A power distribution network and data center collaborative planning method based on security domain analysis

Abstract

This application relates to the field of distribution network planning and discloses a collaborative planning method for distribution networks and data centers based on security domain analysis. The method includes the following steps: S1, constructing a security domain model for a distribution network-data center system including distributed generation sources; S2, based on the constructed security domain model, constructing a two-layer collaborative optimization model for the distribution network-data center system; S3, solving the two-layer collaborative optimization model using an algorithm solver to obtain the collaborative planning results for the distribution network and data center. In step S2, the security domain model of the distribution network-data center system represents the set of all operating points that satisfy the normal operation N-0 constraint and N-1 security constraint. By constructing the security domain model of the distribution network-data center system and proposing joint quantitative indicators, continuous characterization of the system's security margin is achieved, improving the system's anti-disturbance capability and operational stability, and avoiding the risk of local overload.

Description

Technical Field

[0001] This invention relates to the field of power distribution network planning technology, specifically to a collaborative planning method for power distribution networks and data centers based on security domain analysis. Background Technology

[0002] With the development of the digital economy, the scale and energy consumption of data centers continue to grow. Data center loads exhibit high power density and strong dynamic fluctuations. Connecting such loads to traditional power distribution networks on a large scale presents a series of technical challenges. On the one hand, power fluctuations in data center loads may cause distribution network node voltages to exceed specified limits or line current carrying capacity to exceed thermal stability limits, affecting power quality and equipment safety. On the other hand, power outages on the distribution network side directly affect the normal operation of data centers, potentially leading to data loss or service interruptions and causing economic losses. Therefore, improving the operational security of the interaction system between the distribution network and data centers through collaborative planning methods is an important technical issue in constructing new power distribution systems.

[0003] However, some insurmountable problems remain. First, current research generally relies on the traditional N-1 criterion for safety verification. However, in scenarios with large-scale uncertain distributed power sources and highly dynamic data center loads, this criterion is computationally inefficient in handling uncertainty and fails to quantify the system's safety margin, making it difficult to meet the planning flexibility requirements of new power distribution systems. Furthermore, existing research focuses primarily on optimizing the power supply side and energy storage systems; for the load side, especially for high-density, high-reliability loads like data centers, a comprehensive safety analysis framework is still lacking. In addition, while analysis methods such as safety domains have been applied to stability analysis during system operation, their application at the level of collaborative planning between power distribution networks and data centers is insufficient. They fail to effectively embed safety boundaries and safety margins into the planning model, thus limiting the overall performance of the planning scheme. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a collaborative planning method for power distribution networks and data centers based on security domain analysis. This method solves the problem that existing technologies, when conducting collaborative planning for power distribution networks and data centers containing distributed power sources, lack effective quantitative assessment methods for system security margins, making it difficult to formulate planning schemes that balance economy and operational safety.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a collaborative planning method for power distribution networks and data centers based on security domain analysis, comprising the following steps:

[0006] S1. Construct a security domain model for a power distribution network-data center system containing distributed power sources;

[0007] S2. Based on the constructed security domain model of the power distribution network-data center system, construct a two-layer collaborative optimization model for the power distribution network-data center system;

[0008] S3. Solve the two-layer collaborative optimization model using an algorithm solver to obtain the collaborative planning results of the power distribution network and the data center;

[0009] In step S2, the security domain model of the power distribution network-data center system represents the set of all operating points that satisfy the normal operation N-0 constraint and N-1 security constraint, and quantifies the practical security boundary of the system.

[0010] In step S3, the two-layer collaborative optimization model of the power distribution network-data center system includes an upper planning layer and a lower operation layer;

[0011] The upper planning layer optimizes the planning scheme with the objectives of minimizing the total life cycle cost, maximizing the system's maximum energy supply capacity, and minimizing the standard deviation of the all-dimensional safety margin.

[0012] The lower-level operating layer solves for the dynamic adjustment strategy with the goal of minimizing operating costs, and solves for the maximum energy supply capacity index of the system.

[0013] Preferably, in step S1, constructing the security domain model of the power distribution network-data center system including distributed power sources includes:

[0014] The operating point is defined as a vector consisting of the load power of all unbalanced nodes when the distribution network is operating normally, and the load power is limited to a specific range.

[0015] Establish normal operation N-0 constraints, including but not limited to power flow constraints and data center constraints;

[0016] Establish N-1 safety constraints, which take into account that the output of distributed power sources remains unchanged after N-1 failures and that the data center load can be reduced and transferred, and also consider network bandwidth and task transfer time constraints.

[0017] A security domain model for the power distribution network-data center system is constructed based on the operating point, the normal operation N-0 constraint, and the N-1 security constraint.

[0018] Preferably, the power flow constraints in the normal operation N-0 constraint include balance equations for calculating the power of the line and main transformer based on the load power of downstream nodes and the output of distributed generation. The specific formulas are as follows:

[0019]

[0020] In the formula, P i B / P iTF Let i represent the power of the line / main transformer i, where i∈N; P represents the set of downstream nodes of line / main transformer i; j DG The output of all DGs connected to node j, in MW.

[0021] Preferably, the data center constraint in the normal operation N-0 constraint includes:

[0022] The total power consumption of the data center is constrained, which is determined by the power consumption of the servers and the power consumption of the cooling equipment, and the maximum value of the total power consumption is limited.

[0023] Server energy consumption constraints, wherein the server energy consumption is related to the number of servers in the power-on state and the amount of data tasks, and limits the number of servers in the power-on state and the CPU utilization of a single server.

[0024] The power consumption of the cooling system is constrained, wherein the power consumption of the cooling system is related to the cooling power of the data center, and the maximum value of the power consumption of the cooling system is limited.

[0025] Preferably, other constraints in the normal operation N-0 constraint include:

[0026] Line capacity constraints;

[0027] Main transformer capacity constraints;

[0028] Output constraints of distributed power sources.

[0029] Preferably, the N-1 security constraints include:

[0030] The distributed power source maintains its instantaneous output unchanged after an N-1 fault.

[0031] After an N-1 failure, a data center can reduce some of its load and partially transfer it to other data centers, but this is limited by the network bandwidth capacity between data centers and the task transfer latency.

[0032] The N-1 line capacity constraint and the N-1 main transformer capacity constraint are used to convert the operating point at the time of the N-1 fault into the operating point at the time of normal operation through a mapping relationship.

[0033] Preferably, the practical security boundary of the quantified system includes:

[0034] Define a safety upper boundary, which reflects the combined energy supply capacity of energy supply equipment, energy storage equipment and distributed power sources, as well as the dynamic adjustment of data center load reduction and migration;

[0035] Define a safety lower boundary that reflects the minimum load requirements of the power supply equipment and takes into account the data center task migration characteristics;

[0036] Define the safety distance as the shortest distance from the current operating point of the system to each safety boundary;

[0037] The system's safety performance is described using a full-dimensional safety margin vector, which is composed of the minimum distances from the operating point to each safety upper boundary at each time period.

[0038] The standard deviation of the full-dimensional safety margin is used to describe the balance of the system's safety margin in the spatial dimension.

[0039] Preferably, the upper planning layer aims to minimize the total life cycle cost;

[0040] The total life cycle cost includes:

[0041] Investment cost calculated using annualized factors;

[0042] Operating costs are calculated using annualized factors.

[0043] Preferably, the operating costs include:

[0044] Equipment operation and maintenance costs;

[0045] Demand response costs;

[0046] Cost of purchasing electricity from the upstream power grid;

[0047] Carbon emission costs;

[0048] Network loss costs;

[0049] Costs of curtailing wind and solar power.

[0050] Preferably, the lower-level operating layer aims to minimize operating costs, and its constraints include:

[0051] Safety domain constraint, which ensures that all elements of the full-dimensional safety margin vector are greater than zero;

[0052] Distributed power generation output constraints;

[0053] Distribution network node voltage constraints and line current constraints;

[0054] Source node injects grid power constraints;

[0055] Power flow constraints;

[0056] Power purchase capacity constraints;

[0057] Demand response constraints include the relationship between data center load reduction and reduction factor, as well as the upper and lower limits of the reduction factor;

[0058] Energy storage constraints include constraints such as the state of charge / discharge, charge / discharge power, energy range, and sustainability.

[0059] This invention provides a collaborative planning method for power distribution networks and data centers based on security domain analysis.

[0060] It has the following beneficial effects:

[0061] 1. This invention constructs a security domain model for a distribution network-data center system containing distributed power sources and proposes a joint quantitative index system based on the standard deviation of the full-dimensional security margin vector and the maximum power supply capacity of the system. This achieves the effect of continuous spatial quantitative characterization of the system's security margin, solving the technical problems of discretization of security constraints and difficulty in evaluating the global security boundary in traditional planning methods. This significantly improves the system's anti-disturbance capability and operational stability, and avoids the risk of local overload.

[0062] 2. This invention constructs a two-layer collaborative optimization model for power distribution networks and data center systems, and simultaneously optimizes three objectives in the upper-level planning: total life-cycle cost, maximum system power supply capacity, and standard deviation of all-dimensional safety margin. This achieves the effect of balancing economy and safety in planning decisions, and solves the technical problem that planning schemes that only focus on economy cannot guarantee the long-term operational safety of the system. Thus, it achieves a balance between long-term economic benefits and safety benefits by significantly reducing operating costs by precisely improving safety, with only a slight increase in total investment costs. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the DN-DC system of the present invention;

[0064] Figure 2 This is a schematic diagram of the two-layer optimization model relationship of the present invention;

[0065] Figure 3 This is a schematic diagram of the solution process for the planning model of the present invention. Detailed Implementation

[0066] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Please see the appendix Figure 1 -Appendix Figure 3 This invention provides a collaborative planning method for power distribution networks and data centers based on security domain analysis, comprising the following steps:

[0068] S1. Construct a security domain model for a power distribution network-data center system containing distributed power sources;

[0069] S2. Based on the constructed security domain model of the power distribution network-data center system, a two-layer collaborative optimization model for the power distribution network-data center system is constructed.

[0070] S3. Use the algorithm solver to solve the two-layer collaborative optimization model to obtain the collaborative planning results of the power distribution network and the data center;

[0071] In step S2, the security domain model of the power distribution network-data center system represents the set of all operating points that satisfy the normal operation N-0 constraint and N-1 security constraint, and quantifies the practical security boundary of the system.

[0072] In step S3, the two-layer collaborative optimization model of the power distribution network-data center system includes an upper planning layer and a lower operation layer;

[0073] The upper-level planning layer aims to minimize the total life-cycle cost, maximize the system's maximum energy supply capacity, and minimize the standard deviation of the all-dimensional safety margin, thereby optimizing the planning scheme.

[0074] The lower-level operating layer seeks to optimize the dynamic adjustment strategy with the goal of minimizing operating costs, and also seeks to solve for the system's maximum energy supply capacity index.

[0075] The specific technical solution for step S1 above is as follows:

[0076] In this embodiment, constructing a security domain model for a power distribution network-data center system containing distributed power sources is one of the core steps, aiming to quantify the system's operational boundaries and security margins. This model is based on a comprehensive consideration of the system's operating state, normal operation constraints (N-0 constraints), and single-point-of-failure security constraints (N-1 security constraints).

[0077] First, the operating point is defined as a vector consisting of the load power of all unbalanced nodes during normal operation of the distribution network. When there are n unbalanced nodes in the system, the operating point W can be expressed as:

[0078]

[0079] Where i represents the node number, its value range is 1≤i≤n, and i∈N; n is the dimension of the working point; P i Load Let node i be the node number connected to, where i represents the node number, with a value range of 1 ≤ i ≤ n, and i ∈ N; n is the dimension of the working point; P i Load Let P be the load power connected to node i. Additionally, the load power P of each node... iLoad It is confined to a specific range, namely:

[0080]

[0081] In the formula: They are respectively The lower limit / upper limit.

[0082] Secondly, normal operation N-0 constraints were established to ensure the safe operation of the system under normal load conditions. N-0 constraints include, but are not limited to, power flow constraints and data center constraints. Regarding power flow constraints, considering the short power supply radius of the 10kV urban distribution network studied, which mainly relies on load power supply and distributed generation absorption, the distribution of active power flow becomes crucial. Therefore, the power flow calculation is simplified to a power balance equation. The power P of line i... i B The power P of the main transformer i i TF The calculation can be performed using the load power of its downstream nodes and the output of distributed power sources. The specific formula is as follows:

[0083]

[0084] Among them, P i B and P i TF Let i represent the power of line i and main transformer i, respectively, where i∈N; and Let P represent the sets of downstream nodes of line i and main transformer i, respectively; j DG This represents the output power of all distributed power sources connected to node j, in MW.

[0085] Regarding data center constraints, since servers and cooling equipment account for a significant portion of the total power consumption of a data center, the total power consumption of a data center can be expressed as:

[0086]

[0087] Furthermore, the maximum total power consumption of the data center is limited:

[0088]

[0089] in, This represents the total active power consumption; and λ represents the active power of servers and cooling equipment in a data center, respectively; DC P represents the data center's energy efficiency coefficient. DC，maxThis represents the maximum data processing load capacity of the data center, determined by the server parameters. The subscript i represents a node, s represents a scenario, and t represents time. Furthermore, server energy consumption is calculated as follows:

[0090]

[0091] The number of servers in operation and the CPU utilization of a single server are limited: In the formula, and These represent the silent power and peak power of a single server when no tasks are being processed, respectively; ψ i This refers to the server's processing speed. W represents the number of servers in operation during time period t. i,s,t Data volume in the data center; The number of servers is configured; φ is the server redundancy factor; S max This represents the upper limit of CPU utilization for a single server. The operating characteristics of the cooling system's power consumption can be expressed as:

[0092]

[0093] The maximum power consumption of the cooling system is limited:

[0094] in, This refers to the power consumption of the air conditioner. Cooling power for data centers; λ Air The energy efficiency coefficient of the refrigeration equipment; The cooling equipment capacity configured for the data center. In addition to the power flow constraints and data center constraints mentioned above, normal N-0 operation also includes other constraints, such as line capacity constraints, main transformer capacity constraints, and distributed power output constraints.

[0095] Line capacity constraints are expressed as follows:

[0096] The main transformer capacity constraint is expressed as:

[0097] The output constraints of distributed power sources are expressed as follows:

[0098] In the formula, and These represent the capacities of the transmission lines and main transformers, respectively; B and T are the sets of all transmission lines and main transformers. Provide power for distributed power sources; For the predicted value of distributed power sources; This represents the prediction error.

[0099] Secondly, an N-1 safety constraint is established, requiring the system to maintain power supply even in the event of a single point of failure. The N-1 safety constraint considers the following: distributed power sources maintain their instantaneous output after an N-1 failure to ensure the network's available capacity. Its expression is:

[0100]

[0101] In the formula, P is the output of DG after system N-1 failure; t DG To ensure the normal operation of DG. In the event of an N-1 failure, the data center can reduce its load and partially transfer power to other data centers to ensure the continuity of power supply. This load adjustment is represented as follows:

[0102]

[0103] Where, α i Let α be the percentage of data center load that can be reduced at node i, 0 ≤ α i ≤1; β j,i Let be the percentage of task load transferred from data center j to i, and ∑ j≠i β j,i ≤α i , Load transfer is limited by the percentage of load reduction. To ensure the data center operates smoothly during load transfer, α... i Let α be the percentage of data center load that can be reduced at node i, 0 ≤ α i ≤1; β j,i Let be the percentage of task load transferred from data center j to i, and ∑ j≠i β j,i ≤α i , Load transfer is limited by the load reduction ratio. To ensure that network bandwidth between data centers does not exceed their respective capacity and latency limits during load transfer, network bandwidth and task transfer time are limited as follows:

[0104]

[0105]

[0106] In the formula, B j,i and T j,i These represent the network bandwidth capacity and task transfer delay from data center j to i, respectively. In addition, the N-1 security constraints also include N-1 line capacity constraints and N-1 main transformer capacity constraints.

[0107] Since the variable studied in this method is the operating point W at the N-0 fault time, while the model needs to take into account the W at the N-1 fault time. t+1Due to security constraints, W needs to be... t+1 The transformation is to W. This transformation is described by a mapping h, where h: W→W t+1 In traditional N-1 security analysis, W = W t+1 The situation is different. When considering data center demand response, it is not an identity mapping. The specific relationship is as follows:

[0108]

[0109] Where E is the identity matrix; α is a vector composed of the load reduction ratio coefficients of each interruptible load. When component d experiences an N-1 fault, the distribution network will undergo network topology reconfiguration to restore power supply to the non-faulty areas, and its power balance equation will change accordingly.

[0110] The capacity constraint for line N-1 is:

[0111]

[0112] In the formula, ψ d For component d, a failure scenario is presented. This represents the power of loads other than the data center after component d experiences an N-1 fault; The power of line Bi after component d experiences an N-1 fault; Energy storage capacity configured for data centers; This represents the set of downstream nodes of Bi after the fault. The N-1 main transformer capacity constraint is:

[0113]

[0114] In the formula, The power of the main transformer Ti after component d experiences an N-1 fault; Let Ti be the set of downstream nodes. Let the fault set be... If at a certain working point W, for If both the N-1 line capacity constraint and the N-1 main transformer capacity constraint are met, then W satisfies the N-1 safety criterion.

[0115] Ultimately, the security domain model Ω of the power distribution network-data center system DN-DC It is constructed as the set of all operating points in the state space that satisfy the N-0 and N-1 safety criteria. This model can be represented as:

[0116] Ultimately, the security domain model Ω of the power distribution network-data center system DN-DC It is constructed as the set of all operating points in the state space that satisfy the N-0 and N-1 safety criteria. This model can be represented as:

[0117]

[0118] The following conditions must be met simultaneously:

[0119]

[0120] for At the same time, the following conditions must also be met:

[0121]

[0122] In the formula, Θ is the bounded set of all reasonable operating points, i.e., the state space. This safety domain model effectively characterizes the safe operating range of the system under normal and fault conditions, providing a foundation for subsequent collaborative optimization.

[0123] Based on this, this implementation further quantifies the practical safety boundary of the system and defines the maximum power supply capacity index of the system.

[0124] In this context, the practical safety boundary of the power distribution network-data center system represents all critical operating points of the system under the premise of satisfying N-1 safety checks. This boundary is jointly determined by the system structure and key equipment parameters, and is composed of a set of hyperplanes. When a fault occurs, the data center can respond to demand, and energy storage devices can serve as temporary power resources to ensure that the system is not overloaded as much as possible. The general formula for its practical safety boundary is as follows:

[0125]

[0126] In the formula, B i Indicates the i-th power supply line; This is the upper safety boundary of the feeder, reflecting the maximum power supply capacity of the power supply line; The lower safety boundary of the feeder represents the minimum load the feeder can maintain after load reduction, reflecting the minimum load requirement of the line. It is primarily influenced by the minimum load requirements of the equipment and the load transfer characteristics of the data center; C U C is the upper limit of the feeder's capacity. L Minimum load of the feeder; μ i μ j P is a fixed coefficient used to adjust the weight of load transfer or reduction; i The load size of each outlet pipeline.

[0127] For a power distribution network-data center system, the safety upper boundary reflects the comprehensive power supply capacity of power supply equipment, energy storage equipment, and distributed power sources, as well as the dynamic adjustment of data center load reduction and migration, and can be expressed as:

[0128]

[0129] In the formula, C UThis represents the upper limit of the power supply equipment's capacity; the ellipsis (...) indicates other possible constraints. The safety lower boundary reflects the minimum load requirement of the power supply equipment, ensuring its normal operation, and, considering the characteristics of data center task migration, is expressed as:

[0130]

[0131] In the formula, C L This is the minimum load limit.

[0132] The full-dimensional observation of the power distribution network-data center system adopts an indirect observation method based on safe distance. The safe distance is defined as the shortest distance from the current operating point of the system to each safety boundary. This distance is positive when the operating point is within the safety domain and negative when it is outside the safety domain, and can be expressed as:

[0133]

[0134] For the operating points of the system during 24 typical time periods on a typical day, the full-dimensional safety margin (FDSM) vector is used to describe the system's safety performance, expressed as follows:

[0135]

[0136] In the formula, D FDSM The vector formed by the minimum distances from each working point to each safety upper boundary in each time period is defined as FDSM; For each work point at different times, the upper safety boundary is determined. The minimum distance, in MW; Indicates the working point of time period t. The distance is expressed in MW.

[0137] The random fluctuations in load can cause the system operating point to approach or cross any safe upper boundary. To overcome this limitation, it is necessary to ensure that the numerical distribution of each element in the FDSM vector is balanced. This implementation utilizes the standard deviation index σ of the FDSM. FDSM The FDSM standard deviation describes the spatial balance of the system's safety margin; a value closer to 0 indicates a more balanced safety margin. The standard deviation of the FDSM, constructed based on the above formula, is shown below:

[0138]

[0139] In the formula, d SDPLU,av This represents the average value of all elements in the system FDSM. The unit is MW.

[0140] Furthermore, the total energy supply capacity index of the system, denoted as TSC, characterizes the maximum energy supply of the system under the N-1 criterion. Its calculation can be expressed as the following optimization problem:

[0141]

[0142] In the formula, ∑P i The total load supplying energy to this outlet, in MW; P min ≤P≤P max P represents the state space of the system's operating point; max / P min represents the upper and lower limits of the system's operating point constraints, respectively, with each element in the vector in units of MNW; h(P) = 0 indicates a normal system operation constraint; w(P) ≤ 0 indicates an N-1 safety constraint. Solving this optimization problem determines the maximum load the system can provide while ensuring safety.

[0143] The specific technical solution for step S2 above is as follows:

[0144] In this embodiment, the two-layer collaborative optimization model for the power distribution network-data center system aims to collaboratively optimize the planning and operation of the power distribution network and data center to achieve a balance between economy and security. The model includes an upper planning layer and a lower operation layer.

[0145] In this embodiment, a two-layer collaborative optimization model for the power distribution network-data center system is established based on the constructed security domain model of the power distribution network-data center system. This model includes an upper planning layer and a lower operation layer. Configuration parameters are passed from the upper planning layer to the lower operation layer, and the optimization results of the operation layer are fed back to the upper planning layer to calculate the objective function.

[0146] The upper planning layer is responsible for optimizing the planning from a macro perspective, making decisions on the configuration capacity and installation location of data centers and distributed power supplies with multiple objectives, including minimizing the total life cycle cost, maximizing the maximum system power supply capacity, and minimizing the standard deviation of the full-dimensional safety margin (FDSM).

[0147] The primary objective of this model is to minimize the system’s total lifecycle cost, which consists of annualized investment cost and annual operating cost.

[0148] F1=min(C I +C O );

[0149] In the formula, F1 is the economic objective function; C I C represents the annualized investment cost. O This refers to the annual operating cost.

[0150] Annualized investment cost C IIt covers investments in all new or renovated equipment, including power lines, main transformers, distributed power sources, and data centers.

[0151] C I =C I,Line +C I,TF +C I,DG +C I,DC ;

[0152] In the formula, C I,Line C represents the annualized investment cost of the line. I,TF The annualized investment cost for main transformer expansion; C I,DG C represents the annualized investment cost of distributed power sources. I,DC This represents the annualized investment cost for a data center.

[0153] All investment costs are converted to an annualized cost using an annualized equipment factor, as detailed below:

[0154]

[0155] In the formula, b is the annualized factor for the equipment; Ω Line For the set of candidate installation nodes for the system lines; Φ Line This is a set of candidate investment models for the railway line. L represents the investment cost per unit length of a line of type m; ij Let be the length of line ij; Let m be the 0-1 decision variable for whether to invest in line ij.

[0156] C I,TF =bc TF P TF-add ;

[0157] In the formula, b is the annualized factor for the equipment; c TF The configuration cost per unit capacity transformer; P TF-add To expand the capacity of the transformer.

[0158]

[0159] In the formula, b is the annualized factor for the equipment; Ω PV c is the set of candidate photovoltaic installation nodes; PV The configuration cost of a single photovoltaic unit; The number of photovoltaic installations; Ω WT For the set of candidate installation nodes for wind turbines; c WT The configuration cost for a single wind turbine; This refers to the number of wind turbines installed.

[0160]

[0161] In the formula, b is the annualized factor for the equipment; Ω DC For the set of candidate installation nodes in the data center; c Ser The configuration cost of a single server; c represents the number of servers installed. Air The configuration cost of a single refrigeration unit; c is the number of refrigeration equipment installed. ESS The configuration cost of a single energy storage unit; This refers to the number of energy storage installations.

[0162] The calculation method for the annualized equipment factor b is as follows:

[0163]

[0164] In the formula, y is the discount rate; y is the equipment's useful life. Annual operating cost C O This comprehensively considers costs related to equipment operation and maintenance, demand response, purchasing electricity from the upper-level grid, carbon emissions, network losses, and wind and solar power curtailment.

[0165] C O =C O,Eq +C O,DR +C O,Pur +C O,C +C O,Loss +C O,Waste ;

[0166] In the formula, C O,Eq For equipment operation and maintenance costs; C O,DR For demand response costs; C O,Pur For electricity purchase cost; C O,C For carbon emission costs; C O,Loss For network loss costs; C O,Waste Costs associated with curtailing wind and solar power.

[0167] The operating costs are calculated as follows:

[0168] Equipment maintenance cost C O,Eq It consists of transformer operation and maintenance costs, DG operation and maintenance costs, and DC operation and maintenance costs.

[0169]

[0170]

[0171] In the formula, Λ TF / Λ PV / Λ WT / Λ DC These are the sets of nodes where transformers / PV / WT / DC are installed, respectively; f O,PV / fO,WT These are the maintenance costs per unit of electricity generated by PV / WT, respectively. f represents the active power injected into node i by PV / WT during time period t; O,TF / f O,Ser / f O,Air / f O,ESS These represent the operation and maintenance costs of transformers, data center servers, data center cooling equipment, and data center energy storage, respectively. Δt represents the duration of time period t, which is 1 hour.

[0172] DR cost C O,DR DR costs are reflected by reducing load nodes and their corresponding compensation costs.

[0173]

[0174] In the formula, Λ DC,cut For the set of nodes whose load can be reduced; f O,DC,cut Compensation for unit power reduction in load; ΔP i ,s,t represents the magnitude of the active power of the load reduced after DR.

[0175] Electricity purchase cost C O,Pur The cost of electricity purchased this time takes into account time-of-use pricing, as shown in the following calculation formula:

[0176]

[0177] In the formula, D L (1,:) is the set of end nodes of the branch with source node 1 as the first end node; P represents the time-of-use electricity price for period t; 1j,s,t The active power flowing into the local power grid via the line connecting source node 1 during time period t.

[0178] Carbon emission cost C O,C The calculation is performed using the following formula:

[0179]

[0180] In the formula, e represents the system carbon emission cost corresponding to time period t. TPG The carbon emission coefficient of the external power grid. Power purchased for the power grid.

[0181] Regarding network loss cost C O,Loss Although it has been included in the power purchase cost of the upper-level power grid, in order to better optimize network losses, it is expressed as the penalty cost of network losses.

[0182]

[0183] In the formula, Ps ,t Loss,Net R represents the network loss during time period t; i j represents the resistance of line ij; I i j,s,t 2 Let be the magnitude of the current in line ij during time period t, in Ω. DL This is a collection of DC branches.

[0184] Cost of curtailing wind and solar power C O,Waste The cost of wind and solar curtailment is due to the inability to fully utilize wind and solar power. Therefore, the cost of wind and solar curtailment is calculated based on the difference between the predicted power output and the actual demand, as shown below.

[0185]

[0186] In the formula, f O,WT,cut / f O,PV,aut These represent the cost per unit of electricity curtailed (wind / solar power); These represent the sum of the predicted active power output of PV / WT installed at node i during time period t.

[0187] Beyond economic efficiency, the model's second objective is to maximize the system's energy supply capacity index T. TSC This is to improve the system's reliability under extreme conditions.

[0188] F2 = max(T) TSC );

[0189] In the formula, F2 is the reliability objective function; T TSC This represents the maximum energy supply capacity of the system under the N-1 criterion.

[0190] Finally, to ensure the balance and security of system operation, the third objective is to minimize the standard deviation of the total safety margin (FDSM), thereby balancing the safety margins of each line.

[0191]

[0192] In the formula, F3 is the equilibrium objective function; d i Let be the number of days in the i-th typical day.

[0193] To ensure the feasibility of the planning scheme, the upper-level model must also meet a series of constraints. First, the investment and installation capacity of various types of equipment cannot exceed their upper limits, including the expansion capacity of transformers, the capacity of data centers and their internal equipment, and the number of distributed power supplies installed.

[0194] 0≤P TF-add ≤P TF,max ;

[0195] In the formula, P TF-add To expand the capacity of the transformer; P TF,max This refers to the upper limit of transformer installation capacity.

[0196]

[0197] In the formula, The data center capacity installed on node i; This represents the upper limit of the data center capacity for node i.

[0198]

[0199] In the formula, The number of servers installed for node i; This represents the maximum number of servers.

[0200]

[0201] In the formula, The number of cooling devices installed for node i; This is the upper limit for the number of refrigeration equipment.

[0202]

[0203] In the formula, The amount of energy storage installed for node i; This represents the upper limit for the amount of energy stored.

[0204]

[0205] In the formula, The number of photovoltaic cells installed at node i; This represents the upper limit for the number of photovoltaic units; Ω PV,max This represents the upper limit for the number of photovoltaic units; Ω PV This is a set of candidate photovoltaic installation nodes; The number of fans installed for node i; Maximum number of wind turbines; Ω WT This is a set of candidate installation nodes for wind turbines.

[0206] For the investment and construction of the line, each candidate location can only select one type of line for construction:

[0207]

[0208] In the formula, Ω is a 0-1 decision variable for whether to invest in line ij based on model m; Line This is a set of candidate installation nodes for the system lines.

[0209] The lower-level operating layer receives the planning scheme from the upper layer and establishes a steady-state optimization model. Its core task is to solve for the optimal dynamic adjustment strategy by scheduling the active control devices within the system, with the goal of minimizing the total system operating cost, and then feeds back the calculated operating cost and the system's maximum power supply capacity, among other indicators, to the upper layer.

[0210] To achieve this goal and ensure the system operates safely and stably under various scenarios, the lower-level model must adhere to a series of strict constraints. The primary constraint is the safety domain constraint, which requires the system to maintain safe operation even under N-1 faults. This means that the distance from the operating point to the safety boundary at each time period must be greater than zero.

[0211]

[0212] In the formula, This represents the minimum distance from the operating point to the safety upper boundary at each time period. The actual output of a distributed power source cannot exceed its predicted maximum output at the current moment.

[0213]

[0214] In the formula, and These represent the actual active power outputs of photovoltaic and wind turbines during time period t, respectively; P i ,s,t PV,max and Λ represents the predicted maximum active power output of photovoltaic and wind turbines during time period t, respectively; PV and Λ WT These are the sets of nodes where photovoltaic and wind turbines are installed, respectively.

[0215] The operation of a power distribution network must meet a series of physical constraints. The voltage at each node must be maintained within a safe range, and the line current must not exceed its thermal stability limit.

[0216]

[0217] In the formula, U is the square of the voltage at node i during time period t; i ,min 2 and These are the lower and upper squares of the node voltage, respectively; Ω DN It is the set of all distribution network nodes; Let be the square of the current in line ij during time period t; The square of the safe current of line ij; Ω DL This is a collection of DC branches.

[0218] The power exchanged between the source node and the upstream power grid should also be within the specified limits.

[0219]

[0220] In the formula, P 1,j,s,t The active power flowing into the local power grid via the line connecting source node 1 during time period t; and These are the lower and upper limits of the active power injected into the grid by the source node, respectively; It is the set of end nodes of the branch with source node 1 as the first end node.

[0221] In addition, the power of the entire network must be kept in balance, that is, the injected power equals the consumed power.

[0222]

[0223] In the formula, Power purchased from the power grid; P i ,s,t PV Photovoltaic active power; P i ,s,t WT P represents the active power of the wind turbine. i ,s,t BSS,down P is the energy storage discharge power; i ,s,t ESS,up Power for energy storage charging; P i ,s,t Load The original load power; ΔP i ,s,t L0ad The load power reduced in response to demand. The amount of electricity purchased from the upstream grid is limited by the total capacity of the transformers.

[0224] 0≤P t Trans ≤(P TF-add +P TF-0 );

[0225] In the formula, P t Trans P represents the power purchased at time t. TF-add To expand the capacity of the transformer; P TF-0 The original capacity of the transformer.

[0226] Data centers, as a flexible load, can participate in demand response. The amount of load reduction is determined by a reduction factor, which is adjustable within a certain range.

[0227]

[0228] In the formula, The active power of the reduced load; v i,s,t This is the data center load reduction factor; Active power consumption of the data center; Λ DC,cutA set of data center nodes that can participate in demand response; ν min and ν max These are the lower and upper limits of the load reduction factor, respectively.

[0229] The operation of energy storage systems involves several constraints, including the mutual exclusivity of charge and discharge states, the upper limit of charge and discharge power, the dynamic changes in the stored energy, the energy conservation during the operating cycle, and the upper and lower limits of the state of charge.

[0230]

[0231] In the formula, and These are the 0-1 state variables for energy storage charging and discharging, respectively; and These are the energy storage charging and discharging power, respectively; This refers to the upper limit of the energy storage charging and discharging power. Let η be the stored energy quantity at time t; up and η down These represent the charging and discharging efficiencies, respectively; Δt is the duration of the time period. and These represent the lower and upper limits of the energy storage capacity, respectively.

[0232] The specific technical solution for step S3 above is as follows:

[0233] This step aims to solve the two-layer collaborative optimization model of the power distribution network-data center system to obtain the collaborative planning results of the power distribution network and the data center. Specifically, the solution process involves an upper planning layer and a lower operation layer, and the solution process is an iterative optimization process. The upper planning layer uses a modified normalization method and a plane constraint method for multi-objective optimization, while the lower operation optimization and the solution for the system's maximum power supply capacity are completed using the commercial solver Gurobi.

[0234] When the specific solution process starts, the upper planning layer first initializes a population, which consists of multiple individuals, each representing a candidate planning scheme. Each planning scheme includes a set of definite decision variables, specifically the configuration capacity and installation location of data centers and distributed power sources, covering line investment decision variables, main transformer expansion capacity, number of photovoltaic installations, number of wind turbine installations, number of server installations, number of cooling equipment installations, and number of energy storage installations.

[0235] Subsequently, based on the constructed utopian hyperplane, the upper-level planning layer transforms the original multi-objective problem into a set of single-objective optimization subproblems with additional constraints. Each subproblem aims to minimize the lifetime cost F1, but with added constraints that confine the solution space to specific normals originating from the utopian hyperplane. By adjusting the position of these normals, the entire Pareto front can be systematically explored.

[0236] The solution process for each decomposed single-objective subproblem is as follows:

[0237] The upper-level optimizer adjusts a set of planning decision variables, including line investment decisions, main transformer expansion capacity, and the number of data centers and distributed power sources. For each given set of planning decision variables, it is passed as fixed parameters to the lower-level operation layer. Upon receiving the parameters, the lower-level operation layer uses the Gurobi solver to perform two calculations: first, to minimize operating costs while satisfying all operating constraints, it calculates the optimal annual operating cost; second, based on the system's maximum power supply capacity, T... TSC With the goal of maximizing, the specific value of this indicator is calculated.

[0238] After the lower-level calculations are completed, the annual operating cost C will be obtained. O And the system's maximum energy supply capacity index T TSC Returning to the upper planning layer, and combining the known annualized investment cost, the three objective function values ​​corresponding to the planning scheme are calculated: life-cycle cost F1, maximum system energy supply capacity F2, and full-dimensional safety margin (FDSM) standard deviation F3.

[0239] F1=min(C I +C O );

[0240] In the formula, C I C represents the annualized investment cost. O This refers to the annual operating cost.

[0241] F2 = max(T) TSC );

[0242] In the formula, T TSC This represents the maximum energy supply capacity of the system under the N-1 criterion.

[0243]

[0244] In the formula, d represents the variance of the system's total safety margin (FDSM); i Let be the number of days in the i-th typical day.

[0245] The upper-level planning layer systematically solves these decomposed single-objective subproblems with different additional constraints, obtaining a series of efficient solutions evenly distributed on the Pareto front. The set of these solutions directly constitutes the final Pareto front solution set, where each solution is a Pareto optimal planning scheme, representing a specific trade-off between the three objectives of economy, energy supply capacity, and safety.

[0246] Finally, the final collaborative programming result is selected from the output Pareto front solution set. A balanced decision-making method is used, specifically: first, the three objective function values ​​of all schemes in the solution set are normalized; then, the Euclidean distance from each scheme to the ideal point is calculated; and the scheme with the smallest distance is selected as the final collaborative programming result with balanced performance across all objectives.

[0247] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A collaborative planning method for power distribution networks and data centers based on security domain analysis, characterized in that, Includes the following steps: S1. Construct a security domain model for a power distribution network-data center system containing distributed power sources; S2. Based on the constructed security domain model of the power distribution network-data center system, construct a two-layer collaborative optimization model for the power distribution network-data center system; S3. Solve the two-layer collaborative optimization model using an algorithm solver to obtain the collaborative planning results of the power distribution network and the data center; In step S2, the security domain model of the power distribution network-data center system represents the set of all operating points that satisfy the normal operation N-0 constraint and N-1 security constraint, and quantifies the practical security boundary of the system. In step S3, the two-layer collaborative optimization model of the power distribution network-data center system includes an upper planning layer and a lower operation layer; The upper planning layer optimizes the planning scheme with the objectives of minimizing the total life cycle cost, maximizing the system's maximum energy supply capacity, and minimizing the standard deviation of the all-dimensional safety margin. The lower operating layer solves for the dynamic adjustment strategy with the goal of minimizing operating costs, and solves for the maximum energy supply capacity index of the system. In step S1, constructing the security domain model of the power distribution network-data center system with distributed power sources includes: The operating point is defined as a vector consisting of the load power of all unbalanced nodes when the distribution network is operating normally, and the load power is limited to a specific range. Establish normal operation N-0 constraints, including power flow constraints and data center constraints; Establish N-1 safety constraints, which take into account that the output of distributed power sources remains unchanged after N-1 failures and that the data center load can be reduced and transferred, and also consider network bandwidth and task transfer time constraints. Construct a security domain model for the power distribution network-data center system based on the operating point, the normal operation N-0 constraint, and the N-1 security constraint. The practical security boundary of the quantification system includes: Define a safety upper boundary, which reflects the combined energy supply capacity of energy supply equipment, energy storage equipment and distributed power sources, as well as the dynamic adjustment of data center load reduction and migration; Define a safety lower boundary that reflects the minimum load requirements of the power supply equipment and takes into account the data center task migration characteristics; Define the safety distance as the shortest distance from the current operating point of the system to each safety boundary; The system's safety performance is described using a full-dimensional safety margin vector, which is composed of the minimum distances from the operating point to each safety upper boundary at each time period. The standard deviation of the full-dimensional safety margin is used to describe the balance of the system's safety margin in the spatial dimension.

2. The method for collaborative planning of power distribution networks and data centers based on security domain analysis according to claim 1, characterized in that, The power flow constraints in the normal operation N-0 constraint include the balance equations for calculating the power of the line and main transformer based on the load power of downstream nodes and the output of distributed generation. The specific formulas are as follows: ； In the formula, These are respectively the line / main transformer power, ; These are respectively the line / main transformer The set of downstream nodes; For nodes The output of all connected DGs, in MW; For nodes The power of the connected load.

3. The method for collaborative planning of power distribution networks and data centers based on security domain analysis according to claim 1, characterized in that, The data center constraints in the normally operating N-0 constraints include: The total power consumption of the data center is constrained, which is determined by the power consumption of the servers and the power consumption of the cooling equipment, and the maximum value of the total power consumption is limited. Server energy consumption constraints, wherein the server energy consumption is related to the number of servers in the power-on state and the amount of data tasks, and limits the number of servers in the power-on state and the CPU utilization of a single server. The power consumption of the cooling system is constrained, wherein the power consumption of the cooling system is related to the cooling power of the data center, and the maximum value of the power consumption of the cooling system is limited.

4. The method for collaborative planning of power distribution networks and data centers based on security domain analysis according to claim 1, characterized in that, Other constraints in the normal operation N-0 constraint include: Line capacity constraints; Main transformer capacity constraints; Output constraints of distributed power sources.

5. The method for collaborative planning of power distribution networks and data centers based on security domain analysis according to claim 1, characterized in that, The N-1 security constraints include: The distributed power source maintains its instantaneous output unchanged after an N-1 fault. After an N-1 failure, a data center can reduce some of its load and partially transfer it to other data centers, but this is limited by the network bandwidth capacity between data centers and the task transfer latency. The N-1 line capacity constraint and the N-1 main transformer capacity constraint are used to convert the operating point at the time of the N-1 fault into the operating point at the time of normal operation through a mapping relationship.

6. The method for collaborative planning of power distribution networks and data centers based on security domain analysis according to claim 1, characterized in that, The upper planning layer aims to minimize the total life cycle cost. The total life cycle cost includes: Investment cost calculated using annualized factors; Operating costs are calculated using annualized factors.

7. The method for collaborative planning of power distribution networks and data centers based on security domain analysis according to claim 6, characterized in that, The operating costs include: Equipment operation and maintenance costs; Demand response costs; Cost of purchasing electricity from the upstream power grid; Carbon emission costs; Network loss costs; Costs of curtailing wind and solar power.

8. The method for collaborative planning of power distribution networks and data centers based on security domain analysis according to claim 1, characterized in that, The lower-level operating layer aims to minimize operating costs, and its constraints include: Safety domain constraint, which ensures that all elements of the full-dimensional safety margin vector are greater than zero; Distributed power generation output constraints; Distribution network node voltage constraints and line current constraints; Source node injects grid power constraints; Power flow constraints; Power purchase capacity constraints; Demand response constraints include the relationship between data center load reduction and reduction factor, as well as the upper and lower limits of the reduction factor; Energy storage constraints include the state of charge / discharge, charge / discharge power, energy range, and sustainability constraints.

Images