Resource cluster regulation-oriented park shared energy storage distribution robust optimization configuration method

Abstract

The invention provides a resource cluster regulation-oriented park shared energy storage distribution robust optimization configuration method, which comprises the following steps of: 1, establishing a park distributed resource scheduling cluster by a cluster division method considering a source load matching degree, and generating a source load typical scene based on an optimal clustering model; 2, establishing a park shared energy storage flexible regulation and control model for the park distributed resource scheduling cluster established in the step 1; 3, based on the park shared energy storage flexible regulation and control model constructed in the step 2 and the typical source load scene generated in the step 1, establishing a park shared energy storage distribution robust optimization configuration model; 4, solving the park shared energy storage distribution robust optimization configuration model established in the step 3 based on a CCG algorithm; according to the method, the capacity and power of park shared energy storage can be reasonably configured, the necessity of park energy storage configuration is accurately evaluated from the aspects of economy and safety, and the method can adapt to a future power system distributed resource large-scale efficient access and utilization scene.

Description

Technical Field

[0001] The present invention relates to a park shared energy storage distributed blue stick optimization configuration method oriented to resource cluster regulation, belonging to the field of distributed resource optimization configuration and decision-making of power system. Background Art

[0002] With the steady progress of the "whole-county photovoltaic" pilot project, the park, as an important part of promoting the consumption of distributed photovoltaics, has significantly increased its photovoltaic penetration rate and gradually improved its clean energy supply capacity; however, the dual uncertainty of source and load after the access of distributed photovoltaics has brought huge challenges to the power supply reliability and economic operation of the park; energy storage has significant time flexibility, can flexibly adjust the power supply within the park, effectively smooth out source and load fluctuations, and has good adaptability to distributed photovoltaic access; at the same time, cluster control can improve the scheduling efficiency of massive distributed resources accessing the park and protect resource privacy, and is the main form of park operation control in the future.

[0003] Existing energy storage configuration still focuses on centralized control, while the massive access of distributed resources will increase the difficulty of centralized control of power systems and reduce the applicability of centralized energy storage configuration; and existing energy storage planning and configuration technologies for distributed control mostly focus on smart buildings and multiple feeders of distribution networks, and lack technology for shared energy storage configuration in parks using cluster control; in addition, the existing shared energy storage configuration process still adopts a centralized operation mode or a fixed capacity allocation mode, which makes it difficult to fully tap the flexibility of shared energy storage, and the random and robust optimization methods used to deal with source and load uncertainty have certain limitations, making it difficult to effectively balance the economy and reliability of shared energy storage configuration; in response to the above problems, the present invention is aimed at the widespread access of distributed resources to the park, introduces a park distributed resource cluster control architecture, establishes a park distributed resource cluster shared energy storage flexible control model, considers the park cluster carbon quota and power interaction, and proposes a park shared energy storage distributed robust optimization configuration method for distributed resource cluster control, thereby reasonably optimizing the access capacity and power of distributed energy storage, effectively improving the utilization efficiency of energy storage, and accurately evaluating the necessity of park energy storage configuration from the perspective of economy and safety. Summary of the invention

[0004] The present invention provides a distributed and optimized configuration method for shared energy storage in a park oriented to resource cluster regulation, which aims to achieve reasonable configuration of distributed energy storage under the distributed resource cluster regulation mechanism of the park, effectively improve the efficiency of energy storage utilization, and accurately evaluate the necessity of configuring energy storage in the park.

[0005] The technical solution of the present invention is a method for optimizing the configuration of shared energy storage distribution bars in a park for resource cluster control, the method comprising:

[0006] Step 1: Establish a distributed resource scheduling cluster in the park by using a cluster division method that takes into account the source-load matching degree, and generate a typical source-load scenario based on the optimal clustering model;

[0007] Step 2: Based on step 1, establish a distributed resource scheduling cluster in the park to build a flexible control model for shared energy storage in the park;

[0008] Step 3: Based on the flexible control model of the park shared energy storage constructed in step 2 and the typical source-load scenario generated in step 1, a distributed and robust optimization configuration model of the park shared energy storage is established;

[0009] Step 4: Solve the park shared energy storage distributed blue bar optimization configuration model established in step 3 based on the CCG algorithm.

[0010] Furthermore, the cluster division method taking into account the source-load matching degree in step 1 establishes a park distributed resource scheduling cluster, and generates a typical source-load scenario based on the optimal clustering model, specifically including:

[0011] Step S1-1, cluster the resources in the park according to the source-load matching and electrical distance between the resources;

[0012] Step S1-2: for the clusters divided in step S1-1, an original park cluster source-load scenario optimization classification model is established with the goal of minimizing the sum of squares due to error (SSE) within the cluster;

[0013] Step S1-3: Generate typical source-load scenarios and probabilities based on the original park cluster source-load scenario optimization classification model.

[0014] Furthermore, in step S1-1, the resources in the park are divided into clusters according to the source-load matching and electrical distance between the resources, specifically including:

[0015] Step S2-1: Treat each resource in the park as a single cluster, with a total of n clusters;

[0016] Step S2-2: Add resource i to the cluster where the adjacent resource is located, and calculate the increment Δγ of the system comprehensive index after joining the cluster. Where w1 is the weight coefficient of the electrical distance index, w2 is the weight coefficient of the source-charge matching index, is the electrical distance, is the source-charge matching degree;

[0017] Step S2-3: Determine the adjacent resource j corresponding to the maximum increment of the system comprehensive index. For Δγ>0, merge resource i and resource j into one cluster. If i=i+1, return to step S2-2 until there is no cluster to merge.

[0018] Furthermore, the original park cluster source-load scenario optimization classification model established in step S1-2 includes the following formula (1-1):

[0019]

[0020] Where: N sam is the number of original source load scenarios; N S is the number of categories, that is, N sam The original scenes are grouped into N S Class; ω k,m is the Euclidean distance between the original source load scene k and the original source load scene m; y k Is the kth original source load scenario selected as the cluster center state variable; y m is the state variable for whether the mth original source load scenario is selected as the cluster center; x k,m A state variable indicating whether the kth original source load scenario is classified as a class with the mth original source load scenario as the cluster center.

[0021] Furthermore, the source-load typical scenarios and probabilities are generated based on the original park cluster source-load scenario optimization classification model, specifically including the following formula (1-2):

[0022]

[0023] Where: α k is the photovoltaic output vector in the kth original source-load scenario; β k is the load demand vector in the kth original source-load scenario; n s is the number of original source-load scenarios in the sth category; is the initial probability of the typical source-load scenario s; is the photovoltaic output vector under the typical source-load scenario s; Load demand vector under typical source-load scenario s; x k,s A state variable indicating whether the kth original source load scenario is classified as the sth category.

[0024] Furthermore, in step 2, a distributed resource scheduling cluster in the park is established for step 1 to construct a flexible control model for shared energy storage in the park, which specifically includes:

[0025] Step S2-1: Establish a shared energy storage power and capacity dynamic allocation model. The shared energy storage power and capacity dynamic allocation model includes the following formula (1-3):

[0026]

[0027] Where: E ES Allocate capacity for shared energy storage in the park; ES Configure power for shared energy storage in the park; EES,i,t P is the shared energy storage capacity allocated to cluster i during period t; ES,i,t is the shared energy storage capacity and power allocated to cluster i during period t; is the initial amount of shared energy storage allocated to cluster i during period t; is the initial power of shared energy storage during period t; Ω DT is the scheduling period set; N col is the number of park clusters;

[0028] S2-2. For the shared energy storage power and capacity dynamic allocation model established in step S2-1, a shared energy storage power model is established. The shared energy storage power model includes the following formulas (1-4) to (1-6):

[0029]

[0030] Where: P ch,i,t is the energy storage charging power of cluster i during period t; P dis,i,t are the energy storage discharge power of cluster i in period t; P ch,i,t,max is the upper limit of energy storage charging power of cluster i in period t; P dis,i,t,max is the upper limit of energy storage discharge power of cluster i in period t; u ch,i,t is the energy storage charging state of cluster i during period t; u dis,i,t is the energy storage discharge state of cluster i during period t; Ω col is the set of campus clusters; M is a sufficiently large positive number.

[0031] S2-3. Establish a shared energy storage power balance model based on the shared energy storage power model established in step S2-2.

[0032] Furthermore, the shared energy storage power balance model includes the following formulas (1-7) to (1-9):

[0033]

[0034]

[0035] Where: N col is the number of park clusters; η ch is the charging efficiency of shared energy storage; η dis is the shared energy storage discharge efficiency; Δt is the time interval of the scheduling period.

[0036] Furthermore, in step 3, based on the flexible control model of the park shared energy storage constructed in step 2 and the typical source-load scenario generated in step 1, a distributed and robust optimization configuration model of the park shared energy storage is established, which specifically includes:

[0037] Step S3-1, taking into account the average annual investment cost of shared energy storage in the park, the cost of using shared energy storage, the cost of power supply to the cluster, and the cost of purchasing additional carbon quotas, establish a distributed robust optimization model objective function that minimizes investment and operating costs. The distributed robust optimization model objective function that minimizes investment and operating costs is shown in formula (1-10):

[0038]

[0039] Where: C In,ES is the average annual investment cost of shared energy storage in the park; x is a vector consisting of the capacity and power configuration decision variables of shared energy storage in one stage; ρ s is the probability of the typical source-load scenario s; Φ SPL is the uncertain set of typical source-load scenario probabilities; y is the vector composed of park operation decision variables; N S is the number of typical source-load scenarios; C s,i,E is the power supply cost of cluster i in the typical source-load scenario s; C s,i,CO C is the purchase cost of additional carbon quota for cluster i under the typical source-load scenario s; i,s,O,ES is the shared energy storage usage cost of cluster i;

[0040] The details of each cost are as follows:

[0041] 1) The calculation of the average annual investment cost of shared energy storage in the park includes the following formula (1-11):

[0042]

[0043] Where: r is the depreciation rate of energy storage; l ES is the energy storage service life; c PES is the investment cost per unit power of energy storage; c RES is the investment cost per unit capacity of energy storage;

[0044] 2) Cluster power supply cost calculation includes formula (1-12):

[0045]

[0046] Where: b i,g is the unit power generation cost of controllable distributed power in cluster i; P s,i,t,g is the power generation of the controllable distributed generation in cluster i during period t under the typical source-load scenario s; c t,e P is the time-of-use electricity purchase price of the upper power grid during period t; s,i,t,net is the power purchased by cluster i from the upper grid under the typical source-load scenario s; c t,ex P is the past fee for unit interaction power between different clusters during period t, paid by the party receiving the power; s,ij,tis the active power provided by cluster i to cluster j in the typical source-load scenario s; P ex,s,ji,t Compared with the active power P provided by cluster j to cluster i in typical scenario s s,ji,t The absolute value of P s,ji,t is the active power provided by cluster j to cluster i in the typical source-load scenario s;

[0047] in,

[0048] P s,ij,t +P s,ji,t =0 formula (1-13);

[0049] Since the objective function is to minimize the power supply cost, P ex,s,ji,t With P s,ji,t The relationship is expressed as a linear inequality:

[0050]

[0051] Ω DT is a set of scheduling periods;

[0052] 3) The calculation of the cost of purchasing additional carbon quotas includes the following formula (1-15):

[0053] C s,i,CO =c cb W s,i,mark Formula (1-15);

[0054] Where: c cb W is the cost of purchasing additional carbon quotas through the carbon market; s,i,mark is the additional carbon quota purchased by cluster i from the carbon market under the typical source-load scenario s;

[0055] 4) The cost calculation of shared energy storage is as shown in formula (1-16):

[0056]

[0057] Where: c O,PES P is the shared energy storage operation and maintenance cost coefficient; ch,s,i,t is the charging power of cluster i in period t under the typical source-load scenario s; P dis,s,i,t is the discharge power of cluster i in period t under the typical source-load scenario s;

[0058] The uncertainty set of typical source-load scenario probabilities is shown in formula (1-17):

[0059]

[0060] Where: θ1 is the 1-norm allowable deviation of the probability of a typical PV output scenario, θ ∞is the ∞-norm allowable deviation of the probability of typical PV output scenarios, θ1 and θ ∞ The calculation formula is shown in the following formula (2-18):

[0061]

[0062] Where: α1 is the 1-norm confidence, α ∞ is the ∞-norm confidence;

[0063] Step S3-2, in combination with step S3-1, further consider the cluster operation power conservation constraint, cluster carbon emission constraint, and controllable distributed power supply operation constraint to construct a park shared energy storage distributed blue stick optimization configuration model, the park shared energy storage distributed blue stick optimization configuration model is shown in the following formula (1-19):

[0064]

[0065] Where: y s is the vector of decision variables of the park cluster operation under the typical source-load scenario s; s,i is the vector of decision variables of cluster i under the typical source-load scenario s; s is the source-load power vector under the typical source-load scenario s; A, B, C are the coefficient matrices extracted from the model constraints; c is the coefficient vector related to the shared energy storage configuration cost, b i is the coefficient vector related to the operating cost of cluster i, c, b i It can be extracted from the model objective function; a and d are coefficient vectors extracted from the model constraints.

[0066] Furthermore, the step 4 solves the park shared energy storage distributed robust optimization configuration model established in step 3 based on the CCG algorithm, specifically including:

[0067] Step S4-1, splitting the original park shared energy storage distribution blue stick optimization configuration model into a main problem and sub-problems based on the CCG algorithm;

[0068] The main problem is shown in formula (1-20):

[0069]

[0070] Where: α is the auxiliary variable introduced; N Iter is the current iteration number; is the probability of the typical source-load scenario s under the Lth worst probability distribution identified by the subproblem;

[0071] The sub-problem model is shown in formula (1-21):

[0072]

[0073] Where: x * is the current shared energy storage configuration strategy obtained by solving the main problem;

[0074] Since y in the subproblem s and ρ s There is no coupling in the constraints, so the above sub-problem model can be further split into the following single-layer optimization model:

[0075] Sub-problem 1 is to minimize the operation cost of the park cluster under various typical source-load scenarios based on the current shared energy storage configuration strategy; the specific model is shown in formula (1-22):

[0076]

[0077] Where: f 2,s is the cluster operation cost under the typical source-load scenario s;

[0078] Sub-problem 2: The optimal cost of park operation based on Sub-problem 1 Identify the worst probability distribution of typical source-load scenarios. The specific model is shown in formula (1-23):

[0079]

[0080] Step S4-2: Set N Iter =1, The upper bound U of the optimal configuration model of shared energy storage in the park B = + inf, the next L of the park shared energy storage distribution blue bar optimization configuration model B = -inf, iterative convergence criterion ε = 0.5;

[0081] Step S4-3: Solve the main problem and obtain the current park shared energy storage configuration strategy x * , calculate the lower bound L of the shared energy storage distribution optimization model of the park B =c T x * +α * ;

[0082] Step S4-4, substitute x * Solve sub-problem 1 to obtain the campus cluster operation cost of each typical source-load scenario under the current shared energy storage configuration strategy

[0083] Step S4-5: Substitute Solve sub-problem 2 and identify the worst probability distribution of typical source-load scenarios corresponding to the current shared energy storage configuration strategy Upper bound of the distributed blue-rod optimization configuration model for shared energy storage in computing parks

[0084] Step S4-6, determine U B -L B ≤ε? If yes, end the iteration and output the optimal shared energy storage configuration strategy; if not, N Iter =N Iter +1, Return to step S4-3.

[0085] Beneficial technical effects of the present invention:

[0086] 1) This method can reasonably configure the capacity and power of shared energy storage in the park, and accurately evaluate the necessity of configuring energy storage in the park from the perspective of economy and safety, and can adapt to the large-scale and efficient access and utilization of distributed resources in the future power system;

[0087] 2) The dynamic capacity and power allocation flexible control model of shared energy storage in the park for distributed resource cluster scheduling in the park can effectively improve the utilization efficiency of shared energy storage, and at the same time significantly improve the economy of park operation through low charging and high discharging;

[0088] 3) The park shared energy storage configuration model built with the goal of minimizing the sum of investment and operating costs can obtain reasonable energy storage configuration capacity and power, improve the benefits of shared energy storage configuration, and accurately evaluate the necessity of park energy storage configuration;

[0089] 4) The distributed robust optimization method adopted can effectively deal with the uncertainty of source and load within the park, effectively balance the economy of random optimization and the robustness of robust optimization, making the shared energy storage configuration results more reasonable and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Attached Figure 1 It is a schematic diagram of the overall steps of the method of the present invention.

[0091] Attached Figure 2 It is a park cluster control architecture that takes into account the participation of shared energy storage.

[0092] Attached Figure 3 It is a typical source-load scenario and its probability obtained based on the optimal clustering model in the embodiment.

[0093] Attached Figure 4 It is the power configuration result under different capacity configuration upper limits of the shared energy storage in the embodiment.

[0094] Attached Figure 5 It is a shared energy storage power and capacity dynamic allocation strategy under typical scenario 1 in the embodiment.

[0095] Attached Figure 6 This is the shared energy storage charging and discharging strategy under typical scenario 1 in the embodiment. DETAILED DESCRIPTION

[0096] The present invention will be further described clearly and completely below in conjunction with the accompanying drawings and specific embodiments; it should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0097] The present invention proposes a method for optimizing the configuration of shared energy storage distribution bars in a park for resource cluster control, the method comprising the following steps:

[0098] Step 1: Establish a distributed resource scheduling cluster in the park by using a cluster division method that takes into account the source-load matching degree, and generate a typical source-load scenario based on the optimal clustering model;

[0099] Step 2: Based on step 1, establish a distributed resource scheduling cluster in the park to build a flexible control model for shared energy storage in the park;

[0100] Step 3: Based on the flexible control model of the park shared energy storage constructed in step 2 and the typical source-load scenario generated in step 1, a distributed and robust optimization configuration model of the park shared energy storage is established;

[0101] Step 4: Solve the park shared energy storage distributed blue bar optimization configuration model established in step 3 based on the CCG algorithm.

[0102] Example:

[0103] Step 1: Establish a distributed resource scheduling cluster in the park based on the cluster division method taking into account the source-load matching degree, and generate a typical source-load scenario based on the optimal clustering model, including:

[0104] 1) Cluster the resources in the park according to the source-load matching and electrical distance indicators between the resources, including:

[0105] Step S1-1: Treat each resource in the park as a single cluster, with a total of n clusters;

[0106] Step S1-2: Add resource i to the cluster where the adjacent resources are located, and calculate the increment Δγ of the system comprehensive index after joining the cluster. Where w1 is the weight coefficient of the electrical distance index, w2 is the weight coefficient of the source-charge matching index, is the electrical distance, is the source-charge matching degree;

[0107] Step S1-3: Determine the adjacent resource j corresponding to the maximum increment of the system comprehensive index, and if Δγ>0, merge resource i and resource j into one cluster, and if i=i+1, return to step S2-2 until there is no cluster to be merged;

[0108] 2) For the clusters divided in step 1), the original park cluster source-load scenario optimization classification model is established with the goal of minimizing the sum of squares due to error (SSE) within the cluster:

[0109]

[0110] Where: N sam is the number of original source load scenarios; N S is the number of categories, that is, N sam The original scenes are grouped into N S Class; ω k,m is the Euclidean distance between the original source load scene k and the original source load scene m; y k Is the kth original source load scenario selected as the cluster center state variable; y m is the state variable for whether the mth original source load scenario is selected as the cluster center; x k,m A state variable indicating whether the kth original source load scenario is classified as a class with the mth original source load scenario as the cluster center.

[0111] Based on the optimization results of the above model, typical scenarios and probabilities of park source load are generated, including the following formula (2-2):

[0112]

[0113] Where: α k is the photovoltaic output vector in the kth original source-load scenario; β k is the load demand vector in the kth original source-load scenario; n s is the number of original source-load scenarios in the sth category; is the initial probability of the typical source-load scenario s; is the photovoltaic output vector under the typical source-load scenario s; Load demand vector under typical source-load scenario s; x k,s A state variable indicating whether the kth original source load scenario is classified as the sth category.

[0114] In summary, based on 1), the distributed resources and users of the park are divided into three clusters, and the cluster control architecture of the park considering the participation of shared energy storage is as shown in the attached figure. Figure 2 As shown in the figure; further based on 2) the source-load typical scenario generation method, the photovoltaic and load scenarios of the park throughout the year are clustered into 6 typical scenarios. The initial probability and source-load power curve of each typical scene are shown in the attached figure. Figure 3 shown.

[0115] Step 2: Construct a flexible control model for shared energy storage in the park based on the distributed resource scheduling cluster established in step 1. The specific model is as follows:

[0116] S2-1. Establish a shared energy storage power and capacity dynamic allocation model. The shared energy storage power and capacity dynamic allocation model includes the following formula (2-3):

[0117]

[0118] Where: E ES Allocate capacity for shared energy storage in the park; ES Configure power for shared energy storage in the park; E ES,i,t P is the shared energy storage capacity allocated to cluster i during period t; ES,i,t is the shared energy storage capacity and power allocated to cluster i during period t; is the initial amount of shared energy storage allocated to cluster i during period t; is the initial power of shared energy storage during period t; Ω DT is the scheduling period set; N col is the number of park clusters;

[0119] S2-2. For the shared energy storage power and capacity dynamic allocation model established in S2-1, a shared energy storage power model is established. The shared energy storage power model includes the following formulas (2-4) to (2-6):

[0120]

[0121] Where: P ch,i,t is the energy storage charging power of cluster i during period t; P dis,i,t are the energy storage discharge power of cluster i in period t; P ch,i,t,max is the upper limit of energy storage charging power of cluster i in period t; P dis,i,t,max is the upper limit of energy storage discharge power of cluster i in period t; u ch,i,t is the energy storage charging state of cluster i during period t; u dis,i,t is the energy storage discharge state of cluster i during period t; Ω col is the set of park clusters; M is a sufficiently large positive number;

[0122] S2-3. With respect to the shared energy storage power model established in step S2-2, a shared energy storage power balance model is established. The shared energy storage power balance model includes the following formulas (2-7) to (2-9):

[0123]

[0124] Where: N col is the number of park clusters; η ch and η dis are the charging and discharging efficiencies of shared energy storage respectively; Δt is the time interval of the scheduling period.

[0125] Step 3: Based on the shared energy storage cluster control model in step 2 and the typical source-load scenario in step 1, establish a distributed and optimized configuration model for shared energy storage in the park, including:

[0126] S3-1. Taking into account the average annual investment cost of shared energy storage in the park, the cost of using shared energy storage, the cost of power supply to the cluster, and the cost of purchasing additional carbon quotas, a distributed robust optimization model objective function is established to minimize investment and operating costs. The distributed robust optimization model objective function to minimize investment and operating costs is shown in formula (2-10):

[0127]

[0128] Where: C In,ES is the average annual investment cost of shared energy storage in the park; x is a vector consisting of the capacity and power configuration decision variables of shared energy storage in one stage; ρ s is the probability of the typical source-load scenario s; Φ SPL is the uncertain set of typical source-load scenario probabilities; y is the vector composed of park operation decision variables; N S is the number of typical source-load scenarios; C s,i,E is the power supply cost of cluster i in the typical source-load scenario s; C s,i,CO C is the purchase cost of additional carbon quota for cluster i under the typical source-load scenario s; i,s,O,ES is the shared energy storage usage cost of cluster i;

[0129] The details of each cost are as follows:

[0130] 1) Average annual investment cost of shared energy storage in the park (as shown in formula (2-11)

[0131]

[0132] Where: r is the depreciation rate of energy storage; l ES is the energy storage service life; c PES and c RES are the unit power and capacity investment costs of energy storage respectively;

[0133] 2) The cluster power supply cost is shown in formula (1-12):

[0134]

[0135] Where: b i,g is the unit power generation cost of controllable distributed power in cluster i; P s,i,t,g is the power generation of the controllable distributed generation in cluster i during period t under the typical source-load scenario s; c t,e P is the time-of-use electricity purchase price of the upper power grid during period t; s,i,t,netis the power purchased by cluster i from the upper grid under the typical source-load scenario s; c t,ex P is the past fee for unit interaction power between different clusters during period t, paid by the party receiving the power; s,ij,t is the active power provided by cluster i to cluster j in the typical source-load scenario s; P ex,s,ji,t Compared with the active power P provided by cluster j to cluster i in typical scenario s s,ji,t The absolute value of P s,ji,t is the active power provided by cluster j to cluster i in the typical source-load scenario s;

[0136] in,

[0137] P s,ij,t +P s,ji,t =0(2-13);

[0138] Since the objective function is to minimize the power supply cost, P ex,s,ji,t With P s,ji,t The relationship is expressed as a linear inequality:

[0139]

[0140] 3) The cost of purchasing additional carbon quotas is shown in formula (1-15):

[0141] C s,i,CO =c cb W s,i,mark (2-15);

[0142] Where: c cb W is the cost of purchasing additional carbon quotas through the carbon market; s,i,mark is the additional carbon quota purchased by cluster i from the carbon market under the typical scenario s;

[0143] 4) The cost of using shared energy storage is shown in formula (1-16):

[0144]

[0145] Where: c O,PES P is the shared energy storage operation and maintenance cost coefficient; ch,s,i,t is the charging power of cluster i in period t under the typical source-load scenario s; P dis,s,i,t are the discharge power of cluster i in period t under the typical source-load scenario s;

[0146] The probability uncertainty set of typical source-load scenarios is as follows:

[0147]

[0148] Where: θ1 is the 1-norm allowable deviation of the probability of a typical PV output scenario, θ∞ is the ∞-norm allowable deviation of the probability of typical PV output scenarios, and the calculation formula is as follows:

[0149]

[0150] Where: α1 is the 1-norm confidence, α ∞ is the ∞-norm confidence;

[0151] S3-2. Combined with S3-1, further considering the cluster operation power conservation constraint, cluster carbon emission constraint, and controllable distributed power supply operation constraint, a park shared energy storage distributed blue stick optimization configuration model is constructed. The park shared energy storage distributed blue stick optimization configuration model is shown in the following formula (2-19):

[0152]

[0153] Where: y s is the vector of decision variables of the park cluster operation under the typical source-load scenario s; s,i is the vector of decision variables of cluster i under the typical source-load scenario s; s is the source-load power vector under the typical source-load scenario s; A, B, C are the coefficient matrices extracted from the model constraints; c is the coefficient vector related to the shared energy storage configuration cost, b i is the coefficient vector related to the operating cost of cluster i, c, b i It can be extracted from the model objective function; a and d are coefficient vectors extracted from the model constraints.

[0154] Step 4: Solve the park shared energy storage distributed blue bar optimization configuration model established in step 3 based on the CCG algorithm, as follows:

[0155] Step S4-1: Based on the CCG algorithm, the original park shared energy storage distribution blue stick optimization configuration model is divided into a main problem and sub-problems:

[0156] The main problem is shown in formula (2-20):

[0157]

[0158] Where: α is the auxiliary variable introduced; N Iter is the current iteration number; is the probability of the typical source-load scenario s under the Lth worst probability distribution identified by the subproblem;

[0159] The sub-problem model is shown in formula (2-21):

[0160]

[0161] Where: x* is the current shared energy storage configuration strategy obtained by solving the main problem;

[0162] Since y in the subproblem s and ρ s There is no coupling in the constraints, so the above sub-problem model can be further split into the following single-layer optimization model:

[0163] Sub-problem 1 is to minimize the operation cost of the park cluster under various typical source-load scenarios based on the current shared energy storage configuration strategy; the specific model is shown in formula (2-22):

[0164]

[0165] Where: f 2,s is the cluster operation cost under the typical source-load scenario s.

[0166] Sub-problem 2: The optimal cost of park operation based on Sub-problem 1 Identify the worst probability distribution of typical source-load scenarios. The specific model is shown in formula (2-23):

[0167]

[0168] S4-2. The steps for solving the optimal configuration model of shared energy storage for clusters in the park based on the CCG algorithm are as follows:

[0169] (1) Set N Iter =1, The upper bound U of the optimal configuration model of shared energy storage in the park B = + inf, the next L of the park shared energy storage distribution blue bar optimization configuration model B = -inf, iterative convergence criterion ε = 0.5;

[0170] (2) Solve the main problem and obtain the current park shared energy storage configuration strategy x * , calculate the lower bound L of the shared energy storage distribution optimization model of the park B =c T x * +α * ;

[0171] (3) Substitute x * Solve sub-problem 1 to obtain the campus cluster operation cost of each typical source-load scenario under the current shared energy storage configuration strategy

[0172] (4) Substitution Solve sub-problem 2 and identify the worst probability distribution of typical source-load scenarios corresponding to the current shared energy storage configuration strategy Upper bound of the distributed blue-rod optimization configuration model for shared energy storage in computing parks

[0173] (5) Determine U B -L B ≤ε? If yes, end the iteration and output the optimal shared energy storage configuration strategy; if not, N Iter =N Iter +1, Return to step (2).

[0174] Based on the above solution algorithm, the shared energy storage power configuration results under different shared energy storage capacity upper limits in the specific embodiment are obtained as shown in the attached figure. Figure 4 As shown in the attached Figure 4 It can be seen that as the upper limit of energy storage capacity configuration increases, its optimal configuration power also increases. When the upper limit of shared energy storage capacity increases from 1000kWh to 1200kWh, the optimal configuration power of shared energy storage increases from 280.7kW to 336.8kW, an increase of 19.99%; in addition, it can be found that under different capacity configuration upper limits, the optimal power and capacity configuration ratio of shared energy storage is stable at around 0.2807. This value can provide a reference for the dynamic access of subsequent energy storage in the park.

[0175] Furthermore, the shared energy storage operation strategy based on the dynamic allocation mechanism of power and capacity in typical scenario 1 is as follows: Figure 5 As shown in the attached Figure 5 It can be seen that based on the dynamic allocation mechanism of power and capacity, the power and capacity of shared energy storage can be optimally and reasonably allocated according to the source-load characteristics of different clusters during the operation of the park, so as to maximize the power support needs of different clusters and improve the utilization rate of shared energy storage; at the same time, the charging of shared energy storage is mainly concentrated in the 1-5, 16-18 and 24 time periods with lower electricity prices, and the discharging is mainly concentrated in the 11-15 and 19-21 time periods with higher electricity prices, thereby reducing the power supply cost of the park and improving the economic operation efficiency of the park.

[0176] Secondly, the method of the present invention is compared with the random programming and robust optimization methods. The comparison results are shown in Table 1:

[0177] Table 1

[0178]

[0179] It can be seen from Table 1 that the shared energy storage power configuration result of the method of the present invention is greater than 276.4kW of the random optimization method, but less than 283.6kW of the robust optimization method; therefore, in dealing with the uncertainty of source and load power in the park, it has higher reliability than random planning, and has better economy than robust optimization. Therefore, it can make a better balance between the economy and robustness of random planning, and it can also further adjust the economy and reliability of the shared energy storage configuration by adjusting the model confidence. Therefore, the method in this paper has a good application prospect.

[0180] In summary, based on the above specific embodiments, it can be seen that the technical solution of the present invention has the following advantages:

[0181] 1) The dynamic capacity and power allocation flexible control model of shared energy storage in the park for distributed resource cluster scheduling in the park can effectively improve the utilization efficiency of shared energy storage, and at the same time significantly improve the economy of park operation through low charging and high discharging;

[0182] 2) The park shared energy storage distributed blue bar optimization configuration model built with the goal of minimizing the sum of investment and operating costs can obtain reasonable energy storage configuration capacity and power, improve the shared energy storage configuration benefits, and accurately evaluate the necessity of park energy storage configuration;

[0183] 3) The distributed robust optimization method adopted can effectively deal with the uncertainty of source and load within the park, effectively balance the economy of random optimization and the robustness of robust optimization, making the shared energy storage configuration results more reasonable and reliable.

Claims

1. A method for optimizing the configuration of shared energy storage in a park for resource cluster control, characterized in that: The method includes: Step 1: Establish a distributed resource scheduling cluster in the park by using a cluster division method that takes into account the source-load matching degree, and generate a typical source-load scenario based on the optimal clustering model; Step 2: Based on step 1, establish a distributed resource scheduling cluster in the park to build a flexible control model for shared energy storage in the park; Step 3: Based on the flexible control model of the park shared energy storage constructed in step 2 and the typical source-load scenario generated in step 1, a distributed and robust optimization configuration model of the park shared energy storage is established; Step 4: Solve the park shared energy storage distributed blue bar optimization configuration model established in step 3 based on the CCG algorithm.

2. The method for optimizing the configuration of shared energy storage in a park for resource cluster control according to claim 1 is characterized in that: The cluster division method taking into account the source-load matching degree in step 1 establishes a distributed resource scheduling cluster in the park, and generates a typical source-load scenario based on the optimal clustering model, specifically including: Step S1-1, cluster the resources in the park according to the source-load matching and electrical distance between the resources; Step S1-2: for the clusters divided in step S1-1, an optimization classification model for the source-load scenario of the original park cluster is established with the goal of minimizing the sum of quadratic errors within the cluster; Step S1-3: Generate typical source-load scenarios and probabilities based on the original park cluster source-load scenario optimization classification model.

3. The method for optimizing the configuration of shared energy storage in a park for resource cluster control according to claim 2 is characterized in that: In step S1-1, the resources in the park are divided into clusters according to the source-load matching and electrical distance between the resources, specifically including: Step S2-1: Treat each resource in the park as a single cluster, with a total of n clusters; Step S2-2: Add resource i to the cluster where the adjacent resource is located, and calculate the increment Δγ of the system comprehensive index after joining the cluster. Where w1 is the weight coefficient of the electrical distance index, w2 is the weight coefficient of the source-charge matching index, is the electrical distance, is the source-charge matching degree; Step S2-3: Determine the adjacent resource j corresponding to the maximum increment of the system comprehensive index. For Δγ>0, merge resource i and resource j into one cluster. If i=i+1, return to step S2-2 until there is no cluster to merge.

4. The method for optimizing the configuration of shared energy storage distribution in a park for resource cluster control according to claim 2 is characterized in that: The original park cluster source-load scenario optimization classification model established in step S1-2 includes the following formula (1-1): Where: N sam is the number of original source load scenarios; N S is the number of categories, that is, N sam The original scenes are grouped into N S Class; ω k,m is the Euclidean distance between the original source load scene k and the original source load scene m; y k Is the kth original source load scenario selected as the cluster center state variable; y m is the state variable for whether the mth original source load scenario is selected as the cluster center; x k,m A state variable indicating whether the kth original source load scenario is classified as a class with the mth original source load scenario as the cluster center.

5. The method for optimizing the configuration of shared energy storage distribution in a park for resource cluster control according to claim 2 is characterized in that: The source-load typical scenarios and probabilities are generated based on the original park cluster source-load scenario optimization classification model, specifically including the following formula (1-2): Where: α k is the photovoltaic output vector in the kth original source-load scenario; β k is the load demand vector in the kth original source-load scenario; n s is the number of original source-load scenarios in the sth category; is the initial probability of the typical source-load scenario s; is the photovoltaic output vector under the typical source-load scenario s; Load demand vector under typical source-load scenario s; x k,s A state variable indicating whether the kth original source load scenario is classified as the sth category.

6. The method for optimizing the configuration of shared energy storage distribution in a park for resource cluster control according to claim 1 is characterized in that: In step 2, a distributed resource scheduling cluster in the park is established for step 1 to construct a flexible control model for shared energy storage in the park, specifically including: Step S2-1: Establish a shared energy storage power and capacity dynamic allocation model. The shared energy storage power and capacity dynamic allocation model includes the following formula (1-3): Where: E ES Allocate capacity for shared energy storage in the park; ES Configure power for shared energy storage in the park; E ES,i,t P is the shared energy storage capacity allocated to cluster i during period t; ES,i,t is the shared energy storage capacity and power allocated to cluster i during period t; is the initial amount of shared energy storage allocated to cluster i during period t; is the initial power of shared energy storage during period t; Ω DT is the scheduling period set; N col is the number of park clusters; S2-2. For the shared energy storage power and capacity dynamic allocation model established in step S2-1, a shared energy storage power model is established. The shared energy storage power model includes the following formulas (1-4) to (1-6): Where: P ch,i,t is the energy storage charging power of cluster i during period t; P dis,i,t are the energy storage discharge power of cluster i in period t; P ch,i,t,max is the upper limit of energy storage charging power of cluster i in period t; P dis,i,t,max is the upper limit of energy storage discharge power of cluster i in period t; u ch,i,t is the energy storage charging state of cluster i during period t; u dis,i,t is the energy storage discharge state of cluster i during period t; Ω col is the set of park clusters; M is a sufficiently large positive number; S2-3. Establish a shared energy storage power balance model based on the shared energy storage power model established in step S2-2.

7. The method for optimizing the configuration of shared energy storage distribution in a park for resource cluster control according to claim 6 is characterized in that: The shared energy storage power balance model includes the following formulas (1-7) to (1-9): Where: N col is the number of park clusters; η ch charging efficiency for shared energy storage; η dis is the shared energy storage discharge efficiency; Δt is the time interval of the scheduling period.

8. The method for optimizing the configuration of shared energy storage distribution in a park for resource cluster control according to claim 1 is characterized in that: In step 3, based on the flexible control model of the park shared energy storage constructed in step 2 and the typical source-load scenario generated in step 1, a distributed and robust optimization configuration model of the park shared energy storage is established, which specifically includes: Step S3-1, taking into account the average annual investment cost of shared energy storage in the park, the cost of using shared energy storage, the cost of power supply to the cluster, and the cost of purchasing additional carbon quotas, establish a distributed robust optimization model objective function that minimizes investment and operating costs. The distributed robust optimization model objective function that minimizes investment and operating costs is shown in formula (1-10): Where: C In,ES is the average annual investment cost of shared energy storage in the park; x is a vector consisting of the capacity and power configuration decision variables of shared energy storage in one stage; ρ s is the probability of the typical source-load scenario s; Φ SPL is the uncertain set of typical source-load scenario probabilities; y is the vector composed of park operation decision variables; N S is the number of typical source-load scenarios; C s,i,E is the power supply cost of cluster i in the typical source-load scenario s; C s,i,CO C is the purchase cost of additional carbon quota for cluster i under the typical source-load scenario s; i,s,O,ES is the shared energy storage usage cost of cluster i; Step S3-2, in combination with step S3-1, further consider the cluster operation power conservation constraint, cluster carbon emission constraint, and controllable distributed power supply operation constraint to construct a park shared energy storage distributed blue stick optimization configuration model, the park shared energy storage distributed blue stick optimization configuration model is shown in the following formula (1-19): Where: y s is the vector of decision variables of the park cluster operation under the typical source-load scenario s; s,i is the vector of decision variables of cluster i under the typical source-load scenario s; s is the source-load power vector under the typical source-load scenario s; A, B, C are the coefficient matrices extracted from the model constraints; c is the coefficient vector related to the shared energy storage configuration cost, b i is the coefficient vector related to the operating cost of cluster i, c, b i It can be extracted from the model objective function; a and d are coefficient vectors extracted from the model constraints.

9. The method for optimizing the configuration of shared energy storage distribution in a park for resource cluster control according to claim 8, characterized in that: The calculation of the average annual investment cost of the park shared energy storage includes the following formula (1-11): Where: r is the depreciation rate of energy storage; l ES is the energy storage service life; c PES is the investment cost per unit power of energy storage; c RES is the investment cost per unit capacity of energy storage; The calculation of the cluster power supply cost includes the following formula (1-12): Where: b i,g is the unit power generation cost of controllable distributed power in cluster i; P s,i,t,g is the power generation of the controllable distributed generation in cluster i during period t under the typical source-load scenario s; c t,e P is the time-of-use electricity purchase price of the upper power grid during period t; s,i,t,net is the power purchased by cluster i from the upper grid under the typical source-load scenario s; c t,ex P is the past fee for unit interaction power between different clusters during period t, paid by the party receiving the power; s,ij,t is the active power provided by cluster i to cluster j in the typical source-load scenario s; P ex,s,ji,t Compared with the active power P provided by cluster j to cluster i in typical scenario s s,ji,t The absolute value of P s,ji,t is the active power provided by cluster j to cluster i in the typical source-load scenario s; in, P s,ij,t +P s,ji,t =0 formula (1-13); Since the objective function is to minimize the power supply cost, P ex,s,ji,t With P s,ji,t The relationship is expressed as a linear inequality: Ω DT is a set of scheduling periods; The calculation of the additional carbon quota purchase cost includes the following formula (1-15): C s,i,CO =c cb W s,i,mark Formula (1-15); Where: c cb W is the cost of purchasing additional carbon quotas through the carbon market; s,i,mark is the additional carbon quota purchased by cluster i from the carbon market under the typical source-load scenario s; The calculation of the shared energy storage usage cost includes the following formula (1-16): Where: c O,PES P is the shared energy storage operation and maintenance cost coefficient; ch,s,i,t is the charging power of cluster i in period t under the typical source-load scenario s; P dis,s,i,t is the discharge power of cluster i in period t under the typical source-load scenario s; The uncertainty set of typical source-load scenario probabilities is shown in formula (1-17): Where: θ1 is the 1-norm allowable deviation of the probability of a typical scenario of photovoltaic output, θ ∞ is the ∞-norm allowable deviation of the probability of typical PV output scenarios, θ1 and θ ∞ The calculation formula is shown in the following formula (2-18): Where: α1 is the 1-norm confidence, α ∞ is the ∞-norm confidence.

10. The method for optimizing the configuration of shared energy storage in a park for resource cluster control according to claim 1, characterized in that: In step 4, the park shared energy storage distributed robust optimization configuration model established in step 3 is solved based on the CCG algorithm, specifically including: Step S4-1, splitting the original park shared energy storage distribution blue stick optimization configuration model into a main problem and sub-problems based on the CCG algorithm; The main problem is shown in formula (1-20): Where: α is the auxiliary variable introduced; N Iter is the current iteration number; is the probability of the typical source-load scenario s under the Lth worst probability distribution identified by the subproblem; The sub-problem model is shown in formula (1-21): Where: x * is the current shared energy storage configuration strategy obtained by solving the main problem; Since y in the subproblem s and ρ s There is no coupling in the constraints, so the above sub-problem model can be further split into the following single-layer optimization model: Sub-problem 1 is to minimize the operation cost of the park cluster under various typical source-load scenarios based on the current shared energy storage configuration strategy; the specific model is shown in formula (1-22): Where: f 2,s is the cluster operation cost under the typical source-load scenario s; Sub-problem 2: The optimal cost of park operation based on Sub-problem 1 Identify the worst probability distribution of typical source-load scenarios. The specific model is shown in formula (1-23): Step S4-2: Set N Iter =1, The upper bound U of the optimal configuration model of shared energy storage in the park B = + inf, the next L of the park shared energy storage distribution blue bar optimization configuration model B = -inf, iterative convergence criterion ε = 0.5; Step S4-3: Solve the main problem and obtain the current park shared energy storage configuration strategy x * , calculate the lower bound L of the shared energy storage distribution optimization model of the park B =c T x * +α * ; Step S4-4, substitute x * Solve sub-problem 1 to obtain the campus cluster operation cost of each typical source-load scenario under the current shared energy storage configuration strategy Step S4-5: Substitute Solve sub-problem 2 and identify the worst probability distribution of typical source-load scenarios corresponding to the current shared energy storage configuration strategy Upper bound of the distributed blue-rod optimization configuration model for shared energy storage in computing parks Step S4-6, determine U B -L B ≤ε? If yes, end the iteration and output the optimal shared energy storage configuration strategy; if not, N Iter =N Iter +1, Return to step S4-3.

Images