Novel power distribution system multi-time scale reserve capacity optimal configuration method and device

Abstract

The invention discloses a novel power distribution system multi-time scale reserve capacity optimization configuration method and device, and the method comprises the steps: generating an operation scene based on wind and light load information through employing a Wasserstein gradient penalty generative adversarial network, and generating a typical scene set through employing an improved K-means clustering reduction scene; based on the multi-time-scale standby service demand model, constructing a power distribution system multi-time-scale standby capacity double-layer optimization configuration model based on conditional value-at-risk, and applying a conditional value-at-risk theory to control operation risks existing in a typical scene set; performing equivalent transformation on the double-layer optimal configuration model and constraint conditions based on second-order cone relaxation and parameter transfer, and obtaining an optimal configuration result through a solver; and based on an optimal configuration result, realizing optimal configuration of the standby resources with the minimum optimal configuration and operation cost. The device comprises a processor and a memory.

Description

Technical Field

[0001] The present invention relates to the technical field of optimized operation of power distribution systems, and in particular to a novel method and device for optimizing configuration of multi-time scale spare capacity of power distribution systems. Background Art

[0002] In the new power system with a high proportion of new energy access, the failure of DC commutation when the power of new energy stations is delivered, extreme weather and the uncertainty of new energy itself may bring short-term high-power disturbances to the power system, seriously threatening the safe operation of the power grid. The system needs to reserve sufficient backup to deal with the risk of power shortage in operation. Traditional thermal power units have slow ramp rate and long response time. Hydropower and pumped storage power stations have good regulation performance, but they are greatly affected by seasonal and location factors and have limited application scope. Relying solely on the existing regulation capacity of the system is not enough to meet the backup regulation needs of the system after large-scale new energy access in the future. With the development of advanced metering facilities and management and control methods, user-side resources such as distributed power sources (distribution generations, DGs) such as micro gas turbines, energy storages (energy storages, ESs), flexible loads (flexible loads, FL), electric vehicles (electric vehicles, EV) and other user-side resources in the distribution system will become high-quality flexible backup resources. The distribution system can provide backup support for the upper power grid by reasonably dispatching local flexible resources.

[0003] At present, the research focus on reserve is on the power system operation level, optimizing the day-ahead and real-time dispatch strategies under given flexible resources. The research on the optimal configuration of flexible resources in the distribution system mainly focuses on the balance of power and electricity, and the reserve potential of flexible resources on the distribution side has not been fully explored. On the one hand, the individual capacity of flexible resources within the distribution system is small, the distribution is wide, and it is dispersed, and it has certain uncertainties, making it difficult to directly control and utilize; on the other hand, due to the coupling of reserve and real-time energy operation strategies, the distribution system needs to coordinate the optimization of short-term operation strategies and medium- and long-term resource selection when optimizing the configuration of reserve resources.

[0004] There are still some difficulties in solving the above problems: various types of flexible resources are affected by factors such as electricity price policies, incentive signals, and user response willingness. The reserve capacity has significant differences and uncertainty, and the means of quantifying uncertainty still need to be improved; the distribution system provides different combinations of reserve products, which will bring lower operating costs, but it is very complicated to ensure that all services are completed in the best way; there are a large number of nonlinear constraints related to investment decision variables and optimal power flow in the optimization configuration model. Specifically, nonlinear constraints will technically make the feasible solution space in the optimization process more complex and difficult to search, especially when dealing with the dynamic response of distributed energy and reserve resources, which may lead to convergence and stability problems in the solution process. In addition, since these nonlinear constraints involve complex interactions of electrical parameters (such as voltage, current, etc.) in the power system, traditional linearization methods are difficult to effectively capture these details, which may lead to inaccurate approximation of system behavior in technical implementation, thus affecting the safety and reliability of power grid operation. Summary of the invention

[0005] The present invention provides a novel method and device for optimizing the configuration of multi-time scale backup capacity of a power distribution system. Under the background of high proportion of renewable energy access, the present invention constructs an optimization model based on conditional risk value, effectively integrates distributed energy storage, renewable energy, flexible load and other flexible resources, accurately configures capacity for multi-time scale backup demand, enhances the system's rapid response capability to frequency fluctuations and load fluctuations, thereby improving the overall operation stability and reliability of the power grid. See the following description for details:

[0006] A novel multi-time scale reserve capacity optimization configuration method for a power distribution system, the method comprising:

[0007] Construct a distribution system operation and management framework that takes into account multiple business formats, and establish a multi-time scale reserve service demand model;

[0008] This architecture connects grid demand and supply by coordinating and integrating distributed resources, adjustable loads and energy storage. This architecture can achieve efficient management and coordination of internal resources, while also having the ability to interact with external resources in different dimensions. By solving decision-making problems at different operation stages, achieving global optimization, designing and implementing different types of distributed resource incentive mechanisms, and supporting internal resources to meet multi-format service needs such as multi-time scale backup;

[0009] The Wasserstein gradient penalty generative adversarial network is used to generate operation scenarios based on wind and solar load information, and then the improved K-means clustering is used to reduce the scenarios to generate a typical scenario set;

[0010] Based on the multi-time scale reserve service demand model, a two-layer optimization configuration model of the multi-time scale reserve capacity of the distribution system based on the conditional risk value is constructed, and the conditional risk value theory is applied to control the operation risks existing in typical scenarios; based on the second-order cone relaxation and parameter transfer, the two-layer optimization configuration model and constraint conditions are equivalently transformed, and the optimization configuration results are obtained through the solver;

[0011] Based on the optimization configuration results, the optimal configuration of backup resources is achieved with the minimum optimization configuration and operation costs.

[0012] The establishment of a multi-time scale standby service demand model is as follows:

[0013] Construct a reserve service model at a time scale of seconds to meet the emergency frequency response needs. The modeling of this reserve service takes into account the activation signal and price data of the reserve capacity, and requires that the increase in reserve capacity is equal to the decrease in reserve capacity.

[0014] Construct a 10-minute time scale backup service model to make up for power shortages and smooth load fluctuations after an accident. The backup service is required to be fully activated within 5 minutes. The backup service is asymmetric, and backup capacity can only be configured in one direction during a period of time.

[0015] A 30-minute time scale reserve service model is constructed to adjust the reserve service upward or downward in response to net load forecast errors.

[0016] The dual-layer optimization configuration model of multi-time scale reserve capacity of distribution system based on conditional risk value is constructed as follows:

[0017] The upper-level optimization configuration model decides the optimal configuration of the backup resources, with the goal of minimizing the optimal configuration cost of the flexible backup resources of the distribution system;

[0018] In the lower-level model, CVaR is used to characterize the potential operating risks of the distribution system, and the operating costs are minimized by actively managing local flexible reserve resources.

[0019] The Wasserstein gradient penalty is: introducing a gradient penalty term on the discriminator D to obtain the optimization target of the WGAN-GP method;

[0020]

[0021] In the formula, ||·|| represents the 1 norm, γ is the regularization coefficient, which is generally 10. And ε~U[0,1]; ξ is a sample vector; G(z) is a sample vector generated by the generator G; D(ξ) and D(G(z)) are the output samples of the discriminator D corresponding to ξ and G(z) respectively; It is a gradient operator. Input historical data to generate operation scenarios based on historical wind power, photovoltaic and load data.

[0022] The optimal configuration cost of minimizing the flexibility reserve resources of the power distribution system is:

[0023] minf1=C GT,inv +C ES,inv +C HVAC,inv +C DS

[0024]

[0025] In the formula, C GT,inv is the optimal configuration cost of the gas unit; C ES,inv is the optimal configuration cost of the energy storage system; C HVAC,inv is the optimal configuration cost of the air conditioning system, including the initial registration compensation cost paid to air conditioning users and the configuration cost of the response terminal; C DS is the annual operating cost of the distribution system; Ω GΤ is the candidate installation node set of the gas unit; Ω ES is the set of candidate installation nodes for the energy storage system, where battery energy storage and supercapacitors are considered; Ω HVAC is the set of air conditioner users; is a collection of running scenes; c GT,inv is the unit installed capacity cost of the gas unit; c ES,inv Unit installed capacity cost of energy storage system; c HVAC is the initial registration cost paid to a single air-conditioning user and the configuration cost of the response terminal; x GT,i,k With x ES,i,k are 0-1 variables representing whether the gas generator set and energy storage system are optimally configured at node i; x HVAC,i,k is a 0-1 variable representing whether to recruit air-conditioning user k under node i; P GT,i,k,max is the rated power of a single gas unit with optimal configuration at node i; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided; P ES,i,k,max is the rated charge and discharge power of the energy storage system with optimal configuration at node i; λ is the annual discount rate; is the useful life; ρ s is the probability of running scenario s; C DS,s is the running cost of scenario s.

[0026] Wherein, the minimized operating cost is:

[0027]

[0028] CDS,s -α-ε s ≤0,ε s ≥0

[0029] C DS,s =C GT +C ES +C Grid +C Loss

[0030]

[0031] In the formula, C GT , C ES , C Grid and C Loss are the interaction costs with gas generators, energy storage, upper grid, and network loss costs; P GT,i,k,t With ΔP GT,i,k,τ are the power output of the gas unit during period t and the output adjustment after calling the reserve; P ES,i,k,t With ΔP ES,i,k,τ are the power output of energy storage during period t and the output adjustment after calling the reserve; P Grid,t With π P,t are the transmission power and time-of-use electricity price between the distribution system and the upper transmission system; c Loss is the unit network loss cost; I ij,t is the current in branch ij; r ij is the resistance of branch ij; L is the set of branches in the distribution system; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided; C EFRR C is the second-level reserve cost represented by emergency frequency response reserve; FRR C is the 10-minute time scale backup service cost represented by frequency recovery backup; LFR is the load fluctuation reserve cost at a 30-minute time scale; α is the risk value VaR under the confidence level β 1-β , ε s is a non-negative auxiliary variable representing the value of the operating cost exceeding α in scenario s. DS,s When it is less than α, its value is C DS,s -α, otherwise it is 0; κ is the risk aversion coefficient, κ=0 represents a neutral attitude towards risk, and the larger the κ is, the more averse to potential risks it is.

[0032] In the second aspect, a novel multi-time scale reserve capacity optimization configuration for a power distribution system, the device comprising:

[0033] Establishing a module for establishing a multi-time scale standby service demand model;

[0034] The generation module is used to generate operation scenarios based on wind and solar load information using the Wasserstein gradient penalty generative adversarial network, and then use the improved K-means clustering to reduce the scenarios to generate a typical scenario set;

[0035] A construction module is used to construct a two-layer optimization configuration model of multi-time scale reserve capacity of distribution system based on conditional risk value based on multi-time scale reserve service demand model, and apply conditional risk value theory to control the operation risks existing in typical scenarios;

[0036] The acquisition module is used to perform equivalent transformation of the double-layer optimization configuration model and constraint conditions based on second-order cone relaxation and parameter transfer, and obtain the optimization configuration results through the solver;

[0037] The optimization configuration module is used to realize the optimization configuration of the backup resources with the minimum optimization configuration and operation cost based on the optimization configuration result.

[0038] Wherein, the establishment module includes:

[0039] The first construction submodule is used to construct a second-level time scale reserve service model to meet the emergency frequency response requirements. The modeling of the reserve service takes into account the activation signal and price data of the reserve capacity, and requires that the reserve capacity adjustment is equal to the reserve capacity adjustment.

[0040] The second construction submodule is used to construct a 10-minute time scale backup service model to realize the control behavior of making up for power shortage and smoothing load fluctuation after an accident. The backup service requires full activation within 5 minutes; the backup service is asymmetric, and the backup capacity can only be configured in one direction in one period;

[0041] The third construction submodule is used to construct a 30-minute time scale backup service model, and to adjust the backup service upward or downward in response to the net load forecast error.

[0042] In the third aspect, a novel device for optimizing the configuration of multi-time-scale spare capacity in a power distribution system is provided, the device comprising: a processor and a memory, the memory storing program instructions, the processor calling the program instructions stored in the memory to enable the device to execute any one of the methods described in the first aspect.

[0043] In a fourth aspect, a computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes any one of the methods described in the first aspect.

[0044] The beneficial effects of the technical solution provided by the present invention are:

[0045] (1) The present invention proposes a model-free scenario generation method based on Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP). This technology constructs high-precision resource scenarios such as wind and solar energy by accurately simulating and capturing the uncertainty of renewable energy and load demand. The technical advantage of this method is that it fully considers the complex characteristics of the spatiotemporal correlation between renewable energy and load, improves the technical reliability of scenario generation and the accuracy of prediction, and provides a safer and more effective decision-making basis for coping with power fluctuations.

[0046] (2) The present invention comprehensively establishes technical models for various reserve services such as emergency frequency response reserve, frequency recovery reserve and load fluctuation reserve, and deeply analyzes the multi-time scale operation characteristics of flexible reserve resources in the distribution system; by integrating long-term resource selection with short-term operation strategies, it ensures the accurate configuration of reserve capacity, effectively improves the response speed and safety of the power system in an uncertain environment, and thus significantly improves the system's safe operation capability under accident recovery and load fluctuation conditions;

[0047] (3) The present invention fully considers the potential risks in the operation of the distribution system. By introducing an optimization model based on conditional value at risk (CVaR), market operators can formulate optimal operation strategies based on different risk preferences. This not only enhances the emergency response capability of the power grid, but also provides technical guarantees for the safety and stability of the system operation, ensuring the reliable operation of the system when a high proportion of new energy is connected. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flowchart of a novel method for optimizing the allocation of multi-time scale reserve capacity in distribution systems;

[0049] Figure 2 The flowchart of the solution of the double-layer optimal allocation method of the distribution system reserve capacity based on conditional risk value;

[0050] Figure 3 Generate autocorrelation coefficient plots of scenario and historical data for PV;

[0051] Figure 4 This is the heat map of Spearman correlation coefficient of photovoltaic sites;

[0052] Figure 5 It is the probability density distribution map of wind power sites;

[0053] Figure 6 Price charts for multiple backup services;

[0054] Figure 7 Activation price charts for various alternatives;

[0055] Figure 8 Profit graphs for various alternative combinations;

[0056] Fig. 9 The following is a graph showing the percentage of various service revenues;

[0057] Fig.10 This is the cost analysis diagram of the distribution system under different risk factors κ. DETAILED DESCRIPTION

[0058] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention are described in further detail below.

[0059] The embodiment of the present invention takes into account the impact of uncertainty of new energy on the configuration of reserve capacity of new power systems, and constructs a safe and reliable method for optimizing the configuration of reserve capacity of new distribution systems by actively participating in providing multi-time scale reserve capacity through the distribution system. The method can combine the real-time situation of the distribution system, explore the frequency support capability of the distribution system for the power grid, ensure that the power grid can still operate stably and safely when a high proportion of new energy is connected, and enhance the flexibility and adaptability of the system to cope with complex operation scenarios in the future.

[0060] The novel multi-time scale backup capacity optimization configuration method for power distribution system proposed in the embodiment of the present invention is shown in Figure 1 , the method comprises the following steps:

[0061] Example 1

[0062] 101: Build a distribution system operation and management framework that takes into account multiple business formats and establish a multi-time scale reserve service demand model;

[0063] This architecture connects grid demand and supply by coordinating and integrating distributed resources, adjustable loads and energy storage. This architecture can achieve efficient management and coordination of internal resources, while also having the ability to interact with external resources in different dimensions. By solving decision-making problems at different operation stages, achieving global optimization, designing and implementing different types of distributed resource incentive mechanisms, and supporting internal resources to meet multi-format service needs such as multi-time scale backup;

[0064] Further, in step 101, the distribution system provides backup services at time scales of seconds, 10 minutes, and 30 minutes to support emergency frequency response, frequency recovery after accidents, and load fluctuations. The basic steps are:

[0065] 1011: Construct a reserve service model with a time scale of seconds to meet the emergency frequency response requirements. The modeling of this reserve service takes into account the activation signal and price data of the reserve capacity, and requires that the increase in reserve capacity is equal to the decrease in reserve capacity;

[0066] 1012: Construct a 10-minute time scale backup service model to achieve frequency control behaviors such as making up for power shortages and smoothing load fluctuations after accidents. The backup service is required to be fully activated within 5 minutes, and the backup service is asymmetric. The backup capacity can only be configured in one direction during a period of time.

[0067] 1013: Construct a 30-minute time scale backup service model, mainly to deal with net load forecast errors. The backup service can be adjusted upward or downward.

[0068] Among them, the upward adjustment reserve service includes: the upward adjustment reserve capacity and the actual response reserve power on the previous day. The downward adjustment reserve service only considers the actual activation energy of the downward adjustment reserve.

[0069] 102: Use the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) algorithm to generate operating scenarios based on wind and solar load information, and then use the improved K-means clustering method to reduce the scenarios to generate a typical scenario set;

[0070] Among them, WGAN-GP defines a generator (G) and a discriminator (D), where the generator is responsible for generating historical data that is as realistic as possible, and the discriminator is used to distinguish the generated data from the real data. The generator and the discriminator are continuously optimized during the training process, and the gradient penalty is used to stabilize the training process and ensure the Lipschitz constraint. The generator forms a mapping from random noise to the historical data curve.

[0071] The specific steps are:

[0072] 1021: Define historical sample data Ξ={ξ1,ξ2,ξ3}, where ξ1, ξ2, ξ3 represent sample vectors of load, wind power, and photovoltaic data respectively, and determine their probability distribution And generate a probability distribution The noise z is taken as the input of the generator G;

[0073] 1022: Construct training objectives for generative adversarial networks;

[0074] 1023: Introduce a gradient penalty term on the discriminator D to obtain the optimization target of the WGAN-GP method;

[0075] 1024: Use the improved K-means clustering method to reduce the scenarios and generate a set of typical scenarios.

[0076] First, k initial cluster centers are randomly selected, and each data object is classified into the nearest center according to the distance between the original data point and these centers to form k clusters. Then, new cluster centers are formed in the next iteration, and the original k clusters are used as new data points for classification, and the iteration is repeated until the cluster centers no longer change. Finally, representative typical scenario time series data are obtained as the input of the dual-layer optimization configuration model of multi-time scale reserve capacity of the distribution system.

[0077] 103: Construct a two-layer optimization configuration model for multi-time scale reserve capacity of distribution system based on conditional risk value to meet the declared multi-time scale reserve service demand, and apply conditional risk value theory to control the operation risk of typical scenario sets;

[0078] The specific steps are:

[0079] 1031: The upper-level optimization configuration model decides the optimal configuration of the backup resources, and the specific goal is to minimize the optimal configuration cost of the flexible backup resources of the distribution system;

[0080] 1032: In the lower-level model, CVaR is used to characterize the potential operating risks of the distribution system and minimize operating costs by actively managing local flexibility reserve resources.

[0081] 104: The constraints of the dual-layer optimization configuration model for multi-time scale reserve capacity of distribution system based on conditional risk value are as follows:

[0082] Among them, the constraints of the upper-level optimization configuration model are:

[0083] (1) Air conditioning users respond to capacity constraints;

[0084] (2) Installed capacity constraints of gas-fired units and energy storage.

[0085] Among them, the constraints of the lower-level running model are:

[0086] (1) The operating constraints of the gas-fired units: the reserve capacity is determined by the ramp rate and the rated power generation, and the reserve capacity must meet the reserve regulation requirements in each period;

[0087] (2) Energy storage operation constraints. Energy storage operation constraints include: energy and reserve output limits, charge and discharge operation state constraints, and the relationship between the current and previous state of charge. The energy storage reserve capacity is related to factors such as charge and discharge power and state of charge;

[0088] (3) Multi-time scale backup service demand constraints: The backup capacity provided by gas units, energy storage and air conditioning users should meet the backup service demand at the second, 10-minute and 30-minute time scales;

[0089] (4) Power balance constraint to ensure the power balance of the entire system;

[0090] (5) Safety constraints: ensuring that node voltages and branch currents of the distribution system are within safety limits.

[0091] 105: Based on second-order cone relaxation and parameter transfer, the two-layer optimization configuration model and constraint conditions are equivalently transformed, and the solution is obtained through the solver to obtain the optimization configuration result;

[0092] Among them, the embodiment of the present invention is explained by taking the GUROBI commercial solver as an example. GUROBI is a high-performance mathematical programming optimization solver that can solve linear programming (LP), mixed integer programming (MIP), quadratic programming (QP), quadratic constrained programming (QCP) and more complex optimization problems. It is widely used in various industries, such as logistics, finance, energy, and artificial intelligence. In specific implementation, other solvers can also be selected according to the needs of actual applications, and the embodiment of the present invention does not limit this.

[0093] 106: Based on the optimization configuration results, the optimal configuration of backup resources is achieved with the minimum optimization configuration and operating cost, which improves the system's response speed and ability to cope with power fluctuations and emergencies, and significantly improves the safety and stability of the power grid; in addition, this method actively manages backup resources to effectively control the operating risks of the distribution system providing multi-time scale backup services, and realizes the optimization and risk control of power grid operation in an environment with a high proportion of new energy access.

[0094] Example 2

[0095] The solution in Example 1 is further introduced below in conjunction with a specific calculation formula, as described below for details:

[0096] 201: Construct a distribution system operation and management framework that takes into account multiple business formats and establish a multi-time scale reserve service demand model;

[0097] This architecture connects grid demand and supply by coordinating and integrating distributed resources, adjustable loads and energy storage. This architecture can achieve efficient management and coordination of internal resources, while also having the ability to interact with external resources in different dimensions. By solving decision-making problems at different operation stages, achieving global optimization, designing and implementing different types of distributed resource incentive mechanisms, and supporting internal resources to meet multi-format service needs such as multi-time scale backup;

[0098] Further, in step 201, the distribution system provides backup services at time scales of seconds, 10 minutes and 30 minutes to support emergency frequency response, frequency recovery after accidents and load fluctuations. The basic steps are:

[0099] 2011: Construct a reserve service model with a time scale of seconds to meet the emergency frequency response needs. The modeling of this reserve service takes into account the activation signal and price data of the reserve service, and requires that the increase in reserve capacity is equal to the decrease in reserve capacity;

[0100] Among them, the second-level reserve supporting emergency frequency response needs to be activated within 30 seconds of the accident to prevent the frequency from further deteriorating and avoid causing serious low-frequency load shedding. And the activation needs to remain unchanged within 15 minutes after receiving the activation signal. The upper-level transmission system operator compensates the distribution system for the opportunity cost of providing reserve capacity and the energy cost after the reserve is activated. This reserve requires the same upward and downward adjustment of reserve capacity. Formula (1) represents the income from the reserved emergency frequency response reserve capacity and reserve activation of the distribution system. The first term π EFRR,t R EFRR,t is the revenue of the distribution system for providing emergency frequency response reserve capacity. The second term π EFRR+,t,τ ΔP EFRR+,t,τ and the third term π EFRR-,t,τ ΔP EFRR-,t,τ They are the energy income from activating upward and downward reserves of the distribution system.

[0101]

[0102] ΔP EFRR+,t,τ =a EFRR+,t,τ φ EFRR+,t,τ R EFRR,t (2)

[0103] ΔP EFRR-,t,τ =a EFRR-,t,τ φ EFRR-,t,τ R EFRR,t (3)

[0104] In the formula, C EFRR is the second-level reserve cost represented by the emergency frequency response reserve; t∈Τ is the time set with a resolution of 1h in a day; τ∈Γ t is a subset of Τ with a time resolution of 15 minutes; π EFRR,t is the emergency frequency response reserve capacity price during period t; R EFRR,t is the emergency frequency response reserve capacity declared by the distribution system during period t; EFRR+,t,τ With π EFRR-,t,τ is the energy price of the reserve energy; ΔP EFRR+,t,τ With ΔP EFRR-,t,τ is the energy actually activated by the up-regulation and down-regulation reserve; the coefficient φEFRR+,t,τ and φ EFRR-,t,τ The ratio of the activated energy for upward reserve and downward reserve to the declared reserve capacity; a EFRR+,t,τ with a EFRR-,t,τ It is a binary parameter indicating whether the upward and downward reserve are activated. These two parameters are determined by the historical emergency frequency response reserve demand, and the update interval Δτ is 15 minutes. If the frequency is lower than the threshold, the upward reserve is activated. At this time, a EFRR+,t,τ =1. If the frequency is higher than the threshold, the down-regulation standby is activated. EFRR-,t,τ =1.

[0105] 2012: Construct a 10-minute time-scale backup service model to achieve frequency control behaviors such as making up for power shortages and smoothing load fluctuations after accidents. Taking the frequency recovery backup after an accident as an example, this backup service requires full activation within 5 minutes and is asymmetric. Backup capacity can only be configured in one direction during a period:

[0106]

[0107] ΔP FRR+,t,τ =a FRR+,t,τ φ FRR+,t,τ R FRR+,t (5)

[0108] ΔP FRR-,t,τ =a FRR-,t,τ φ FRR-,t,τ R FRR-,t (6)

[0109] In the formula, C FRR R is the 10-minute time scale backup service cost represented by frequency recovery backup; FRR+,t With R FRR-,t is the frequency recovery reserve capacity provided by the distribution system during period t for upward and downward adjustments; π FRR,t is the price of the frequency recovery reserve capacity during period t; FRR+,t,τ and π FRR-,t,τ are the activation prices of the upward and downward reserve during period t; ΔP FRR+,t,τ and ΔP FRR-,t,τ is the actual activation amount of the up- and down-frequency recovery standby; φ FRR+,t,τ and φ FRR-,t,τ is the ratio of the frequency recovery reserve activation amount to the declared reserve capacity; a FRR+,t,τ and a FRR-,t,τ is the corresponding activation signal, Δτ is the update time interval of the activation signal

[0110] 2013: Construct a 30-minute time scale reserve service model, mainly to deal with net load forecast errors. The reserve service can be adjusted upward or downward. Among them, the upward-adjusted reserve service includes: the upward-adjusted reserve capacity and the actual response reserve power on the day before. The downward-adjusted reserve service only considers the actual activation energy of the downward-adjusted reserve.

[0111]

[0112] ΔP LFR+,t,τ =φ LFR+,t,τ R LFR+,t (8)

[0113] In the formula, C LFR is the load fluctuation reserve cost at the 30-minute time scale; π LFR+,t is to increase the price of reserve capacity; π LFR+,t,τ and π LFR-,t is the standby activation price; ΔP LFR+,t,τ and ΔP LFR-,t is the actual activation amount of the upward reserve and downward reserve; φ LFR+,t,τ It is the ratio of the load fluctuation reserve activation to the declared reserve capacity; a LFR+,t,τ and a LFR-,t,τ It is an activation signal for upregulation and downregulation.

[0114] 202: Use the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) algorithm to generate operating scenarios based on wind and solar load information, and then use the improved K-means clustering method to reduce the scenarios to generate a typical scenario set;

[0115] Among them, WGAN-GP defines a generator (G) and a discriminator (D), where the generator is responsible for generating historical data that is as realistic as possible, and the discriminator distinguishes the generated data from the real data. The generator and the discriminator are continuously optimized during the training process, and gradient penalties are used to stabilize the training process and ensure the Lipschitz constraint. The generator forms a mapping from random noise to the historical data curve.

[0116] 2021: Define historical sample data Ξ={ξ1,ξ2,ξ3}, where ξ1, ξ2, ξ3 represent sample vectors of load, wind power, and photovoltaic data respectively, and determine their probability distribution And generate a probability distribution The noise z is taken as the input of the generator G;

[0117] 2022: Constructing training objectives for generative adversarial networks;

[0118]

[0119] In the formula, It follows the probability distribution The mathematical expectation of ξ; ξ is the sample vector; G(z) is the sample vector generated by the generator G; D(ξ) and D(G(z)) are the output samples of the discriminator D corresponding to ξ and G(z), respectively.

[0120] 2023: Introduce a gradient penalty term on the discriminator D to obtain the optimization objective of the WGAN-GP method;

[0121]

[0122] In the formula, ||·|| represents the 1 norm, γ is the regularization coefficient, which is generally 10. And ε~U[0,1]; ξ is a sample vector; G(z) is a sample vector generated by the generator G; D(ξ) and D(G(z)) are the output samples of the discriminator D corresponding to ξ and G(z) respectively; It is a gradient operator. Input historical data to generate operation scenarios based on historical wind power, photovoltaic and load data.

[0123] 2024: Use the improved K-means clustering method to reduce the scenarios and generate a set of typical scenarios.

[0124] First, k initial cluster centers are randomly selected, and each data object is classified into the nearest center according to the distance between the original data point and these centers to form k clusters. Then, new cluster centers are formed in the next iteration, and the original k clusters are used as new data points for classification, and the iteration is repeated until the cluster centers no longer change. Finally, representative typical scenario time series data are obtained as the input of the dual-layer optimization configuration model of multi-time scale reserve capacity of the distribution system.

[0125] The K-Means algorithm cannot give the best number of clusters. The embodiment of the present invention selects a suitable clustering index and analyzes the clustering results within the range of cluster numbers to determine the best number of clusters. The range of cluster numbers is usually N is an integer between , and N is the total number of samples in the dataset.

[0126] The Calinski-Harabasz (CH) index is selected as the index for determining the best clustering.

[0127]

[0128] Where, T k and P kThey represent the sum of squares of the inter-class and intra-class deviations of the number of classifications k. As the value of k continues to grow, P k P k The indicator shows a downward trend. At the same time, T k The index shows an upward trend. At a certain k value, the CH index may reach its maximum value, which is usually considered to be a reference for identifying the optimal number of clusters.

[0129] 203: Construct a two-layer optimization configuration model for multi-time scale reserve capacity of distribution system based on conditional risk value to meet the declared multi-time scale reserve service demand, and apply conditional risk value theory to control the operation risks existing in typical scenario sets;

[0130] The model can be regarded as a Stackelberg game, in which the upper layer acts as a leader to deal with the optimization configuration problem; the lower layer acts as a follower to formulate the best operation plan. The upper layer optimization configuration model decides the optimal configuration of the reserve resources, and the specific goal is to minimize the optimal configuration cost of the flexible reserve resources of the distribution system. This level is subject to several logical constraints of the optimization configuration problem, some of which are associated with the lower layer operation constraints. In the lower layer model, CVaR is used to characterize the potential operation risks of the distribution system, and the optimal energy-reserve operation strategy is formulated by actively managing local flexible reserve resources to minimize the operation cost.

[0131] 2031: The upper-level optimization configuration model decides the optimal configuration of the reserve resources, and the specific goal is to minimize the optimal configuration cost of the flexible reserve resources of the distribution system;

[0132] minf1=C GT,inv +C ES,inv +C HVAC,inv +C DS (12)

[0133]

[0134]

[0135] In the formula, C GT,inv is the optimal configuration cost of the gas unit; C ES,inv is the optimal configuration cost of the energy storage system; C HVAC,inv is the optimal configuration cost of the air conditioning system, including the initial registration compensation cost paid to air conditioning users and the configuration cost of the response terminal; C DS is the annual operating cost of the distribution system; Ω GΤ is the candidate installation node set of the gas unit; Ω ES is the set of candidate installation nodes for the energy storage system, where battery energy storage and supercapacitors are considered; Ω HVAC is the set of air conditioner users; is a collection of running scenes; c GT,inv is the unit installed capacity cost of the gas unit; c ES,inv Unit installed capacity cost of energy storage system; c HVAC is the initial registration cost paid to a single air-conditioning user and the configuration cost of the response terminal; x GT,i,k With x ES,i,k are 0-1 variables representing whether the gas generator set and energy storage system are optimally configured at node i; x HVAC,i,k is a 0-1 variable representing whether to recruit air-conditioning user k under node i; P GT,i,k,max is the rated power of a single gas unit with optimal configuration at node i; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided; P ES,i,k,max is the rated charge and discharge power of the energy storage system with optimal configuration at node i; λ is the annual discount rate; is the useful life; s is the probability of running scenario s; C DS,s is the running cost of scenario s.

[0136] 2032: In the lower-level model, CVaR is used to characterize the potential operating risks of the distribution system and minimize operating costs by actively managing local flexibility reserve resources.

[0137]

[0138] C DS,s -α-ε s ≤0,ε s ≥0 (18)

[0139] C DS,s =C GT +C ES +C Grid +C Loss (19)

[0140]

[0141] In the formula, C GT , C ES , C Grid and C Loss are the interaction costs with gas generators, energy storage, upper grid, and network loss costs; P GT,i,k,t With ΔP GT,i,k,τ are the power output of the gas unit during period t and the output adjustment after calling the reserve; P ES,i,k,t With ΔP ES,i,k,τ are the power output of energy storage during period t and the output adjustment after calling the reserve; P Grid,t With πP,t are the transmission power and time-of-use electricity price between the distribution system and the upper transmission system; c Loss is the unit network loss cost; I ij,t is the current in branch ij; r ij is the resistance of branch ij; L is the set of branches in the distribution system; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided; C EFRR C is the second-level reserve cost represented by emergency frequency response reserve; FRR C is the 10-minute time scale backup service cost represented by frequency recovery backup; LFR is the load fluctuation reserve cost at the 30-minute time scale; α is the risk value VaR under the confidence level β 1-β , ε s is a non-negative auxiliary variable representing the value of the operating cost exceeding α in scenario s. DS,s When it is less than α, its value is C DS,s -α, otherwise it is 0; κ is the risk aversion coefficient, κ=0 represents a neutral attitude towards risk, and the larger the κ is, the more averse to potential risks it is.

[0142] 204: The constraints of the dual-layer optimization configuration model for multi-time scale reserve capacity of distribution system based on conditional risk value are as follows:

[0143] Among them, the constraints of the upper-level optimization configuration model are:

[0144] (1) Air conditioning user response capacity constraints:

[0145]

[0146] Where P HVAC,i,k,max is the air conditioning resource response potential extracted from the load data of air conditioning user k; different users have different response potentials, and the user's response degree is affected by the initial registration compensation cost. k,c is the response degree of user k under the initial registration compensation cost c; R HVAC,i is the spare capacity provided by the air conditioning resources under node i.

[0147] (2) Constraints on installed capacity of gas-fired units and energy storage:

[0148] 0≤K GT ≤K GT,max (25)

[0149] 0≤K ES ≤K ES,max (26)

[0150] In the formula, KGT,max With K ES,max are the maximum number of gas units and energy storage units configured at a single node; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided.

[0151] Among them, the constraints of the lower-level running model are:

[0152] (1) The operating constraints of the gas-fired units: the reserve capacity is determined by the ramp rate and the rated power generation, and the reserve capacity must meet the reserve regulation requirements in each period;

[0153] 0≤P GT,i,k,t +R GT+,i,k,t ≤ξ i,k,t P GT,i,k,max (27)

[0154] 0≤P GT,i,k,t -R GT-,i,k,t ≤ξ i,k,t P GT,i,k,max (28)

[0155]

[0156]

[0157] In the formula, R GT+,i,k,t With R GT-,i,k,t They are the upward and downward reserve capacities provided by the gas-fired units; i,k,t It is a 0-1 variable representing the operating status of the gas unit; It is an auxiliary 0-1 variable to ensure that the gas units are not adjusted up or down at the same time; ΔP GT+,i,k,τ and ΔP GT-,i,k,τ is the upward or downward adjustment of the gas unit; r U,i,k With r D,i,k They are the upward or downward climbing rates of the gas unit respectively.

[0158] (2) Energy storage operation constraints. Energy storage operation constraints include energy and reserve output limits, charge and discharge operation state constraints, and the relationship between the current and previous state of charge. The energy storage reserve capacity is related to factors such as charge and discharge power and state of charge;

[0159] 0≤P ES,d,i,k,t ≤θ i,k,t P ES,i,k,max (33)

[0160] 0≤P ES,c,i,k,t ≤(1-θ i,k,t ) ES,i,k,max (34)

[0161] 0.1E i,k,max ≤E i,k,t ≤0.9E i,k,max (35)

[0162]

[0163] E i,k,0 =E i,k,end =0.5E i,k,max (37)

[0164] 0≤R ES+,i,k,t ≤P ES,i,k,max -P ES,d,i,k,t +P ES,c,i,k,t (38)

[0165] 0≤R ES-,i,k,t ≤P ES,i,k,max -P ES,c,i,k,t +P ES,d,i,k,t (39)

[0166]

[0167] Where P ES,c,i,k,t With P ES,d,i,k,t is the charging and discharging power of energy storage k at node i during period t; θ i,k,t is a 0-1 variable that characterizes the energy storage charging and discharging state; E i,k,t is the capacity of energy storage k at node i during period t; E i,k,max is the capacity limit of energy storage; E i,k,end and E i,k,0 are the energy storage capacity at the end and the beginning of the day respectively; R ES+,i,k,t With R ES-,i,k,t are the upward and downward reserve capacities provided by energy storage k at node i during period t; η c With η d is the charging and discharging efficiency of energy storage; ΔP ES+,i,k,τ and ΔP ES-,i,k,τ is the upward and downward adjustment amount of energy storage.

[0168] (3) Multi-time scale reserve service demand constraints: The reserve capacity provided by gas units, energy storage and air conditioning users should meet the reserve service demand at the second, 10-minute and 30-minute time scales to support emergency frequency response, frequency recovery after accidents and load fluctuation control;

[0169]

[0170] In the formula, Ω GΤ is the candidate installation node set of the gas unit; Ω ESis the set of candidate installation nodes for the energy storage system, where battery energy storage and supercapacitors are considered; Ω HVAC is the set of air-conditioning users; ΔP GT+,i,k,τ is the upward adjustment of the gas unit; ΔP GT-,i,k,τ is the downward adjustment of the gas unit; ΔP ES+,i,k,τ is the upward adjustment of energy storage; ΔP ES-,i,k,τ is the downward adjustment of energy storage; ΔP HVAC,i,τ is the power adjustment of the air conditioner user; ΔP EFRR+,t,τ With ΔP EFRR-,t,τ is the energy actually activated by the up-regulation and down-regulation reserve; ΔP FRR+,t,τ and ΔP FRR-,t,τ is the actual activation amount of the frequency recovery standby by increasing and decreasing the frequency; ΔP LFR+,t,τ and ΔP LFR-,t It is the actual activation amount of upward reserve and downward reserve.

[0171] (4) Power balance constraint to ensure the power balance of the entire system;

[0172]

[0173] q j,t =Q L,j,t -Q Grid,t (47)

[0174]

[0175] Where P ij,t and Q ij,t are the active and reactive power flowing through branch ij during period t, p j,t and q j,t are the active and reactive power injected into node j during period t, P L,j,t and Q L,j,t are respectively the active and reactive power requirements of the load during period t, x ij is the reactance of branch ij, U i,t is the voltage value at node i; I ij,t is the current in branch ij; r ij is the resistance of branch ij; P PV,j,t and P WP,j,t is the power of photovoltaic and wind power at node j in the distribution system; P Grid,t and Q Grid,t It is the active and reactive power interacting between the distribution system and the upper transmission system.

[0176] (5) Safety constraints: ensuring that node voltages and branch currents of the distribution system are within safety limits.

[0177] U i,min ≤Ui,t ≤U i,max (50)

[0178] I ij,min ≤I ij,t ≤I ij,max (51)

[0179] Where U i,max and U i,min is the upper and lower limits of voltage, I ij,max and I ij,min are the upper and lower limits of the current.

[0180] 205: Based on second-order cone relaxation and parameter transfer, the two-level optimization problem is equivalently transformed and solved by the commercial solver GUROBI to obtain the optimal configuration result;

[0181] Among them, the Distflow line power flow equation shown in formula (49) is nonlinear, and its non-convex form cannot be directly solved by commercial solvers. Therefore, the second-order cone relaxation (SOCR) method is used to convert it into cone constraints.

[0182] ||2P ij,t 2Q ij,t l ij,t -v i,t ||2≤l ij,t +v i,t (52)

[0183] Where P ij,t and Q ij,t is the active and reactive power flowing through branch ij; l ij,t is the square of the current flowing through branch ij, v i,t is the square of the voltage at node i.

[0184] By converting the double-layer model into a single-layer model through parameter transfer, the solution efficiency can be improved. The conversion process of the double-layer model is explained by taking a gas unit as an example.

[0185]

[0186] In the formula, the parameter P in the upper model is GT,i,k,max Replaced by After parameter replacement, there is a nonlinear term x in equations (27)-(30) GT,i,k ξ i,k,t , The model is non-convex. Therefore, the nonlinear term x in equation (27) is GT,i,k ξ i,k,t Fortuny-Amat method is used for further linearization.

[0187] yGT,i,k,t =x GT,i,k ξ i,k,t (54)

[0188] y GT,i,k,t ≤x GT,i,k (55)

[0189] y GT,i,k,t ≤ξ i,k,t (56)

[0190] y GT,i,k,t ≥x GT,i,k +ξ i,k,t -1 (57)

[0191] In the formula, y GT,i,k,t is to replace the nonlinear term x GT,i,k ξ i,k,t 0-1 auxiliary variable; x GT,i,k is a 0-1 variable that indicates whether the gas generator set is optimally configured at node i; i,k,t It is a 0-1 variable that represents the operating status of the gas unit.

[0192] 206: Based on the spare capacity configuration results provided by the optimization model, the optimal configuration of flexible spare resources is achieved to ensure that the spare capacity at different time scales can meet the needs of real-time load fluctuations, accident recovery and new energy fluctuations.

[0193] In summary, the optimal configuration of backup resources is achieved with the minimum configuration and operating costs, which improves the system's response speed and ability to cope with power fluctuations and emergencies, and improves the safety and stability of the power grid. In addition, this method actively manages backup resources, effectively controls the operating risks of the distribution system in providing multi-time scale backup services, and achieves grid operation optimization and risk control in an environment with a high proportion of new energy access.

[0194] Example 3

[0195] The embodiments of the present invention mainly optimize the configuration of flexible resources such as energy storage and gas units in the distribution system when large-scale wind power and photovoltaic power are connected to the power grid, and guide them to participate in providing backup services.

[0196] According to the model-free scenario generation method based on historical information of wind and solar load proposed in the embodiment of the present invention, the WGAN-GP method is used to simulate the original scenarios of wind, solar and load, and then the generated operation scenarios are reduced by the improved K-means algorithm to generate typical operation scenarios after large-scale access of distributed wind power and photovoltaic power to the distribution system; the response time characteristics of flexible resources are used to meet the multi-time scale backup service requirements; on this basis, a two-layer optimization configuration model for the backup capacity of the distribution system based on conditional risk value is proposed to achieve coordinated optimization of medium- and long-term resource selection and short-term operation strategies that take into account operation risks. The solution process of the method proposed in the embodiment of the present invention is as follows: Figure 2 shown.

[0197] The embodiment of the present invention is based on a real power distribution system in northern China to verify the effectiveness of the method. The price and standby activation data in the example refer to the Ensto-e database. The remaining system parameters are shown in Table 1.

[0198] Table 1 System parameters

[0199]

[0200] Figure 3 The autocorrelation coefficient between the generated distributed photovoltaic output scenario and the historical data is shown. The autocorrelation coefficient between the generated scenario and the historical sample data is close, showing the time correlation characteristics of the real historical data and maintaining its diversity. It can be seen that the scenario generated by the WGAN-GP method can better cover the actual output scenario and characterize the correlation characteristics of the time dimension. Figure 4 The Spearman coefficients of the generated scenarios and historical data between the three PV sites are shown, and the Spearman correlation coefficients of the two are close. Figure 5 It is a scatter plot of the probability density distribution of the generated scenarios and historical data between the two wind farm sites. The probability density distribution of the generated scenarios and historical data between the two sites is generally consistent. It can be seen that the scenario generation method based on WGAN-GP can effectively characterize the correlation of spatial dimensions.

[0201] Figure 6 and Figure 7 The standby service price and standby activation price are shown. Figure 8 The revenues of various reserve service combinations in energy market and reserve market are given. The results show that providing multiple reserve services has extremely high price potential. Fig. 9 The percentages of revenues from various types of reserve services are given, with the majority of revenues from reserve services coming from the declared reserve capacity. Fig.10The optimal configuration capacity and cost analysis of flexible resources under different κ values ​​are given. As the κ value increases, the selected unit capacity increases, which shows that the more risk-averse the distribution system is, the more it needs more unit capacity to hedge uncertainty. Correspondingly, the cost increases, but in the face of the impact of uncertainty in new energy sources such as wind power and photovoltaics, the system can rely on the reasonable configuration of multi-time scale backup capacity to effectively respond to frequency fluctuations and load fluctuations. The system can meet the promised capacity in the second, 10-minute and 30-minute backup service response, which ensures that the system can quickly resume stable operation under sudden accidents or high-power disturbances, avoiding the risk of system instability caused by too low or too high frequency, and effectively preventing the system from safety risks caused by the access of new energy.

[0202] Example 4

[0203] A novel multi-time scale reserve capacity optimization configuration for distribution system, the device comprises:

[0204] Establishing a module for establishing a multi-time scale standby service demand model;

[0205] The generation module is used to generate operation scenarios based on wind and solar load information using the Wasserstein gradient penalty generative adversarial network, and then use the improved K-means clustering to reduce the scenarios to generate a typical scenario set;

[0206] A construction module is used to construct a two-layer optimization configuration model of multi-time scale reserve capacity of distribution system based on conditional risk value based on multi-time scale reserve service demand model, and apply conditional risk value theory to control the operation risks existing in typical scenarios;

[0207] The acquisition module is used to perform equivalent transformation of the double-layer optimization configuration model and constraint conditions based on second-order cone relaxation and parameter transfer, and obtain the optimization configuration results through the solver;

[0208] The optimization configuration module is used to realize the optimization configuration of the backup resources with the minimum optimization configuration and operation cost based on the optimization configuration result.

[0209] The establishment modules include:

[0210] The first construction submodule is used to construct a second-level time scale reserve service model to meet the emergency frequency response requirements. The modeling of the reserve service takes into account the activation signal and price data of the reserve capacity, and requires that the reserve capacity adjustment is equal to the reserve capacity adjustment.

[0211] The second construction submodule is used to construct a 10-minute time scale backup service model to realize the control behavior of making up for power shortage and smoothing load fluctuation after an accident. The backup service requires full activation within 5 minutes; the backup service is asymmetric, and the backup capacity can only be configured in one direction in one period;

[0212] The third construction submodule is used to construct a 30-minute time scale backup service model, and to adjust the backup service upward or downward in response to the net load forecast error.

[0213] To sum up, under the background of a high proportion of renewable energy access, the embodiments of the present invention effectively integrate flexible resources such as distributed energy storage, renewable energy, and flexible loads by constructing an optimization model based on conditional risk value, accurately configure capacity for backup needs on multiple time scales, and enhance the system's ability to respond quickly to frequency fluctuations and load fluctuations, thereby improving the overall operating stability and reliability of the power grid.

[0214] Example 5

[0215] A novel device for optimizing the configuration of multi-time scale spare capacity of a power distribution system, the device comprising: a processor and a memory, wherein program instructions are stored in the memory, and the processor calls the program instructions stored in the memory to enable the device to execute the following method steps in Example 1:

[0216] Establish a multi-timescale standby service demand model;

[0217] The Wasserstein gradient penalty generative adversarial network is used to generate operating scenarios based on wind and solar load information, and then the improved K-means clustering is used to reduce the scenarios to generate a typical scenario set, thereby obtaining the uncertainty characteristics of wind and solar loads.

[0218] Based on the multi-time scale reserve service demand model, a two-layer optimization configuration model of the multi-time scale reserve capacity of the distribution system based on conditional risk value is constructed to meet the declared multi-time scale reserve service demand, and the conditional risk value theory is applied to control the operation risks existing in typical scenarios; based on second-order cone relaxation and parameter transfer, the two-layer optimization configuration model and constraint conditions are equivalently transformed, and the optimization configuration results are obtained through the solver;

[0219] Based on the optimization configuration results, the optimal configuration of backup resources is achieved with the minimum optimization configuration and operation costs.

[0220] Among them, the multi-time scale standby service demand model is established as follows:

[0221] Construct a reserve service model at a time scale of seconds to meet the emergency frequency response needs. The modeling of this reserve service takes into account the activation signal and price data of the reserve capacity, and requires that the increase in reserve capacity is equal to the decrease in reserve capacity.

[0222] Construct a 10-minute time scale backup service model to make up for power shortages and smooth load fluctuations after an accident. The backup service is required to be fully activated within 5 minutes. The backup service is asymmetric, and backup capacity can only be configured in one direction during a period of time.

[0223] A 30-minute time scale reserve service model is constructed to adjust the reserve service upward or downward in response to net load forecast errors.

[0224] Among them, the dual-layer optimization configuration model of multi-time scale reserve capacity of distribution system based on conditional risk value is constructed as follows:

[0225] The upper-level optimization configuration model decides the optimal configuration of the backup resources, with the goal of minimizing the optimal configuration cost of the flexible backup resources of the distribution system;

[0226] In the lower-level model, CVaR is used to characterize the potential operating risks of the distribution system, and the operating costs are minimized by actively managing local flexible reserve resources.

[0227] Among them, Wasserstein gradient penalty is: introducing a gradient penalty term on the discriminator D to obtain the optimization target of the WGAN-GP method;

[0228]

[0229] In the formula, ||·|| represents the 1 norm, γ is the regularization coefficient, which is generally 10. And ε~U[0,1]; ξ is a sample vector; G(z) is a sample vector generated by the generator G; D(ξ) and D(G(z)) are the output samples of the discriminator D corresponding to ξ and G(z) respectively; It is a gradient operator. Input historical data to generate operation scenarios based on historical wind power, photovoltaic and load data.

[0230] Among them, the optimal configuration cost of minimizing the flexible reserve resources of the distribution system is:

[0231] minf1=C GT,inv +C ES,inv +C HVAC,inv +C DS

[0232]

[0233] In the formula, C GT,inv is the optimal configuration cost of the gas unit; C ES,inv is the optimal configuration cost of the energy storage system; C HVAC,invis the optimal configuration cost of the air conditioning system, including the initial registration compensation cost paid to air conditioning users and the configuration cost of the response terminal; C DS is the annual operating cost of the distribution system; Ω GΤ is the candidate installation node set of the gas unit; Ω ES is the set of candidate installation nodes for the energy storage system, where battery energy storage and supercapacitors are considered; Ω HVAC is the set of air conditioner users; is a collection of running scenes; c GT,inv is the unit installed capacity cost of the gas unit; c ES,inv Unit installed capacity cost of energy storage system; c HVAC is the initial registration cost paid to a single air-conditioning user and the configuration cost of the response terminal; x GT,i,k With x ES,i,k are 0-1 variables representing whether the gas generator set and energy storage system are optimally configured at node i; x HVAC,i,k is a 0-1 variable representing whether to recruit air-conditioning user k under node i; P GT,i,k,max is the rated power of a single gas unit with optimal configuration at node i; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided; P ES,i,k,max is the rated charge and discharge power of the energy storage system with optimal configuration at node i; λ is the annual discount rate; is the useful life; s is the probability of running scenario s; C DS,s is the running cost of scenario s.

[0234] Among them, the minimum operating cost is:

[0235]

[0236] C DS,s -α-ε s ≤0,ε s ≥0

[0237] C DS,s =C GT +C ES +C Grid +C Loss

[0238]

[0239]

[0240] In the formula, C GT , C ES , C Grid and C Lossare the interaction costs with gas generators, energy storage, upper grid, and network loss costs; P GT,i,k,t With ΔP GT,i,k,τ are the power output of the gas unit during period t and the output adjustment after calling the reserve; P ES,i,k,t With ΔP ES,i,k,τ are the power output of energy storage during period t and the output adjustment after calling the reserve; P Grid,t With π P,t are the transmission power and time-of-use electricity price between the distribution system and the upper transmission system; c Loss is the unit network loss cost; I ij,t is the current in branch ij; r ij is the resistance of branch ij; L is the set of branches in the distribution system; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided; C EFRR C is the second-level reserve cost represented by emergency frequency response reserve; FRR C is the 10-minute time scale backup service cost represented by frequency recovery backup; LFR is the load fluctuation reserve cost at a 30-minute time scale; α is the risk value VaR under the confidence level β 1-β , ε s is a non-negative auxiliary variable representing the value of the operating cost exceeding α in scenario s. DS,s When it is less than α, its value is C DS,s -α, otherwise it is 0; κκ is the risk aversion coefficient, κ=0 represents a neutral attitude towards risk, and the larger κ is, the more averse to potential risks it is.

[0241] It should be pointed out here that the device description in the above embodiment corresponds to the method description in the embodiment, and the embodiment of the present invention will not be described in detail here.

[0242] The execution subjects of the above-mentioned processor and memory can be devices with computing functions such as computers, single-chip microcomputers, and microcontrollers. In specific implementation, the embodiments of the present invention do not limit the execution subjects and are selected according to the needs of actual applications.

[0243] The data signal is transmitted between the memory and the processor via a bus, which is not described in detail in the embodiment of the present invention.

[0244] Based on the same inventive concept, an embodiment of the present invention further provides a computer-readable storage medium, the storage medium includes a stored program, and when the program is running, the device where the storage medium is located is controlled to execute the method steps in the above embodiment.

[0245] The computer-readable storage medium includes but is not limited to a flash memory, a hard disk, a solid-state drive, and the like.

[0246] It should be pointed out here that the description of the readable storage medium in the above embodiment corresponds to the description of the method in the embodiment, and the embodiment of the present invention will not be described in detail here.

[0247] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present invention are generated.

[0248] The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions may be stored in a computer-readable storage medium or transmitted via a computer-readable storage medium. The computer-readable storage medium may be any available medium that can be accessed by the computer or a data storage device such as a server or a data center that includes one or more available media. The available medium may be a magnetic medium or a semiconductor medium, etc.

[0249] Unless otherwise specified, the models of the components in the embodiments of the present invention are not limited, and any device that can perform the above functions may be used.

[0250] Those skilled in the art will appreciate that the accompanying drawing is only a schematic diagram of a preferred embodiment, and the serial numbers of the embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0251] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A novel method for optimizing the allocation of multi-time scale reserve capacity in a power distribution system, characterized in that: The method comprises: Construct a distribution system operation and management framework that takes into account multiple business formats, and establish a multi-time scale reserve service demand model; The architecture connects grid demand and supply by coordinating and integrating distributed resources, adjustable loads and energy storage. The architecture manages and coordinates internal resources, and has the ability to interact with the outside world in different dimensions. It solves decision-making problems at different operation stages, achieves global optimization, designs and implements different types of distributed resource incentive mechanisms, and supports internal resources to meet multi-format service needs such as multi-time scale backup. The Wasserstein gradient penalty generative adversarial network is used to generate operation scenarios based on wind and solar load information, and then the improved K-means clustering is used to reduce the scenarios to generate a typical scenario set; Based on the multi-time scale reserve service demand model, a two-layer optimization configuration model of the multi-time scale reserve capacity of the distribution system based on the conditional risk value is constructed, and the conditional risk value theory is applied to control the operation risks existing in typical scenarios; based on the second-order cone relaxation and parameter transfer, the two-layer optimization configuration model and constraint conditions are equivalently transformed, and the optimization configuration results are obtained through the solver; Based on the optimization configuration results, the optimal configuration of backup resources is achieved with the minimum optimization configuration and operation costs.

2. According to claim 1, a novel method for optimizing the configuration of multi-time scale reserve capacity of a power distribution system is characterized in that: The multi-time scale standby service demand model is established as follows: Construct a reserve service model at a time scale of seconds to meet the emergency frequency response needs. The modeling of this reserve service takes into account the activation signal and price data of the reserve capacity, and requires that the increase in reserve capacity is equal to the decrease in reserve capacity. Construct a 10-minute time scale backup service model to make up for power shortages and smooth load fluctuations after an accident. The backup service is required to be fully activated within 5 minutes. The backup service is asymmetric, and backup capacity can only be configured in one direction during a period of time. A 30-minute time scale reserve service model is constructed to adjust the reserve service upward or downward in response to net load forecast errors.

3. According to the novel multi-time scale reserve capacity optimization configuration method of power distribution system in claim 1, it is characterized in that: The dual-layer optimization configuration model of multi-time scale reserve capacity of distribution system based on conditional risk value is constructed as follows: The upper-level optimization configuration model decides the optimal configuration of the backup resources, with the goal of minimizing the optimal configuration cost of the flexible backup resources of the distribution system; In the lower-level model, CVaR is used to characterize the potential operating risks of the distribution system, and the operating costs are minimized by actively managing local flexible reserve resources.

4. According to the novel multi-time scale reserve capacity optimization configuration method of power distribution system in claim 1, it is characterized in that: The Wasserstein gradient penalty is: introducing a gradient penalty term on the discriminator D to obtain the optimization target of the WGAN-GP method; In the formula, ||·|| represents the 1 norm, γ is the regularization coefficient, which is generally 10. And ε~U[0,1]; ξ is a sample vector; G(z) is a sample vector generated by the generator G; D(ξ) and D(G(z)) are the output samples of the discriminator D corresponding to ξ and G(z) respectively; It is a gradient operator; by inputting historical data, it can generate operating scenarios based on historical wind power, photovoltaic and load data.

5. According to claim 3, a novel method for optimizing the allocation of multi-time scale reserve capacity in a power distribution system is characterized in that: The optimal configuration cost of minimizing the flexibility reserve resources of the distribution system is: minf1=C GT,inv +C ES,inv +C HVAC,inv +C DS In the formula, C GT,inv is the optimal configuration cost of the gas unit; C ES,inv is the optimal configuration cost of the energy storage system; C HVAC,inv is the optimal configuration cost of the air conditioning system, including the initial registration compensation cost paid to air conditioning users and the configuration cost of the response terminal; C DS is the annual operating cost of the distribution system; Ω GΤ is the candidate installation node set of the gas unit; Ω ES is the set of candidate installation nodes for the energy storage system, where battery energy storage and supercapacitors are considered; Ω HVAC is the set of air-conditioning users; s∈S is the set of operation scenarios; c GT,inv is the unit installed capacity cost of the gas unit; c ES,inv Unit installed capacity cost of energy storage system; c HVAC is the initial registration cost paid to a single air-conditioning user and the configuration cost of the response terminal; x GT,i,k With x ES,i,k are 0-1 variables representing whether the gas generator set and energy storage system are optimally configured at node i; x HVAC,i,k is a 0-1 variable representing whether to recruit air-conditioning user k under node i; P GT,i,k,max is the rated power of a single gas unit with optimal configuration at node i; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided; P ES,i,k,max is the rated charge and discharge power of the energy storage system with optimal configuration at node i; λ is the annual discount rate; is the useful life; s is the probability of running scenario s; C DS,s is the running cost of scenario s.

6. According to the novel method for optimizing the allocation of multi-time scale reserve capacity of a power distribution system according to claim 1, it is characterized in that: The minimized operating cost is: C DS,s -a-e s ≤0,e s ≥0 C DS,s =C GT +C ES +C Grid +C Loss In the formula, C GT , C ES , C Grid and C Loss They are the interaction costs with gas units, energy storage, and the upper grid, as well as the network loss costs; P GT,i,k,t With △P GT,i,k,τ are the power output of the gas unit during period t and the output adjustment after calling the reserve; P ES,i,k,t With △P ES,i,k,τ are the power output of energy storage during period t and the output adjustment after calling the reserve; P Grid,t With π P,t are the transmission power and time-of-use electricity price between the distribution system and the upper transmission system; c Loss is the unit network loss cost; I ij,t is the current in branch ij; r ij is the resistance of branch ij; L is the set of branches in the distribution system; K GT,max With K ES,max is the number of gas units and energy storage configurations to be decided; C EFRR C is the second-level reserve cost represented by emergency frequency response reserve; FRR C is the 10-minute time scale backup service cost represented by frequency recovery backup; LFR is the load fluctuation reserve cost at a 30-minute time scale; α is the risk value VaR under the confidence level β 1-β , ε s is a non-negative auxiliary variable representing the value of the operating cost exceeding α in scenario s. DS,s When it is less than α, its value is C DS,s -α, otherwise it is 0; κ is the risk aversion coefficient, κ=0 represents a neutral attitude towards risk, and the larger the κ is, the more averse to potential risks it is.

7. A novel multi-time scale reserve capacity optimization configuration for power distribution system, characterized in that: The device comprises: Establishing a module for establishing a multi-time scale standby service demand model; The generation module is used to generate operation scenarios based on wind and solar load information using the Wasserstein gradient penalty generative adversarial network, and then use the improved K-means clustering to reduce the scenarios to generate a typical scenario set; A construction module is used to construct a two-layer optimization configuration model of multi-time scale reserve capacity of distribution system based on conditional risk value based on multi-time scale reserve service demand model, and apply conditional risk value theory to control the operation risks existing in typical scenarios; The acquisition module is used to perform equivalent transformation of the double-layer optimization configuration model and constraint conditions based on second-order cone relaxation and parameter transfer, and obtain the optimization configuration results through the solver; The optimization configuration module is used to realize the optimization configuration of the backup resources with the minimum optimization configuration and operation cost based on the optimization configuration result.

8. The novel multi-time scale reserve capacity optimization configuration device for power distribution system according to claim 7 is characterized in that: The establishment module includes: The first construction submodule is used to construct a second-level time scale reserve service model to meet the emergency frequency response requirements. The modeling of the reserve service takes into account the activation signal and price data of the reserve capacity, and requires that the reserve capacity adjustment is equal to the reserve capacity adjustment. The second construction submodule is used to construct a 10-minute time scale backup service model to achieve the control behavior of making up for power shortage and smoothing load fluctuation after an accident. The backup service requires full activation within 5 minutes; the backup service is asymmetric, and the backup capacity can only be configured in one direction in one period; The third construction submodule is used to construct a 30-minute time scale backup service model, and to adjust the backup service upward or downward in response to the net load forecast error.

9. A novel device for optimizing the allocation of multi-time scale reserve capacity in a power distribution system, characterized in that: The device comprises: a processor and a memory, wherein program instructions are stored in the memory, and the processor calls the program instructions stored in the memory to enable the device to execute any one of the methods of claims 1-6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is enabled to perform the method according to any one of claims 1 to 6.

Images