A power system optimization method, system and computer-readable storage medium

Abstract

This application belongs to the field of power grid optimization, and discloses a power system optimization method, system and computer-readable storage medium. The method includes: performing clustering processing with the fluctuation feature vector of the load as the element for clustering to obtain the typical scenario conventional load curve; obtaining the adjustable load amount and the tolerable delay time of the adjustable load from the typical scenario conventional load curve, and obtaining the load spatio-temporal transfer matrix according to the immediate power consumption and the delayed power consumption of the adjustable load and the immediate power consumption and the delayed power consumption after spatial transfer; establishing a two-layer programming model, where the upper-layer programming model of the two-layer programming model is oriented to minimizing the total cost of construction and operation; the lower-layer programming model of the two-layer programming model is oriented to minimizing the fluctuation of the transmission power of the superior grid connection line, and using the load spatio-temporal transfer matrix and the typical scenario conventional load curve as the constraints of the lower-layer programming model, which can achieve the effect of improving the operation stability of the power system.

Description

Technical Field

[0001] This application relates to the technical field of power grid optimization, and in particular, to a power system optimization method, system, and computer-readable storage medium. Background Art

[0002] Based on the dual-carbon goal, a huge transformation has taken place in China's energy industry. The dual-carbon goal stipulates quantitative indicators for the penetration rate and consumption rate of new energy and the carbon emissions of the power grid. To achieve the dual-carbon goal, the proportion of new energy in the power grid has gradually increased, while the power generation proportion of traditional fossil energy power generation models will gradually decrease. Due to the randomness and instability of new energy power generation, under the access of a high proportion of new energy, the traditional power grid has problems such as insufficient bearing capacity and insufficient system inertia in safe and stable operation. There are a large number of flexibility resources in the power system, and among them, the demand response of load users is an effective means to address the flexibility shortage and power supply-demand mismatch problems caused by the large-scale grid connection of new energy.

[0003] Since the beginning of this century, the rapid development of science and technology has continuously increased the network traffic and computing power requirements. The number of computing power infrastructure represented by data centers has increased dramatically around the world. In the digital age, the demand for network communication and computing response has soared, the scale of data centers has been continuously expanding, and their power consumption and energy consumption have gradually become a part that cannot be ignored. China has vigorously developed the planning and construction of data centers. At present, data centers are one of the important power loads in China. Different from traditional adjustable loads such as air conditioners that only have time transfer characteristics, data centers can have the potential for spatio-temporal transfer of loads through information transmission channels such as optical fibers. The load of the data center includes real-time response loads and delayable response loads. Through reasonable and scientific scheduling management of the delayable loads, the electricity demand can be effectively reduced during the peak load period, and the electricity demand can be increased during the low load period or the high new energy period, effectively balancing the power supply-demand imbalance problem caused by the fluctuation of new energy and improving the effective consumption rate of new energy; at the same time, the data center can realize the spatial transfer of loads through information transmission, can achieve power flow optimization and congestion management, and reduce network losses.

[0004] At present, domestic and foreign scholars have carried out a large number of studies on the participation of data centers in demand-side response, generally focusing on the energy management and operation scheduling strategies of data centers. There is no planning method for a new type of park power grid with high-penetration new energy access combined with the spatio-temporal transfer characteristics of data centers.

[0005] The inventor found that there are at least the following problems in the prior art: Due to the randomness and instability of the load, the high proportion of loads with randomness and stability access makes the traditional power grid have insufficient bearing capacity, resulting in problems such as insufficient system inertia in safe and stable operation. Summary of the Invention

[0006] To this end, embodiments of the present application provide a power system optimization method, system, and computer-readable storage medium, which can solve the technical problem of the operation stability of the existing power system. The specific technical solution is as follows:

[0007] In a first aspect, embodiments of the present application provide a power system optimization method, and the method includes:

[0008] Performing clustering processing with the fluctuation feature vector of the load as the element for clustering to obtain a typical scenario conventional load curve;

[0009] Obtaining the adjustable load amount and the tolerable delay time of the adjustable load from the typical scenario conventional load curve, and obtaining a load spatio-temporal transfer matrix according to the immediate power consumption and delayed power consumption of the adjustable load and the immediate power consumption and delayed power consumption after spatial transfer;

[0010] Establishing a two-layer programming model. The upper-layer programming model of the two-layer programming model is oriented to minimizing the total cost of construction and operation, and uses whether to build energy storage, whether to build a line, and the energy storage construction capacity as decision variables; the lower-layer programming model of the two-layer programming model is oriented to minimizing the fluctuation of the transmission power of the superior grid connection line, and uses the power of each energy storage at each scheduling moment of each typical scenario, the power of the adjustable load of each data center at each scheduling moment of each typical scenario, the transmission power of the superior grid connection line at each scheduling moment of each typical scenario, and the power flow variables at each scheduling moment of each typical scenario as decision variables. The power of the adjustable load of each data center at each scheduling moment of each typical scenario is obtained from the load spatio-temporal transfer matrix; and the typical scenario conventional load curve serves as a constraint for the lower-layer programming model.

[0011] Optionally, the typical scenario conventional load curve includes the conventional load curve of the data center. The performing clustering processing with the fluctuation feature vector of the load as the element for clustering to obtain a typical scenario conventional load curve includes:

[0012] Obtaining the number of data center servers, the rated power of the servers, and the maximum power of the servers, as well as the server utilization rate at each moment during historical operation, the adjustable load ratio at each moment of the servers, and the tolerable delay time of the adjustable load at each moment. Taking days as the unit, generating a daily server utilization rate vector n n =[η n,1 ,...,η n,T T , the server adjustable ratio vector b n =[β n,1 ,...,β n,T T and the tolerable delay time vector of the adjustable load at each moment ​​Construct a matrix with the daily server utilization rate vector, server adjustable ratio vector, and adjustable load tolerance delay time vector at each moment of the data center. Use the singular vectors of the matrix as the objects for clustering to obtain the regular load curve of the data center.

[0013] Optionally, the regular load curve of the typical scenario includes the regular load curve of new energy. The clustering process using the load fluctuation feature vector as the clustering element to obtain the regular load curve of the typical scenario includes:

[0014] Obtain the volatility, new energy dispersion, and relative peak-valley difference of the new energy power generation curve. Generate the fluctuation feature vector of the new energy power generation curve in units of days, and perform clustering on the fluctuation feature vector to obtain the regular load curve of new energy.

[0015] Optionally, the upper-layer planning model is:

[0016] In the formula: x ess,n is the decision variable for whether to build energy storage at node n, x line, b represents the decision variable for whether to build the line of branch b, E ess,n represents the capacity of energy storage construction at node n, B is the set of lines to be built, and N is the total number of network nodes; C total represents the total cost of new park power grid construction and operation, C total is:

[0017] C total = C inv + C loss + C grid

[0018] In the formula: C inv is the equivalent annual value construction cost of energy storage and lines, C loss represents the operation loss cost of the power grid, including the operation loss of energy storage and line loss, C grid represents the power purchase cost from the superior power grid; the calculation formulas for various costs are as follows:

[0019]

[0020]

[0021]

[0022]

[0023] In the formula: r is the discount rate, y m is the investment period, c ess is the investment cost per unit capacity of energy storage, cline is the line investment cost, c loss is the unit network loss power cost, c eloss is the charge and discharge loss cost per unit power of the energy storage conversion. S represents the set of typical scenarios, σ s represents the probability of the s-th typical scenario, I ij,s,t is the current of the line between nodes ij at time t in the s-th typical scenario, r ij is the resistance of the line between nodes ij, P ess,n,s,t represents the charge and discharge power of the n-node energy storage at time t in the s-th typical scenario; N line is the set of lines to be built; T represents the calculation period, B represents the set of all lines, N ess represents the set of energy storage nodes;; c grid,s,t represents the purchase price of electricity from the superior power grid at time t in the s-th typical scenario, P grid,s,t represents the power transmitted by the superior power grid at time t in the s-th typical scenario.

[0024] Optionally, the objective function of the lower-layer planning model is:

[0025]

[0026] Optionally, the constraints of the lower-layer planning model include energy storage operation constraints, power flow constraints, operating voltage constraints, branch current constraints, active power balance of electricity, and adjustable load constraints and operation constraints of the data center. The conventional load curve of the typical scenario is used as the output of the conventional load in the power flow constraint and the active power balance constraint, and is used for the adjustable load constraint of the data center in the lower-layer planning model.

[0027] Optionally, the lower-layer planning model is simplified by cone optimization.

[0028] Optionally, the power flow constraint is transformed into a cone constraint form by variable substitution.

[0029] In a second aspect, an embodiment of the present application provides a power system optimization system, the system includes:

[0030] A first calculation module, configured to perform clustering processing with the fluctuation feature vector of the load as the element of clustering to obtain a conventional load curve of a typical scenario;

[0031] A second calculation module, configured to obtain an adjustable load amount and a tolerable delay time of the adjustable load from the conventional load curve of the typical scenario, and obtain a load spatio-temporal transfer matrix according to the immediate power consumption and delay power consumption of the adjustable load and the immediate power consumption and delay power consumption after spatial transfer;

[0032] A planning module is used to establish a bilevel planning model. The upper-level planning model of the bilevel planning model aims to minimize the total cost of construction and operation, with the decisions of whether to build energy storage, whether to build lines, and the capacity of energy storage construction as decision variables. The lower-level planning model of the bilevel planning model aims to minimize the fluctuation of the transmission power of the superior grid tie line, with the power of each energy storage at each scheduling moment in each typical scenario, the power of the adjustable load of each data center at each scheduling moment in each typical scenario, the transmission power of the superior grid tie line at each scheduling moment in each typical scenario, and the power flow variables at each scheduling moment in each typical scenario as decision variables. The power of the adjustable load of each data center at each scheduling moment in each typical scenario is obtained from the load spatio-temporal transfer matrix. And the conventional load curve of the typical scenario serves as a constraint for the lower-level planning model.

[0033] In a third aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program, which when executed by a processor implements the steps of the power system optimization method described in any one of the foregoing.

[0034] In summary, compared with the prior art, the beneficial effects brought by the technical solutions provided by the embodiments of the present application at least include:

[0035] By combining the spatio-temporal transfer characteristics of data centers to study the planning method of a new type of park power grid with high-penetration new energy access, although the spatio-temporal transfer of the load of the data center is reflected at the operation level, the flexible load of the data center affects the planning scheme, and the load transfer ability of the data center can improve the flexibility of the power grid, effectively balance the power supply and demand imbalance caused by new energy fluctuations, thereby reducing the construction of equipment such as power sources, lines, and energy storage, making the planning scheme more economical. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a schematic flowchart of a power system optimization method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0037] This specific embodiment is only an interpretation of the present application and does not limit the present application. Those skilled in the art can make non-creative modifications to this embodiment as needed after reading this specification, but as long as they are within the scope of the claims of the present application, they are protected by the patent law.

[0038] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are some but not all of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the scope of protection of this application.

[0039] In addition, the term "and / or" in this application is merely an association relationship describing associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this application generally represents an "or" relationship between the associated objects before and after, unless otherwise specified.

[0040] The terms "first", "second", etc. in this application are used to distinguish identical or similar items with basically the same functions and effects. It should be understood that there is no logical or chronological dependence between "first", "second", and "nth", nor are the quantity and execution order limited.

[0041] The term "at least one" in this application means one or more, and the meaning of "multiple" means three or more. For example, multiple first positions refer to three or more first positions.

[0042] The embodiments of this application will be further described in detail below with reference to the accompanying drawings of the specification.

[0043] Referring to Figure 1 , in an embodiment of this application, an electric power system optimization method is provided. The main steps of the method are described as follows:

[0044] S1: Using the fluctuation feature vector of the load as the element for clustering processing to obtain the typical scenario conventional load curve;

[0045] S2: Obtaining the adjustable load amount and the tolerable delay time of the adjustable load from the typical scenario conventional load curve, and obtaining the load spatio-temporal transfer matrix according to the immediate power consumption and delayed power consumption of the adjustable load and the immediate power consumption and delayed power consumption after spatial transfer;

[0046] S3: Establish a bilevel programming model. The upper-level programming model of the bilevel programming model aims to minimize the total cost of construction and operation, with the decisions of whether to build energy storage, whether to build lines, and the energy storage capacity as decision variables. The lower-level programming model of the bilevel programming model aims to minimize the fluctuation of the transmission power of the superior grid tie line, with the power of each energy storage at each scheduling moment in each typical scenario, the power of the adjustable load of each data center at each scheduling moment in each typical scenario, the transmission power of the superior grid tie line at each scheduling moment in each typical scenario, and the power flow variables at each scheduling moment in each typical scenario as decision variables. The power of the adjustable load of each data center at each scheduling moment in each typical scenario is obtained from the load spatio-temporal transfer matrix. And the typical scenario conventional load curve serves as a constraint for the lower-level programming model.

[0047] Specifically, in this embodiment, the load includes new energy loads, data center loads, etc. The new energy loads include photovoltaic power generation, wind power generation, and other loads. The fluctuation feature vector is a vector representing the fluctuation characteristics of the load curve. Clustering the fluctuation feature vectors of the load can reduce the dimension of the clustering elements.

[0048] When obtaining the typical scenario conventional load curve, according to the different loads, the typical scenario conventional load curves of different typical scenarios are formed. For example, for the data center load, the conventional load curve of the data center is generated; for new energy, the conventional load curve of new energy is generated. New energy includes different sub-scenarios. For example, for photovoltaic power generation, the conventional load curve of new energy generates the conventional load curve of photovoltaic power generation in the typical scenario corresponding to photovoltaic power generation. In other implementation manners of this application, there are other typical scenarios, which will not be elaborated here.

[0049] The typical scenario conventional load curve includes the adjustable load amount and the tolerance time of the adjustable load in this typical scenario. A linear model is established for the energy consumption of the server on the same day, and an energy consumption model is established for the data center. The adjustable load of the data center mainly refers to the adjustable load of the server.

[0050] The load spatio-temporal transfer matrix describes the spatio-temporal transfer characteristics of the adjustable load. The spatio-temporal transfer characteristics refer to the characteristics of the adjustable load being transferred in time and space during scheduling, specifically manifested as the load of a certain server can be delayed in response or realized by the response of a remote server for spatio-temporal transfer.

[0051] This application studies the planning method of a new type of park power grid with high-penetration new energy access by combining the spatio-temporal transfer characteristics of the data center. Although the spatio-temporal transfer of the data center load is reflected at the operation level, the flexible load of the data center affects the planning scheme. The load transfer ability of the data center can improve the flexibility of the power grid and effectively balance the power supply-demand imbalance caused by new energy fluctuations, thereby reducing the construction of equipment such as power sources, lines, and energy storage, making the planning scheme more economical.

[0052] Further, in another embodiment, the typical scenario conventional load curve includes the conventional load curve of the data center, and S1 includes:

[0053] S11: Obtain the number of data center servers, the rated power of the servers, and the maximum power of the servers, as well as the server utilization rate at each moment, the adjustable load ratio at each moment, and the tolerable delay time of the adjustable load at each moment during historical operation. Generate a daily server utilization rate vector n of the data center in days n =[η n,1 ,...,η n,T T , the server adjustable ratio vector b n =[β n,1 ,...,β n,T T and the tolerable delay time vector of the adjustable load at each moment Construct a matrix from the daily server utilization rate vector, the server adjustable ratio vector, and the tolerable delay time vector of the adjustable load at each moment. Use the singular vectors of the matrix as the objects for clustering to obtain the conventional load curve of the data center, that is, the typical scenario of the data center, to overcome the instability of the data center load.

[0054] Further, in another embodiment, the typical scenario conventional load curve includes the conventional load curve of new energy, and S1 further includes:

[0055] S12: Obtain the volatility, new energy dispersion degree, and relative peak-valley difference of the new energy power generation curve. Generate a fluctuation characteristic vector of the new energy power generation curve in days, and perform clustering processing on the fluctuation characteristic vector to obtain the conventional load curve of new energy, that is, the typical scenario corresponding to new energy, to overcome the instability of new energy.

[0056] Further, in another embodiment, the upper-layer planning model is:

[0057] In the formula: x ess,n is the decision variable for whether to construct energy storage at node n, x line,b represents the decision variable for whether to construct a line on branch b, E ess,n ​​denotes the capacity of energy storage construction at node n, B is the set of lines to be constructed, and N is the total number of network nodes; C total The total cost of new industrial park power grid construction and operation is expressed as:

[0058] C total = C inv + C loss + C grid

[0059] In the formula: C inv is the equivalent annual construction cost of energy storage and lines, C loss represents the operation loss cost of the power grid, including the operation loss of energy storage and line loss, C grid represents the power purchase cost from the superior power grid; the calculation formulas for various costs are as follows:

[0060]

[0061]

[0062]

[0063]

[0064] In the formula: r is the discount rate, y m is the investment period, c ess is the investment cost per unit capacity of energy storage, cl ine is the line investment cost, c loss is the cost per unit of network loss power, c eloss is the conversion unit power charge and discharge loss cost of energy storage, S represents the set of typical scenarios, σ s represents the probability of the s-th typical scenario, I ij,s,t is the current of the line between nodes ij at time t in the s-th typical scenario, r ij is the resistance of the line between nodes ij, P ess,n,s,t represents the charge and discharge power of the n-node energy storage at time t in the s-th typical scenario; N line is the set of lines to be constructed; T, B, N ess respectively represent the calculation period, the set of all lines, and the set of energy storage nodes; c grid,s,t represents the power purchase price from the superior power grid at time t in the s-th typical scenario, P grid,s,t represents the power transmitted by the superior power grid at time t in the s-th typical scenario.

[0065] Furthermore, in another embodiment, the objective function of the lower-layer planning model is:

[0066]

[0067] The constraints of the lower-layer planning model include energy storage operation constraints, power flow constraints, operating voltage constraints, branch current constraints, active power balance of electricity, adjustable load constraints and operating constraints of the data center. The typical scenario conventional load curve serves as the output of the typical scenario conventional load in the power flow constraint and the active power balance constraint, and is used for the adjustable load constraint of the data center in the lower-layer planning model.

[0068] Specifically, the lower-layer planning model is simplified by cone optimization, and the power flow constraint is transformed into a cone constraint form by variable substitution.

[0069] An example of this application is as follows:

[0070] 1. Using the fluctuation feature vector of the load as the element for clustering to obtain the typical scenario conventional load curve;

[0071] Loads have uncertainties, such as new energy loads and other loads. As a type of load, the conventional load of the data center also has uncertainties and uncontrollabilities. Therefore, when planning a new type of park power grid with a high proportion of new energy access to optimize the power system resources, it is necessary to establish typical scenarios for new energy and other loads, as well as typical adjustable scenarios for the data center.

[0072] Since the daily curves of new energy, such as photovoltaic power generation and wind power generation, and other loads have a high dimension, and the curves on different days vary greatly, if the method of Euclidean distance between curves is directly used for clustering, the computational complexity is huge, and the clustering result cannot guarantee the accuracy of the clustering result, which affects the planning and optimization effects. Due to the high dimension of the original source-load curve (generally in units of 15 minutes, there are 96 points in a day, and the dimension is 96), the direct clustering effect is not good. Therefore, using some eigenvalue to describe the source-load curve can reduce the data dimension and improve the clustering effect, making the generated typical scenario more representative. In this application, a typical scenario generation method based on the fluctuation feature vector is adopted, using the fluctuation feature vector of the source-load curve as the clustering element to reduce the dimension of the clustering element. Specifically, collect the historical data of the source-load, perform data processing, and perform min-max normalization on the original data to convert it into a value between [0,1] for subsequent calculations.

[0073] In this embodiment, the index eigenvalues of the new energy fluctuation feature vector are described by the volatility, new energy dispersion, and relative peak-valley difference of the new energy generation curve. In actual operation, according to actual analysis requirements, indicators describing new energy can be appropriately added.

[0074] In this embodiment, taking photovoltaic power generation and wind power generation as examples, represents the fluctuation feature vector of the photovoltaic power generation cluster on the d-th day, denotes the fluctuation feature vector of the wind power generation cluster on the d-th day, where α PV,n and α WT,n respectively represent the n-th index describing the fluctuation feature vector of photovoltaic power generation and the n-th index describing the fluctuation feature vector of wind power generation. In this embodiment, there are three indices describing the fluctuation feature vector, that is, n = 3. The following is the index calculation method for the new energy fluctuation feature vector adopted in this application:

[0075] Volatility of new energy power generation curve:

[0076]

[0077] In the formula: P RE,t represents the output power of a certain type of new energy at time t, and P REN represents the rated value of the output power of a certain type of new energy.

[0078] New energy dispersion:

[0079]

[0080]

[0081] In the formula: E RE represents the total installed capacity of a certain type of new energy in the park, E RE,i represents the installed capacity of a certain type of new energy connected to the i-th node, and N RE represents the set of nodes in the park connected to a certain type of new energy. Equation (3) realizes the normalization of the indices of the new energy fluctuation feature vector.

[0082] Relative peak-valley difference:

[0083]

[0084] In the formula: P REmax and P REmax represent the maximum and minimum output powers of a certain new energy, and P REN represents the typical value of the output power of a certain new energy.

[0085] denotes the fluctuation feature vector of other load clusters on the d-th day, α LDnIt represents the nth index describing the fluctuation eigenvector of other load clusters, which can be described by conventional load characteristic indicators such as daily maximum load, daily minimum load, daily average load, and daily peak-valley difference, or described by utilization factors, stage factors, load density, etc. The calculation methods of the indicators have been defined in many literatures and will not be elaborated here. For the conventional non-adjustable load in the data center, the same method is used to describe the load fluctuation eigenvector. The fluctuation eigenvectors of new energy sources such as photovoltaic power generation and wind power generation, other loads, and the conventional load curves of the data center are respectively clustered to obtain the typical daily wind, light and other new energy sources, other loads, and the typical scenario conventional load curves of the data center.

[0086] The load of the data center includes adjustable load and non-adjustable load. For the non-adjustable load, the above method is used to generate the typical curve. The characteristics of the adjustable load in the data center are related to factors such as time, weather, and humanities, and there are also uncertainties and randomness. That is, the adjustable load capacity at different times is random, and the types of adjustable loads are also uncertain. Therefore, in order to consider as many uncertain scenarios as possible in the planning scheme, the adjustable load scenarios of the data center should be analyzed to generate the typical conventional load curves of the data center for power system resource optimization. Collect the relevant operation information of the data center, including the number of servers in the data center, the rated power and maximum power of the servers, the server utilization rate at each moment in historical operation, the adjustable ratio of the server load at each moment, and the tolerable delay time of the adjustable load at each moment. Generate the daily server utilization rate vector n n =[η n,1 ,...,η n,T T , the server adjustable ratio vector b n =[β n,1 ,...,β n,T T and the tolerable delay time vector of the adjustable load at each moment Construct a matrix from the above three vectors, and use the singular vectors of this matrix as the objects of clustering to obtain the typical conventional load curves of the data center.

[0087] This step generates typical curves for wind, light, other loads, and the conventional load of the data center. The double-layer planning model provides data support for the typical curves, clarifying how much the adjustable load is at each time point and what the tolerable delay time of the adjustable load is.

[0088] 2. Obtain the adjustable load amount and the tolerable delay time of the adjustable load from the typical scenario conventional load curve, and obtain the load spatio-temporal transfer matrix according to the immediate power consumption and delay power consumption of the adjustable load and the immediate power consumption and delay power consumption after spatial transfer.

[0089] ​​The data center includes servers, refrigeration equipment, lighting equipment, and other supporting equipment. Currently, numerous literatures have conducted research on the equivalent modeling of the energy consumption of servers in the data center, including linear models considering the server load conditions, non-linear CPU energy consumption models for dynamic voltage and frequency regulation, etc. Although complex energy consumption models are more accurate, their non-linearity brings great difficulties to the solution of problems. For the energy management scheduling control strategy applicable to the data center, a linear model of the energy consumption of a single server is adopted in the planning scenario of the present invention, that is:

[0090]

[0091] In the formula: is the power consumption of a single server, P n and P peak represent the rated power consumption and full-load power consumption of a single server respectively, and η represents the utilization rate of the server.

[0092] In addition to the power consumption of the servers in the data center, there are also the power consumption of refrigeration equipment, lighting, and other equipment. Therefore, the energy consumption model of the data center can be expressed as:

[0093] P dc,n,t = P server,t + P others,t (6)

[0094] In the formula: P dc,n,t represents the total energy consumption of the nth data center at time t, P server,t is the total energy consumption of the servers in the data center at time t, which is the sum of the power consumption of all single servers, and P others,t represents the other energy consumption of the data center at time t.

[0095] The refrigeration energy consumption, lighting energy consumption, etc. of the data center account for a relatively small proportion of the total energy consumption of the data center. Generally speaking, the adjustable load of the data center mainly refers to the adjustable load of the servers. The load of the servers includes real-time response load and delayable response load. The real-time response load includes real-time conversations, commercial transactions, and other user demands that require immediate response; the delayable response load includes large-scale data processing, medical image processing, etc. Therefore, the load with spatio-temporal transfer potential in the data center is the delayable response load. Different types of loads require different numbers of servers and server processing times, and the delay tolerance is also different. Therefore, the energy consumption model of the data center can be expressed as:

[0096]

[0097] Substituting into formula (6) gives:

[0098] P dc,n,t = ρ n,t Pserver,t +(1 - ρ n,t )P server,t +P others,t (8)

[0099] Wherein: represents the adjustable load of the nth data center at time t, and ρ n,t represents the adjustable proportion of the server load in the nth data center at time t.

[0100] It should be noted that other energy consumption P in the data center others,t , like other conventional loads, has randomness and non - adjustability. Therefore, it is necessary to handle its uncertainty according to the conventional load curve of a typical data center.

[0101] The spatio - temporal transfer characteristics of the delay - response load of a certain type of server can be represented by the load spatio - temporal transfer matrix T (the spatio - temporal transfer characteristics refer to the characteristics of the adjustable load being transferred in time and space during scheduling, specifically manifested as the load of a certain server can be delayed in response or transferred in time and space by the response of a remote server). For the case of a single data center, only considering the time - transfer characteristics of the load, it is expressed as:

[0102]

[0103]

[0104]

[0105] Wherein: p ii represents the power consumption of the load generated at time i with immediate response, and p ij represents the power consumption of the load demand generated at time i transferred to time j; T represents the total optimization duration; represents the tolerance delay time of the adjustable load of type k. is the total power consumption of the adjustable load generated at time t, including the parts of immediate response and delayed response.

[0106] Considering the information transmission of data centers located in different geographical locations, the delay - response load can achieve spatial transfer of the load through the information transmission function of the data center, realizing "Eastern data processed in Western regions". The spatio - temporal transfer characteristics of the adjustable load of the data center can be represented by the following load spatio - temporal transfer matrices T a and T b expressed as (taking two data centers as an example):

[0107]

[0108]

[0109]

[0110]

[0111] Wherein: and respectively represent the power consumption immediately responded by the load generated at the i-th moment in data center a and data center b. and respectively represent the power consumption of the load demand transferred to data center a and data center b at the j-th moment for response at the i-th moment. T represents the total optimization duration. represents the tolerance delay time of the adjustable load of type k. P i is the total power consumption of the load response generated at the i-th moment.

[0112] Currently, there are many studies on the forms of data centers participating in demand-side response, mainly including direct load control, electricity price guidance, and power market models. This application adopts the method of electricity price guidance to guide the load regulation of data centers through real-time electricity prices.

[0113] According to the node real-time electricity price signal, the data center operator can adjust the load. Taking two data centers as an example, when a certain adjustable load demand is generated in data center a at the t-th moment, through the node real-time electricity price signal, if the node real-time electricity price of data center b is lower, it can be transferred to data center b. If the t-th moment is the peak electricity consumption period, within the tolerance delay time range of the adjustable load, the power consumption time of the adjustable load can be reasonably allocated to reduce the electricity cost of the data center. Therefore, given the electricity price signals of each node at each time point, the data center optimizes the scheduling of the adjustable load with the goal of minimizing the electricity cost, generating the load spatio-temporal transfer matrices T a and T b . This application focuses on the impact of the adjustable load regulation potential of the data center on the planning rather than the research on the operation problems of the data center. Therefore, in the two-layer planning model involving the scheduling of the adjustable load of the data center, the prediction error of the node real-time electricity price signal is not considered.

[0114] Through the method of this step, a linear power consumption calculation model of the data center server, an adjustable load calculation model of the data center, and a load spatio-temporal transfer matrix are constructed.

[0115] 3. Establish a bilevel programming model. The upper-level programming model of the bilevel programming model aims to minimize the total cost of construction and operation, with the decisions of whether to build energy storage, whether to build lines, and the capacity of energy storage construction as decision variables; the lower-level programming model of the bilevel programming model aims to minimize the fluctuation of the transmission power of the connection line of the superior power grid, with the power of each energy storage at each scheduling moment in each typical scenario, the power of the adjustable load of each data center at each scheduling moment in each typical scenario, the transmission power of the connection line of the superior power grid at each scheduling moment in each typical scenario, and the power flow variables at each scheduling moment in each typical scenario as decision variables. The power of the adjustable load of each data center at each scheduling moment in each typical scenario is obtained from the load spatio-temporal transfer matrix; and the conventional load curve of the typical scenario is used as a constraint of the lower-level programming model.

[0116] Upper-level programming model:

[0117] The upper-level programming model is expressed as the following formula, aiming to minimize the total cost of construction and operation. The decision variables include whether to build energy storage, whether to build lines, and the capacity of energy storage construction.

[0118]

[0119] In the formula: x ess,n is the decision variable of whether to build energy storage at node n, x line,b represents the decision variable of whether to build a line on branch b, E ess,n represents the capacity of energy storage construction at node n, B is the set of lines to be built, and N is the total number of network nodes; C total represents the total cost of construction and operation of the new park power grid, expressed as:

[0120] C total = C inv + C loss + C grid (17)

[0121] In the formula: C inv is the equivalent annual value construction cost of energy storage and lines, C loss represents the operation loss cost of the power grid, including the operation loss of energy storage and line loss, C grid represents the power purchase cost from the superior power grid. The calculation formulas of various costs are as follows:

[0122]

[0123]

[0124]

[0125]

[0126] where: r is the discount rate, y m is the investment period, c ess is the investment cost per unit capacity of energy storage, c line is the line investment cost, c loss is the cost per unit of power loss, c eloss is the cost of charge and discharge losses per unit power of energy storage conversion, S represents the set of typical scenarios, σ s represents the probability of the s-th typical scenario, I ij,s,t is the current of the line between nodes ij at time t in the s-th typical scenario, r ij is the resistance of the line between nodes ij, P ess,n,s,t represents the charge and discharge power of the n-node energy storage at time t in the s-th typical scenario; N line is the set of lines to be built; T, B, N ess respectively represent the calculation period, the set of all lines, and the set of energy storage nodes; c grid,s,t represents the purchase price of electricity from the superior power grid at time t in the s-th typical scenario, P grid,s,t represents the power transmitted by the superior power grid at time t in the s-th typical scenario.

[0127] Lower-level planning model:

[0128] The lower-level planning model is expressed as the following formula. Due to the randomness and uncertainty of the output of new energy and its weak load-following ability, the large-scale access of new energy makes the net load of the park power grid fluctuate greatly. Therefore, from the perspective of suppressing the fluctuation of new energy and improving the flexibility and controllability of the park power grid, taking the connection line between the park power grid and the superior power grid as the monitoring point, with the goal of minimizing the fluctuation of the power transmitted by the superior power grid connection line, the decision variables include the power P ess,n,s,t of each energy storage at each scheduling moment in each typical scenario, the power of the adjustable load of each data center at each scheduling moment in each typical scenario the power P grid,s,t transmitted by the superior power grid connection line at each scheduling moment in each typical scenario, and the branch current I ij,s,t , the node voltage U i,s,t . The objective function can be expressed as:

[0129]

[0130] The lower-level planning model is subject to the following constraints:

[0131] 1) Energy storage operation constraints:

[0132] Taking a typical electrochemical energy storage battery as an example, its operation constraint conditions are as follows: (assuming that the charging power of the energy storage battery is positive and the discharging power is negative).

[0133] Pess,n,min ≤P ess,n,s,t ≤P ess,n,max (23)

[0134] Q ess,n,min ≤Q ess,n,s,t ≤Q ess,n,max (24)

[0135] e ess,n,s,t+1 -e ess,n,s,t =ηP ess,n,s,t+1 Δt (25)

[0136]

[0137] SOC n,min ≤SOC n,s,t ≤SOC n,max (27)

[0138] SOC n,s,0 =SOC n,s,T (28)

[0139] Wherein: P ess,n,max and P ess,n,min are respectively the upper and lower limits of the active power of the nth energy storage, Q ess,n,max and Q ess,n,min are respectively the upper and lower limits of the reactive power of the nth energy storage, SOC n,min and SOC n,max are the lower and upper limits of the state of charge of the energy storage system, η is the efficiency of the battery, Δt is the time step, e ess,n,s,t and SOC n,s,t are respectively the remaining battery capacity and the state of charge of the nth ESS at the tth moment of the sth typical scenario.

[0140] 2) Power flow constraint:

[0141]

[0142]

[0143] Wherein: i = 2, 3,..., N represents the nodes in the power grid, N i is the set of adjacent nodes of node i; U i,s,t , θ ij,s,t are respectively the voltage amplitude of node i and the phase angle difference between nodes ij at the tth moment of the sth typical scenario; G ii , B ii , G ij , B ij are respectively the self-conductance, self-susceptance, mutual conductance and mutual susceptance in the nodal admittance matrix; P wind,i,s,t , Ppv,i,s,t , P load,i,s,t , Q wind,i,s,t , Q pv,i,s,t , Q load,i,s,t respectively represent the active and reactive powers of the typical curves of wind power, photovoltaic power, and other loads at node i at time t in the s-th typical scenario; P dc,i,s,t and Q dc,i,s,t respectively represent the active and reactive powers of the data center at node i at time t in the s-th typical scenario.

[0144] 3) Operating voltage constraint:

[0145] U imin ≤ U i,s,t ≤ U imax i = 1, 2,..., N (31)

[0146] In the formula: U imax and U imin are respectively the upper and lower limits of the voltage amplitude at node i.

[0147] 4) Branch current constraint:

[0148] I ij,s,t 2 = U i,s,t 2 + U j,s,t 2 - 2U i,s,t U j,s,t cos(θ ij,s,t )(G 2 ij + B 2 ij ) ≤ I 2 ijmax (32)

[0149] In the formula: I ij,s,t is the current amplitude flowing through the branch between node i and node j at time t in the s-th typical scenario; I ijmax is the upper limit of the current amplitude of branch ij.

[0150] 5) Active power balance of electricity:

[0151]

[0152] 6) Adjustable load constraint and operating constraint of the data center:

[0153]

[0154] 0 ≤ P dc,x,s,t ≤ P dc,x,max (35)

[0155] Q dc,x,s,t = μ s,t P dc,x,s,t (36)

[0156]

[0157]

[0158]

[0159] where: N dc is the set of nodes for data center construction, is the normal load output of the s-th typical scenario at time t of the data center of node x; represents the power transferred from the adjustable load of the s-th typical scenario at time i of the data center of node x to time t, that is, the element in the load spatio-temporal transfer matrix; P dc,n,max represents the maximum power of the data center of node n. T x represents the load spatio-temporal transfer matrix of the x-th data center.

[0160] 4. Simplify the lower-layer planning model using cone optimization;

[0161] The aforementioned bi-level planning model is a high-dimensional mixed-integer non-linear bi-level optimization model, which is computationally complex, has a large amount of calculation, high solution difficulty, and long solution time. Therefore, it is necessary to simplify this bi-level planning model. The main difficulty of this bi-level planning model lies in the solution of the lower-layer planning model. The lower-layer planning model is a non-convex problem, and the power flow constraint and branch current constraint are non-linear constraints. Therefore, it is necessary to simplify the constraints to convert the lower-layer planning model into a convex problem. The present application adopts the following simplification method based on cone optimization.

[0162] For the power flow constraint, it is transformed into the form of cone constraint by means of variable substitution:

[0163]

[0164] After variable substitution, the power flow constraint is transformed into the following formula by variable conversion:

[0165]

[0166]

[0167] The operating voltage constraint and branch current constraint are rewritten as:

[0168] U 2 i,min ≤ X i,s,t ≤ U 2 i,max (39)

[0169]

[0170] For second-order cone variables, there are the following coupling relationships:

[0171] Y ij,s,t = Y ji,s,t (41)

[0172] Z ij,s,t = Z ji,s,t (42)

[0173] 2X i,s,t X j,s,t = Y ij,s,t 2 + Z ij,s,t 2 (43)

[0174] 2X i,s,t X j,s,t ≥ Y ij,s,t 2 + Z ij,s,t 2 (44)

[0175] To meet the requirements of the second-order cone model, the nonlinear relationships in equations (29), (30), and (32) are processed. The constraints (29) and (30) are relaxed into equations (37) and (38), and (32) is relaxed into equation (40). Therefore, the nonlinear constraints of the power flow equations in the original problem can be transformed into cone constraints, and the original problem is transformed into a convex problem, which can be solved using artificial intelligence algorithms or commercial solvers.

[0176] This application takes into account the spatio-temporal transfer characteristics of the flexible load in the data center in the planning of the new campus power grid, fully considers the regulation capabilities of the data center, energy storage, and other flexibility resources, and can effectively balance the power supply-demand imbalance caused by the uncertainty of new energy, making the optimization effect of the power system resources better. This application analyzes the spatio-temporal transfer characteristics of the data center based on the electricity price signal guidance, considers the response time of different types of adjustable loads and the tolerance delay time of the adjustable load, and proposes a data center load spatio-temporal transfer model and typical scenarios suitable for the planning scenario.

[0177] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0178] In an embodiment of this application, a power system optimization system is provided, which corresponds one-to-one to the power system optimization method in the above embodiment. The power system optimization system includes:

[0179] The first calculation module is used to perform clustering processing with the fluctuation feature vector of the load as the clustering element to obtain the typical scenario regular load curve;

[0180] The second calculation module is used to obtain the adjustable load amount and the tolerable delay time of the adjustable load from the typical scenario regular load curve, and obtain the load spatio-temporal transfer matrix according to the immediate power consumption and delayed power consumption of the adjustable load and the immediate power consumption and delayed power consumption after spatial transfer;

[0181] The planning module is used to establish a two-layer planning model. The upper-level planning model of the two-layer planning model is oriented to minimizing the total cost of construction and operation, and the decision variables are whether to build energy storage, whether to build a line, and the energy storage construction capacity; the lower-level planning model of the two-layer planning model is oriented to minimizing the fluctuation of the transmission power of the superior grid connection line, and the decision variables are the power of each energy storage at each scheduling moment of each typical scenario, the power of the adjustable load of each data center at each scheduling moment of each typical scenario, the transmission power of the superior grid connection line at each scheduling moment of each typical scenario, and the power flow variables at each scheduling moment of each typical scenario. The power of the adjustable load of each data center at each scheduling moment of each typical scenario is obtained from the load spatio-temporal transfer matrix; and the typical scenario regular load curve is used as a constraint of the lower-level planning model.

[0182] Further, in another embodiment, the first calculation module is further used to: obtain the number of data center servers, the rated power of the servers and the maximum power of the servers, as well as the server utilization rate at each moment in historical operation, the adjustable load ratio at each moment, and the tolerable delay time of the adjustable load at each moment. Generate a daily server utilization rate vector n of the data center in days n =[η n,1 ,...,η n,T T , the server adjustable ratio vector b n =[β n,1 ,...,β n,T T And the tolerable delay time vector of the adjustable load at each moment Construct a matrix from the daily server utilization rate vector, the server adjustable ratio vector, and the tolerable delay time vector of the adjustable load at each moment, and use the singular vectors of the matrix as the objects of clustering to obtain the regular load curve of the data center.

[0183] Further, in another embodiment, the first calculation module is further used to: obtain the volatility, the new energy dispersion degree, and the relative peak-valley difference of the new energy power generation curve, generate a new energy power generation curve fluctuation feature vector in days, and perform clustering processing on the fluctuation feature vector to obtain the regular load curve of the new energy. ​​

[0184] Furthermore, in another embodiment, the upper-layer planning model is as follows:

[0185] where: x ess,n is the decision variable for whether to construct energy storage at node n, and x line,b represents the decision variable for whether to construct the line of branch b. E ess,n represents the capacity of energy storage construction at node n, B is the set of lines to be constructed, and N is the total number of network nodes; C total represents the total cost of new park power grid construction and operation as follows:

[0186] C total = C inv + C loss + C grid

[0187] where: C inv is the equivalent annual value construction cost of energy storage and lines, C loss represents the operation loss cost of the power grid, including the operation loss of energy storage and line loss, and C grid represents the power purchase cost from the superior power grid; the calculation formulas for various costs are as follows:

[0188]

[0189]

[0190]

[0191]

[0192] where: r is the discount rate, y m is the investment period, c ess is the investment cost per unit capacity of energy storage, c line is the line investment cost, c loss is the cost per unit of network loss electricity, c eloss is the conversion unit power charge and discharge loss cost of energy storage, S represents the set of typical scenarios, and σ s represents the probability of the s-th typical scenario, I ij,s,t is the current of the line between nodes ij at time t in the s-th typical scenario, r ij is the resistance of the line between nodes ij, P ess,n,s,t represents the charge and discharge power of the energy storage at node n at time t in the s-th typical scenario; T, B, N ess respectively represent the calculation period, the set of all lines, and the set of energy storage nodes; c grid,s,t represents the power purchase price from the superior power grid at time t in the s-th typical scenario, P grid,s,tIt represents the power transmitted by the superior power grid at the t-th moment of the s-th typical scenario.

[0193] Furthermore, in another embodiment, the objective function of the lower-layer planning model is:

[0194]

[0195] Furthermore, in another embodiment, the constraints of the lower-layer planning model include energy storage operation constraints, power flow constraints, operating voltage constraints, branch current constraints, active power balance of electricity, and adjustable load constraints and operation constraints of the data center. The conventional load curve of the typical scenario is used as the output of the conventional load in the power flow constraint and the active power balance constraint of electricity, and is used for the adjustable load constraint of the data center in the lower-layer planning model.

[0196] Furthermore, in another embodiment, cone optimization is used to simplify the lower-layer planning model.

[0197] Furthermore, in another embodiment, the power flow constraint is transformed into a cone constraint form by means of variable substitution.

[0198] All or part of the various modules of the above-mentioned power system optimization system can be implemented by software, hardware, and their combinations. The above-mentioned modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above-mentioned various modules.

[0199] In an embodiment of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps of the power system optimization method described in the above embodiment. The computer-readable storage medium includes ROM (Read-Only Memory), RAM (Random-Access Memory), CD-ROM (Compact Disc Read-Only Memory), magnetic disks, floppy disks, etc.

[0200] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above-mentioned functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the system described in the present application is divided into different functional units or modules to complete all or part of the functions described above.

Claims

1. An electric power system optimization method, characterized in that, The method includes: Performing clustering processing with the fluctuation feature vector of the load as the clustering element to obtain the typical scenario conventional load curve; Obtaining the adjustable load amount and the tolerable delay time of the adjustable load from the typical scenario conventional load curve, and obtaining the load spatio-temporal transfer matrix according to the instantaneous power consumption and the delayed power consumption of the adjustable load and the instantaneous power consumption and the delayed power consumption after spatial transfer; A bi-level programming model is established. The upper-level programming model of the bi-level programming model aims to minimize the total cost of construction and operation, with the decision variables being whether to build energy storage, whether to build lines, and the capacity of energy storage construction; the lower-level programming model of the bi-level programming model aims to minimize the fluctuation of the transmission power of the superior grid tie line, with the decision variables being the power of each energy storage at each scheduling moment in each typical scenario, the power of the adjustable load of each data center at each scheduling moment in each typical scenario, the transmission power of the superior grid tie line at each scheduling moment in each typical scenario, and the power flow variables at each scheduling moment in each typical scenario. The power of the adjustable load of each data center at each scheduling moment in each typical scenario is obtained from the load spatio-temporal transfer matrix; and the typical scenario conventional load curve serves as a constraint for the lower-level programming model. Specifically, the upper-level programming model is as follows: Where: x ess,n is the decision variable for whether to build energy storage at node n, and x line,b represents the decision variable for whether to build the line of branch b. E ess,n represents the capacity of energy storage built at node n. N line is the set of lines to be built, and N is the total number of network nodes; C total represents the total cost of the construction and operation of the new industrial park power grid. C total is: C total = C inv + C loss + C grid Where: C inv is the equivalent annual construction cost of energy storage and lines, and C loss represents the operating loss cost of the power grid, including the operating losses of energy storage and line losses, and C grid represents the power purchase cost from the superior power grid; the calculation formulas for various costs are as follows: Where: r is the discount rate, y m is the investment period, c ess is the investment cost per unit capacity of energy storage, c line is the line investment cost, c loss is the cost per unit of power loss electricity, c eloss is the cost of charge and discharge loss per unit power of energy storage conversion, S represents the set of typical scenarios, σ s represents the probability of the s-th typical scenario, I ij,s,t is the current of the line between nodes ij at time t in the s-th typical scenario, r ij is the resistance of the line between nodes ij, P ess,n,s,t represents the charge and discharge power of the n-node energy storage at time t in the s-th typical scenario; N line is the set of lines to be built; T represents the calculation period, B represents the set of all lines, N ess represents the set of energy storage nodes; c grid,s,t represents the power purchase price from the superior power grid at time t in the s-th typical scenario, P grid,s,t represents the power transmitted by the superior power grid at time t in the s-th typical scenario; The objective function of the lower-layer planning model is: Specifically, the constraints of the lower-layer planning model include energy storage operation constraints, power flow constraints, operating voltage constraints, branch current constraints, active power balance of electricity, adjustable load constraints and operation constraints of the data center. The typical scenario conventional load curve serves as the output of the typical scenario conventional load in the power flow constraints and the active power balance constraints of electricity, as well as the adjustable load constraints of the data center for the lower-layer planning model.

2. The power system optimization method according to claim 1, characterized in that, The typical scenario conventional load curve includes the conventional load curve of the data center. The performing clustering processing with the fluctuation feature vector of the load as the clustering element to obtain the typical scenario conventional load curve includes: Obtain the number of servers in the data center, the rated power of the servers, and the maximum power of the servers, as well as the server utilization rate at each moment during historical operation, the adjustable proportion of the server load at each moment, and the tolerable delay time of the adjustable load at each moment. Generate a daily server utilization rate vector \(n\) of the data center in days. n = [\(\eta\) n,1 ,...,\(\eta\) n,T T , the server adjustable proportion vector \(b\) n = [\(\beta\) n,1 ,...,\(\beta\) n,T T and the tolerable delay time vector of the adjustable load at each moment Construct a matrix from the daily server utilization rate vector, the server adjustable proportion vector, and the tolerable delay time vector of the adjustable load at each moment. Use the singular vectors of the matrix as the objects for clustering to obtain the conventional load curve of the data center.​​ 3. The power system optimization method according to claim 2, characterized in that The typical scenario conventional load curve includes the conventional load curve of new energy. The performing clustering processing with the fluctuation feature vector of the load as the clustering element to obtain the typical scenario conventional load curve includes: Obtaining the volatility, the new energy dispersion degree, and the relative peak-valley difference of the new energy power generation curve, generating the fluctuation feature vector of the new energy power generation curve in units of days, and performing clustering processing on the fluctuation feature vector to obtain the conventional load curve of new energy.

4. The power system optimization method according to claim 1, characterized in that, Using cone optimization to simplify the lower-layer planning model.

5. The power system optimization method according to claim 4, wherein Converting the power flow constraint into a cone constraint form by means of variable substitution.

6. An electric power system optimization system, characterized in that, The system includes: A first calculation module, configured to perform clustering processing with the fluctuation feature vector of the load as the clustering element to obtain the typical scenario conventional load curve; A second calculation module, configured to obtain the adjustable load amount and the tolerable delay time of the adjustable load from the typical scenario conventional load curve, and obtain the load spatio-temporal transfer matrix according to the instantaneous power consumption and the delayed power consumption of the adjustable load and the instantaneous power consumption and the delayed power consumption after spatial transfer; The planning module is used to establish a bilevel programming model. The upper-level programming model of the bilevel programming model aims to minimize the total cost of construction and operation, with the decisions of whether to build energy storage, whether to build lines, and the energy storage capacity as decision variables. The lower-level programming model of the bilevel programming model aims to minimize the fluctuation of the transmission power of the upper-level grid tie line, with the power of each energy storage at each scheduling moment in each typical scenario, the power of the adjustable load of each data center at each scheduling moment in each typical scenario, the transmission power of the upper-level grid tie line at each scheduling moment in each typical scenario, and the power flow variables at each scheduling moment in each typical scenario as decision variables. The power of the adjustable load of each data center at each scheduling moment in each typical scenario is obtained from the load spatio-temporal transfer matrix. And the conventional load curve of the typical scenario is used as a constraint for the lower-level programming model. Specifically, the upper-level programming model is as follows: where: x ess,n is the decision variable for whether to construct energy storage at node n, and x line,b represents the decision variable for whether to construct the line of branch b. E ess,n represents the capacity of energy storage construction at node n. N line is the set of lines to be constructed, and N is the total number of network nodes; C total represents the total cost of construction and operation of the new industrial park power grid. C total is: C total = C inv + C loss + C grid Where: C inv is the equivalent annual construction cost of energy storage and lines, C loss represents the operating loss cost of the power grid, including the operating losses of energy storage and line losses, C grid represents the power purchase cost from the superior power grid; the calculation formulas for various costs are as follows: Where: r is the discount rate, y m is the investment period, c ess is the investment cost per unit capacity of energy storage, c line is the investment cost of the line, c loss is the cost of per unit network loss electricity, c eloss is the charge and discharge loss cost per unit power of energy storage conversion. S represents the set of typical scenarios, σ s represents the probability of the s-th typical scenario, I ij,s,t is the current of the line between nodes ij at time t in the s-th typical scenario, r ij is the resistance of the line between nodes ij, P ess,n,s,t represents the charge and discharge power of the n-node energy storage at time t in the s-th typical scenario; N line is the set of lines to be built; T represents the calculation period, B represents the set of all lines, N ess represents the set of energy storage nodes; c grid,s,t represents the electricity purchase price from the superior power grid at time t in the s-th typical scenario, P grid,s,t represents the power transmitted by the superior power grid at time t in the s-th typical scenario; The objective function of the lower-layer planning model is: Specifically, the constraints of the lower-layer planning model include energy storage operation constraints, power flow constraints, operating voltage constraints, branch current constraints, active power balance of electricity, and adjustable load constraints and operation constraints of the data center. The typical scenario conventional load curve serves as the output of the typical scenario conventional load in the power flow constraints and the active power balance constraints of electricity, and is used for the adjustable load constraints of the data center in the lower-layer planning model.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the power system optimization method according to any one of claims 1-5 are implemented.

Images