape; at the same time, as the regulation amount increases, the slope of the curve gradually increases because the virtual power plant preferentially aggregates distributed resources with low risk and low cost, and only starts to call high-risk and high-cost distributed resources after their potential is exhausted. The results verify the effectiveness and superiority of the method proposed in the present invention in the aggregated resource scheduling of the virtual power plant considering the response willingness of flexible load users.

[0158] In view of the actual background of the coexistence of the wide access of flexible loads and the uncertainty of renewable power generation in the new power system, the present invention constructs a method for quantifying the aggregated cost of a virtual power plant considering the fuzzy dynamic willingness of multiple demand responses. The TSK fuzzy inference system is used to depict the behaviors of electric vehicle and air conditioner users, and the CVaR risk measurement method is introduced to construct a risk model for calculating the aggregated cost of the virtual power plant that integrates response willingness, uncertainty, network constraints and risk control, realizing the quantitative characterization of the relationship between the regulation service volume and cost of the aggregation entity.

[0159] The above-described embodiments only represent the implementation manners of the present invention, but should not be construed as limiting the scope of the invention patent of the present invention. It should be noted that for those skilled in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention.

Claims

1. A method for quantifying the aggregation cost of a virtual power plant considering the fuzzy dynamic willingness of multiple demand responses, characterized in that The method described above includes the following steps: S1. Build a dynamic response willingness model for flexible load users to quantify the participation levels of flexible loads such as electric vehicles and air-conditioning clusters under the drive of multiple factors. The steps include: S1.1 During the process of the virtual power plant providing peak shaving auxiliary services, use the TSK fuzzy inference system to model the response willingness of flexible load users and build a dynamic response willingness model for flexible load users. S1.2 Under the framework of the TSK fuzzy inference system, build a flexible load response willingness model driven by multiple factors to estimate the subjective participation tendency of flexible load users and provide a basis for subsequent adjustable capacity evaluation and scheduling decisions. S2. Generate annual output sample scenarios of wind power and photovoltaic power, and extract a representative set of typical scenarios; then perform Cartesian product construction on different wind power and photovoltaic typical scenarios to generate a set of calculated typical scenarios; the obtained set of typical scenarios is used as input to provide basic data for risk quantification considering uncertainty in the subsequent virtual power plant aggregation cost calculation risk model. S3. Build an objective function for the virtual power plant aggregation cost calculation risk model, and comprehensively consider factors such as resource call cost, network loss, regulation deviation penalty, and system tail risk loss based on the basic data provided by the set of typical scenarios obtained in S2. S4. Build constraint conditions for the virtual power plant aggregation cost calculation risk model, and build a flexible load scheduling margin evaluation model based on the dynamic response willingness model of flexible load users built in S1; to ensure that the feasible solutions of the objective function of the virtual power plant aggregation cost calculation risk model built in S3 meet the system operation constraints and equipment operation characteristics, build a constraint system from multiple levels such as power boundary, equipment limit, network power flow, and risk measurement.

2. A virtual power plant aggregation cost quantification method considering fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that In the S1.1, the TSK fuzzy inference system is composed of multiple fuzzy rules, and each fuzzy rule adopts the general expression form such as the k-th rule R k : Among them, x1, x2,..., x n are input variables, which can be price, power demand, temperature difference, etc., represents the fuzzy set corresponding to the nth input variable in the kth rule, and y k is the output function of the kth rule, generally a linear combination of the inputs; For a given input \(x=(x_1,x_2,\cdots,x\) n ), the membership degree of each fuzzy rule is: Finally, the output λ of the dynamic response willingness model of flexible load users is the weighted average of the outputs of each fuzzy rule as the response willingness: where K is the total number of rules, and y k (x) is the function of rule k; The response willingness λ ∈ [0, 1] of flexible load users represents the degree of their participation in load regulation under the current incentive conditions; by setting appropriate input variables and fuzzy rules, the response levels of group or individual users can be dynamically estimated and used to build dynamic scheduling constraints or adjustable margin ranges.

3. A method for quantifying the aggregation cost of a virtual power plant considering the fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that, In S1.2, for two types of typical flexible loads, electric vehicles and air conditioners, build key factor sets that affect the response willingness respectively: For electric vehicle users, consider input variables: state of charge SOC, incentive price level Price. After normalization processing of the input variables, they are fuzzily mapped to variables: "LOW", "MID", "HIGH", and corresponding triangular membership functions are set for description. For air conditioner users, consider input variables: state of charge SOC, and analogize the air-conditioning cluster to virtual energy storage; incentive electricity price level Price. After normalization processing of the input variables, they are fuzzily mapped to variables: "LOW", "MID", "HIGH", and corresponding triangular membership functions are set for description. In the fuzzy rule layer, establish an "IF-THEN" rule set through expert experience and behavior analysis, as shown in formula (4): The output function of each rule adopts a linear form and is written as: where f k (x) is the output function of the k-th rule; a k is the linear coefficient vector of the input variable; b k is the rule bias term; x is the vector formed by the input variables; The response willingness of flexible load users is calculated according to formulas (2)-(3). In the actual demand response process, there must be a certain degree of uncertainty in the response willingness. Using the triangular membership function to describe the uncertainty degree of the response willingness of flexible load users, the fuzzy expression of the response willingness is as follows: Wherein: is the fuzzy expression of the response willingness of flexible load users; λ1, λ2, and λ3 are the triangular membership degree parameters of the response willingness, r1 and r3 are proportionality coefficients, and 0 < r1 < 1, r3 > 1. λ1 is the willingness membership degree parameter under the proportionality coefficient r1; λ2 is the willingness membership degree parameter with the proportionality coefficient of 1; λ3 is the proportionality coefficient of the willingness membership degree parameter under the proportionality coefficient r3. The proportionality coefficients are obtained by means of an expert system.

4. A method for quantifying the aggregation cost of a virtual power plant considering the fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that In S2, the Monte Carlo simulation method is used to generate the annual output sample scenarios of wind power and photovoltaic power, and the K-means clustering algorithm is used to extract the representative typical scenario set. The cross-combination strategy is used to construct the Cartesian product of different wind power and photovoltaic typical scenarios to generate the calculation typical scenario set.

5. A method for quantifying the aggregation cost of a virtual power plant considering the fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that S3 includes the following steps: S3.1 Based on the determined uncertainty scenarios, to achieve the reliable dispatch of the virtual power plant, a multi-objective cost optimization function is constructed. The multi-objective cost optimization function is as follows: Where S is the set of scenarios, and π s is the probability of the set of typical scenarios; T is the set of scheduling periods; is the call cost of DER; is the network loss cost; is the penalty for the regulation power deviation; L is the risk preference coefficient, which is used to measure the relationship between the aggregation cost and the risk of cost variation. L ∈ [0, 1], and the larger its value, the more risk-averse the decision maker is; CVaR β is the CVaR value at the confidence level of β; S3.2 According to the multi-objective cost optimization function in S3.1, the adjustment costs of each type of DER equipment such as diesel generators, energy storage systems, and wind power / photovoltaic power are constructed. Calculated according to their unit call cost and call power, their call cost is: where c m is the cost of the regulation amount of the m-th DER unit, and M is the total number of DERs; ΔP t m is the regulation power of the m-th DER responding to the virtual power plant, and ΔP t m = {ΔP t DG , ΔP t ES , ΔP t RE , ΔP t EV , ΔP t AC}, where ΔP t DG , ΔP t ES , ΔP t RE , ΔP t EV and ΔP t AC are the regulation powers of the generator, energy storage device, wind power / solar power, electric vehicle cluster, and air conditioner cluster at time t, respectively; The flexible load participates in the regulation of the virtual power plant through incentive-based demand response, and its regulation cost is the incentive cost: In the formula, is the stepped price of the regulation load corresponding to the nth level; is the stepped price of the regulation load corresponding to the (n + 1)th level; D n is the interval of the load adjustment amount section at the nth level; N is the stepped compensation price level; is the stepped price of the regulation load corresponding to the first level; ΔP t m is the regulation power of the mth DER responding to the virtual power plant; S3.3 According to the multi-objective cost optimization function in S3.1, the network loss cost is constructed: Among them, represents the network loss cost; I ij,t,s represents the current between nodes i and j; R ij,t,s represents the resistance of the line between nodes i and j; S3.4 Adjust the power deviation penalty according to the multi-objective cost optimization function in S3.1 It is: wherein, is the unit deviation cost, is the deviation power of the m-th DER, and M represents the total number of DERs; S3.5 Introduce CVaR as a risk control tool in the virtual power plant aggregation dispatch model to measure the tail risk and strengthen the decision-making robustness.

6. A method for quantifying the aggregation cost of a virtual power plant considering the fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that In S3.5, the calculation method of CVaR is: Where VaR is the set VaR value, and η s is the value by which the aggregated cost exceeds VaR under scenario s, and its value is positive; β is the confidence level; S represents scenario s; CVaR β represents the CVaR value at the β confidence level.

7. A method for quantifying the aggregation cost of a virtual power plant considering the fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that S4 includes the following steps: S4.1 When characterizing the aggregation cost of the virtual power plant, it is necessary to clarify the calculation boundary. Since the regulation level of distributed resources has strong time-series characteristics, each scheduling moment is cycled, and the maximum and minimum values of the connection node P with the superior power grid are calculated separately to obtain the adjustable boundary for the virtual power plant to participate in peak shaving services, as shown in formulas (13) to (17): g,t The maximum and minimum values are obtained to get the adjustable boundary for the virtual power plant to participate in peak shaving services, as shown in Formulas (13) to (17): P g,t = ΔP mod,t + P base,t (17) In the formula, is the upper bound of the interactive power of the virtual power plant and the connection node of the superior power grid; is the lower bound of the interactive power of the virtual power plant and the connection node of the superior power grid; P g,t is the interactive power of the virtual power plant and the connection node of the superior power grid at time t; P base,t is the baseline power of the virtual power plant at time t, which can be obtained by the day-ahead optimal scheduling calculation of the virtual power plant; ΔP mod,t is the peak shaving service volume provided by the virtual power plant for the superior power grid at time t; is the upper bound of the peak shaving service volume provided by the virtual power plant for the superior power grid at time t; is the lower bound of the peak shaving service volume provided by the virtual power plant for the superior power grid at time t; Quantile(S,q) is the q-quantile of the data set S; S4.2 Construct the operating constraints of diesel generators; S4.3 Construct the operating constraints of energy storage systems; S4.4 Construct the operating constraints of wind power / photovoltaic power; S4.5 According to the flexible load user dynamic response willingness model constructed in S1, the constraints of the electric vehicle cluster are obtained as: Where: SOC EV (t) is the state of charge of the electric vehicle cluster in the t time period; η ch is the charging efficiency of the electric vehicle; P t EV is the charging power of the electric vehicle group in the t time period; ΔP t EV is the regulation power of the electric vehicle cluster in the t time period; K(t) is the parking probability of the vehicle in the t time period; M unit is the mechanical energy consumed per kilometer by the vehicle; η em is the efficiency of converting electrical energy into mechanical energy; v is the driving speed of the electric vehicle; B av is the average battery capacity of the vehicle group; and are the upper and lower limits of the dispatchable margin of the electric vehicle group in the t time period respectively; and are the maximum / minimum values of the aggregated charging power of the electric vehicle group in the t time period respectively; is the original power consumption of the electric vehicle cluster in the t time period; p ch is the rated charging power of each electric vehicle; S4.6 According to the flexible load user dynamic response willingness model constructed in S1, the air-conditioning cluster constraints are obtained as: In the formula: The air conditioner model can be equivalent to a virtual energy storage model, and SOC AC (t) is the state of charge of the equivalent energy storage of the air conditioner in the t-th period; and are the adjustable upper and lower limits of the temperature in the t-th period respectively; is the outdoor temperature in the t-th period; is the indoor temperature set value in the t-th period; R a is the equivalent thermal resistance of the room; η is the energy efficiency ratio of the air conditioner; C is the equivalent heat capacity of the room; P t AC is the power consumption of the air conditioner cluster in the t-th period; N AC is the number of air conditioners; E(·) is the mathematical expectation of the variable; and are the upper and lower limits of the dispatchable margin of the air conditioner cluster in the t-th period respectively; is the original aggregated power of the air conditioner cluster before participating in demand response in the t-th period; ΔP t AC is the adjustment power of the air conditioner cluster in the t-th period; Therefore, according to the flexible load user dynamic response willingness model constructed in S1, and the constraints of the electric vehicle cluster in S4.5 and the air-conditioning cluster constraints in S4.6, formulas (32)-(34) and formulas (41)-(43) are the constructed flexible load dispatch margin evaluation models; S4.7 In the virtual power plant dispatch modeling, the Distflow model applicable to the radial network structure is used; that is, the Distflow power flow constraint is introduced into the virtual power plant aggregation dispatch model, and its expression is as follows: where \(k\in\{k|j\rightarrow k\}\) is the set of child nodes \(k\) with node \(j\) as the parent node, and \(i\in\{i|i\rightarrow j\}\) is the parent node \(i\) with \(j\) as the child node; \(x\) ij and \(r\) ij are the reactance value and resistance value of line \(L\) ij ; \(P\) j,s,t and \(Q\) j,s,t are the active and reactive injection powers at node \(j\) under scenario \(s\) at time period \(t\) respectively; \(\delta\) is the balancing node flag, which is 1 when node \(j\) is the balancing node and 0 otherwise; \(P\) g,t and \(Q\) g,t are the active power and reactive power of the generator at the balancing node at time period \(t\) respectively; \(P\) g,max and \(P\) g,min are the maximum and minimum values of the active power of the balancing node generator respectively, and \(Q\) g,max and \(Q\) g,min are the maximum and minimum values of the reactive power of the balancing node generator respectively; \(N\) j is the set of DER devices connected to node \(j\); and are the active and reactive loads at node \(j\) at time period \(t\) respectively; \(U\) j,s,t is the voltage of node \(j\) under scenario \(s\) at time period \(t\), and are the maximum / minimum values of the voltage of node \(i\) respectively; is the maximum value of the complex power flowing through line \(L\) ij ; S4.8 Construct the CVaR constraint as: Among them, η s is the value by which the aggregation cost exceeds VaR under scenario s, and its value is positive; β is the confidence level; S represents scenario s; CVaR β represents the CVaR value at the β confidence level; πs is the probability of the typical scenario set; represents the network loss cost; is the unit deviation cost; is the DER adjustment cost.

8. A method for quantifying the aggregation cost of a virtual power plant considering the fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that The operating constraints of the diesel generator in S4.2 are: Wherein, P t DG is the power of the generator in the t period; is the power of the generator in the t - 1 period and λ are the upper and lower limit coefficients of the generator operating power respectively; is the upper limit of the generator ramp rate; ΔP t DG is the regulation power of the generator during period t; is the baseline operating power of the generator during period t, which is obtained from the day-ahead optimal scheduling calculation of the VPP.

9. A method for quantifying the aggregation cost of a virtual power plant considering fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that, The operating constraints of the energy storage system in S4.3 are: -P t ES,max ≤P t ES ≤P t ES,max (21) SOC ES (t0) = SOC ES (T) (26) Wherein, P t ES is the charging and discharging power of the energy storage device in the t period. The power direction is defined as positive during discharging and negative during charging. is the charging and discharging power of the energy storage device in the (t - 1) period; ΔP t ES is the regulation power of the energy storage device in the t period; is the baseline operating power of the energy storage device in the t period, which can be obtained by the day-ahead optimal scheduling of the VPP; E t is the electricity quantity of the energy storage device in the t period; η c and η d are the charging efficiency and discharging efficiency of the energy storage device; SOC ES (t) is the state of charge of the energy storage device at time t; and are the upper and lower limits of the state of charge of the energy storage device; E N is the rated capacity of the energy storage device; σ t is a variable representing the conversion of the charging and discharging state of the energy storage. It is 1 when the charging and discharging state of the energy storage converts from t to t + 1, and 0 when the state remains unchanged. σ max is the maximum value of the conversion of the charging and discharging times of the energy storage within adjacent r time periods, which is used to limit the frequent state conversion of the energy storage.

10. A virtual power plant aggregation cost quantification method considering fuzzy dynamic willingness of multiple demand responses according to claim 1, characterized in that, The operating constraints of wind power / photovoltaic power in S4.4 are: (1 - ω m )P t RE,max ≤P t RE ≤P t RE,max (28) Wherein, P t RE,max is the power upper limit of the renewable energy unit in period t, and is the typical scenario set in S2; P t RE is the power of the renewable energy unit in period t; is the power of the renewable energy unit in period t-1; is the upper limit of the ramp rate of the renewable energy unit; ω m is the maximum curtailment rate of wind / solar power that meets the requirements; ΔP t RE is the regulation power of renewable energy units during period t.

Images