A power transmission and distribution network collaborative scheduling method for heterogeneous resource partition clustering

Abstract

The application discloses a kind of power transmission and distribution network collaborative scheduling methods for heterogeneous resource partition clustering, comprising: step S1, constructing high load energy model and civil load model;Step S2, constructing flexible load two-stage partition clustering model;Step S3, constructing flexible load adjustable potential evaluation index based on partition clustering method;Step S4, constructing day-ahead-intra-day multi-time scale power transmission and distribution network collaborative scheduling model;Step S5, constructing day-ahead-intra-day multi-time scale power transmission and distribution network collaborative scheduling strategy and solution.The application effectively improves the efficiency of power grid coordinated scheduling of mass heterogeneous resources.Combining the internal mass heterogeneous resources of power transmission and distribution network, the adjustable potential of the flexible load partition clustering method based on digital model fusion is evaluated, and the day-ahead-intra-day multi-time scale demand response is considered to obtain the day-ahead-intra-day multi-time scale power transmission and distribution network collaborative scheduling strategy, which coordinates the internal mass heterogeneous resources of power transmission and distribution network to participate in system scheduling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power transmission and distribution network, in particular to a power transmission and distribution network collaborative scheduling method for heterogeneous resource partition clustering. BACKGROUND

[0002] The power industry is a major source of carbon emissions. In order to achieve the "double carbon" goal, a large number of new energy and controllable loads are connected to the power grid. With the intelligentization of the power grid, source-load-storage coordination has become an unstoppable trend. Demand response is an important means for load-side participation in power grid dispatching, and the main content is flexible load response, which has attracted widespread attention due to its fast response speed and low cost. With the dispersed access of massive heterogeneous resources to the power grid, how to deeply tap the regulation potential of massive flexible loads to improve the dispatching capability of the load side of the power grid has become a widely concerned problem.

[0003] Currently, in the field of demand response research in distribution networks, demand response is mainly divided into price-based demand response and incentive-based demand response. Price-based demand response mainly encourages users to stagger peak electricity consumption through time-of-use pricing, real-time pricing and dynamic pricing; incentive-based demand response mainly intervenes in the transfer and reduction of electricity consumption of residential, commercial and industrial users through incentive compensation. The core of demand response is to accurately evaluate the adjustable potential of flexible load and establish a load aggregation model.

[0004] In the field of multi-time scale demand response research, the day-ahead-day-in multi-time scale power transmission and distribution network demand response strategy model based on the two-stage flexible load partition clustering method provides a good demonstration reference for the model design of flexible resource aggregation participating in power grid regulation, but most of the resource modeling of distribution network side is relatively thin, and further modeling is needed to reflect the flexible adjustment capability of distribution network. SUMMARY

[0005] The present application aims to provide a power transmission and distribution network collaborative scheduling method for heterogeneous resource partition clustering to solve the above problems.

[0006] The technical solution of the present application is: a power transmission and distribution network collaborative scheduling method for heterogeneous resource partition clustering, comprising:

[0007] Step S1, constructing a high-load capacity load model and a residential load model;

[0008] Step S2, constructing a two-stage partition clustering model of flexible load;

[0009] Step S3, constructing a flexible load adjustable potential evaluation index based on the partition clustering method;

[0010] Step S4, constructing a day-ahead-day-in multi-time scale power transmission and distribution network collaborative scheduling model;

[0011] Step S5, constructing day-ahead-day-of multi-time scale power transmission and distribution network coordinated dispatching strategy and solving.

[0012] Preferably, in step S1, the high-load energy consumption model includes a discretely adjustable load participation dispatching model and a continuously adjustable load dispatching model; the civil load model includes a heat storage electric boiler load adjustment model, an electric vehicle load adjustment model, and an air conditioning load adjustment model; wherein,

[0013] The discretely adjustable load participation dispatching model is as follows:

[0014]

[0015] In formula (1), the first sub-formula is the upper and lower limit constraint of the output of the discretely adjustable load, wherein respectively represent the minimum and maximum values of the output of the discretely adjustable load, and t is the power consumption of the discretely adjustable load at the time period; the second sub-formula is the minimum running time constraint of the discretely adjustable load, wherein represents the power adjustment flag of the discretely adjustable load at the time period t; represents the minimum running time of the load power; and the third sub-formula is the adjustment flag definition, wherein is the participation adjustment power of the discretely adjustable load at the time period t; and the fourth sub-formula is the adjustment frequency limit of the discretely adjustable load, wherein represents the maximum adjustable frequency of the discretely adjustable load;

[0016] The continuously adjustable load dispatching model includes a continuously adjustable load power upper and lower limit constraint and an adjustment power ramping limit, and is specifically as follows:

[0017]

[0018] In formula (2), represents the load power of the continuously adjustable load at the time period t, respectively represent the upper and lower limits of the continuously adjustable load power, represents the continuously adjustable load power adjustment rate limit value;

[0019] The heat storage electric boiler load adjustment model includes a heat storage boiler power conservation constraint within one day, an actual running constraint, an adjustment rate constraint, and a heat storage capacity constraint, and is specifically as follows:

[0020]

[0021] In formula (3), the first sub-formula is the heat storage electric boiler power conservation constraint within one day, wherein represents the actual heating electric power and the predicted heating electric power of the heat storage electric boiler load at the time period t; and the second sub-formula is the actual running constraint of the heat storage electric boiler, wherein The third formula represents the upper and lower limits of the heating power of the thermal storage electric boiler during time period t; the third formula is the regulation rate constraint of the thermal storage electric boiler, where... The fourth formula represents the maximum load power regulation rate of the thermal storage electric boiler; the fifth formula represents the heat storage capacity constraint of the thermal storage tank. Indicates the maximum and minimum heat storage capacity of the thermal storage tank;

[0022] The electric vehicle load regulation model includes the power conservation constraints, charge / discharge indicators, and power operation constraints of the electric vehicle throughout the day, as shown in the following formula:

[0023]

[0024] In equation (4), the first sub-equation is the power conservation constraint for electric vehicles within a day, where The first formula represents the actual and predicted values ​​of the charging and discharging power of the electric vehicle load during time period t; the second formula is the definition formula for the charging and discharging of electric vehicles, where... This indicates the charging and discharging status of electric vehicles. 1, -1, and 0 represent charging, discharging, and off-grid status, respectively. and These are the rated charge and discharge power, respectively; T in T out These represent the times when the electric vehicle connects to and disconnects from the power grid, respectively; the third and fourth sub-formulas represent the power operation constraints of the electric vehicle, where... Indicates the minimum and maximum storage capacity of the power battery; This indicates the amount of electricity stored when an electric vehicle is connected to the grid. This represents the user's expected energy storage capacity when the device is off-grid.

[0025] In the air conditioning load regulation model, the relationship between real-time electrical power and cooling capacity (or heating capacity) can be described as follows:

[0026]

[0027] In equation (5), Q(t) is the real-time cooling capacity (or heating capacity) of the variable frequency air conditioner; P(t) is the real-time electrical power of the variable frequency air conditioner; f is the operating frequency of the compressor; a and b are constant coefficients for cooling capacity (heating capacity); m and n are constant coefficients for electrical power. Assume the operating frequency of the variable frequency air conditioner varies within the range of [f...]. min ,f max ]; Room temperature is maintained at T set The vicinity, let's assume that the range is [T] min ,T max ];

[0028] The user sets the temperature T. set When the operating frequency f of the inverter air conditioner remains constant, it is related to the indoor temperature T. i The relationship is:

[0029]

[0030] In formula (6), e is the controller accuracy, and is usually taken as 1.

[0031] Preferably, the step S2 comprises: load partitioning based on the quick search and density peak finding algorithm and density-based load clustering; wherein,

[0032] The load partitioning based on the quick search and density peak finding algorithm is specifically:

[0033] Based on two basic assumptions: there is a large local density in the cluster; and the distance between different cluster centers is far apart;

[0034] Each data point derives two feature quantities, the local density ρ i and the distance δ i between nodes with higher local density.

[0035] An electrical distance expectation matrix D * is adopted as a similarity matrix. For a power distribution network with a node number n, when n is small and the spatial density distribution is relatively dense, the local density of node i is defined as:

[0036]

[0037] In formula (8), d c is the truncation distance, which needs to be set artificially. D * is the i-th row and j-th column element of D

[0038] χ(·) is a truncation function, and is defined as:

[0039]

[0040] After the local density of each node is determined, the distance δ i of each node is further calculated.

[0041] First, each node is arranged in ascending order according to the local density value, and the δ i of node i is defined as the electrical distance between the i-th node and the i+1-th node:

[0042]

[0043] For the node with the highest local density, the distance is defined as the distance between the node and the node with the farthest electrical distance:

[0044]

[0045] Solving electrical distance D * The power flow equation of the Newton-Raphson method in polar coordinates is shown as follows:

[0046]

[0047] In formula (12), Δθ, ΔU, ΔP and ΔQ are respectively the transformation vectors of the node phase angle, voltage, active power and reactive power, are respectively the change of the node phase angle caused by the change of the node unit active power injection and the node unit reactive power injection, and are respectively the active power voltage sensitivity and the reactive power voltage sensitivity:

[0048]

[0049] The coupling strength between nodes should be reflected by the influence of the state quantity between nodes, and the voltage sensitivity α i,j between node i and node j is expressed as:

[0050] ΔU i = α i,j ΔU j (14)

[0051] In formula (14), ΔU i and ΔU j respectively represent the voltage change of node i and node j;

[0052] The active power and reactive power comprehensive sensitivity is used to reflect the relationship between voltages:

[0053]

[0054] In formula (15), and are respectively the voltage sensitivity between node i and node j caused by the unit active power and the unit reactive power change;

[0055] The electrical distance is defined by the voltage sensitivity, and the electrical distance of a node to itself is 0. The electrical distance between node i and node j is defined as:

[0056] d i,j = d j,i = log(α i,j · α j,i ) (16)

[0057] The density-based load clustering represents the density between sample data by a set of neighborhood parameters, a radius parameter Eps and a minimum density value MinPts of points in the core point neighborhood, and divides the sample data into clusters; specifically including:

[0058] Step one: randomly select an unprocessed point x in the given sample data set, calculate its corresponding Eps and MinPts, and determine whether it is a core point or a noise point;

[0059] Step two: if x is a core point, calculate all density-reachable data points in the neighborhood of x, and establish a new class cluster;

[0060] Step three: if x is a noise point, do not do anything;

[0061] Step four: repeat steps one to three until all data points are traversed;

[0062] Constructing a power distribution network consumer load aggregation model:

[0063]

[0064] In the formula, C N is the number of flexible load clustering clusters in the region; N k,c is the total number of flexible loads in the kth cluster after clustering; P k,i is the adjustable capacity of the ith flexible load in the kth cluster; N NP is the total number of noise point loads in the region; P j is the capacity of the jth noise point load in the region.

[0065] Preferably, the flexible load adjustable potential evaluation index in step S3 includes cluster peak shaving rate, cluster distance, and cluster energy density; specifically:

[0066] Cluster peak shaving rate: the cluster peak shaving rate is a peak shaving rate representing the smallest dispatching object cluster, and is defined as follows:

[0067]

[0068] In formula (18), γ C is the cluster peak shaving rate; P C,max represents the maximum power of each cluster flexible load before dispatching; P C ' ,max represents the maximum power of each cluster flexible load after dispatching;

[0069] Cluster distance: the cluster distance is defined as the average electrical distance expectation value of each flexible load in the cluster to the dispatching center, and the larger the value, the higher the dispatching cost, and the definition formula is as follows:

[0070]

[0071] In formula (19), δ C is the cluster distance; C N is the total number of flexible loads in the cluster; is the expected value of electrical distance from the i-th flexible load in the cluster to the dispatch center; C Z is the total number of flexible loads in the region; is the expected value of electrical distance from the j-th flexible load in the region to the dispatch center O; C

[0072] Cluster energy density: the cluster energy density refers to the schedulable energy possessed by each cluster of flexible loads per unit space area, and a large cluster energy density represents that the cluster of flexible loads has high-quality characteristics. The definition is as follows:

[0073]

[0074] In formula (20), ρ CP is the cluster energy density.

[0075] Preferably, the step S4 includes a day-ahead scheduling plan model and an intra-day scheduling plan model; specifically:

[0076] The day-ahead scheduling plan model can be expressed in the following form:

[0077]

[0078] In formula (21), the objective function is specifically expressed as follows:

[0079] F1=C G +C H +C J +C WT +C PV +C ESS +C DR (22)

[0080] In formula (22), C G is the operation cost of the thermal power unit; C H and C J are the start-stop costs; C WT and C PV are the abandoned wind and light costs; C ESS is the energy storage maintenance cost; C DR is the incentive DR calling cost;

[0081] wherein C DR is expressed as follows:

[0082]

[0083] In formula (21), ξ1=[ζ1,ζ2,…,ζ n] is a day-ahead decision variable, including thermal power unit output, wind and solar output, flexible load scheduling, energy storage charging and discharging power, and 0-1 decision variable, wherein the 0-1 decision variable includes thermal power unit start-stop state, flexible load calling state and energy storage charging and discharging state; Φ1 = [ψ1, ψ2, …, ψ n ] is a day-ahead power transmission and distribution network mass resource known parameter, including thermal power unit cost coefficient, output upper and lower limit, minimum allowable running time and shutdown time, climbing rate, wind and solar output prediction value, load prediction value, line transmission power limit value, energy storage charging and discharging efficiency, maximum and minimum state of charge;

[0084] The day-ahead scheduling plan model can be expressed in the following form:

[0085]

[0086] The objective function in formula (24) is specifically expressed as follows:

[0087] F2 = C G +C WT +C PV +C ESS +C DR (25)

[0088] In formula (25), C G is the operation cost of the thermal power unit; C H and C J are start-stop costs; C WT and C PV are wind and solar abandonment costs; C ESS is the energy storage maintenance cost; C DR is the incentive DR calling cost;

[0089] Wherein C DR is expressed as follows:

[0090]

[0091] In formula (24), ξ2 = [ζ1, ζ2, …, ζ n ] is a day-ahead decision variable, including thermal power unit output, wind and solar output, flexible load scheduling, energy storage charging and discharging power; Φ2 = [ψ1, ψ2, …, ψ n ] is a day-ahead power transmission and distribution network mass resource known parameter, including thermal power unit cost coefficient, output upper and lower limit, minimum allowable running time and shutdown time, climbing rate, wind and solar output prediction value, load prediction value, line transmission power limit value, energy storage charging and discharging efficiency, maximum and minimum state of charge, and the start-stop state of the thermal power unit, the flexible load calling state and the energy storage charging and discharging state determined in the day-ahead.

[0092] Preferably, in step S5,

[0093] The day-ahead and intraday multi-timescale coordinated dispatch strategy for transmission and distribution networks includes day-ahead 24-hour load dispatch and intraday 4-hour load dispatch; specifically:

[0094] The day-ahead 24-hour load dispatch includes: executing once every 24 hours, dividing into 24 time periods to plan the power of each load agent for the next day. The dispatch targets loads with slower response times and longer advance notice times. The day-ahead dispatch determines the dispatch plan for flexible loads (electrolytic aluminum and silicon carbide) that can be transferred within the next 24 hours in the transmission network, and the dispatch plan for flexible loads (thermal storage electric boilers and electric vehicles) that can be transferred within the next 24 hours in the distribution network. It should also determine the start-up and shutdown of conventional generating units in the transmission and distribution networks and formulate output plans for conventional generating units to guide their power generation.

[0095] 4-hour load dispatching: Executed every 4 hours with a resolution of 15 minutes. Based on the effect of day-ahead load dispatching, the output results of conventional units during the dispatching period are corrected based on the latest new energy output and load forecast data. At the same time, the dispatching plan for flexible loads such as ferroalloy loads that can be reduced in the transmission network in the next 4 hours and the dispatching plan for flexible loads such as air conditioning loads that can be reduced in the distribution network in the next 4 hours are determined. The day-ahead output plan of conventional units is rolled out and optimized.

[0096] The solution to the current scheduling plan model includes: using the objective cascading method to transform the model into a multi-level, multi-agent coordination optimization problem;

[0097] The objective function of the power transmission network has recently been modified as follows:

[0098]

[0099] The objective function for the distribution network has recently been modified as follows:

[0100]

[0101] In the formula, ν k,t and ω k,t For algorithm multipliers; and For coupling variables obtained from adjacent regions, the superscript "-" indicates that the term is a known quantity;

[0102] By setting the penalty function, the coupling variables are made to be as close as possible to the boundary power values ​​transmitted from adjacent regions during the calculation process, and finally achieve consistency.

[0103] The solution to the intraday scheduling model includes:

[0104] The objective function for the intraday power transmission network is modified as follows:

[0105]

[0106] The day-ahead power distribution network objective function is modified as:

[0107]

[0108] In the formula, ν k,t and ω k,t are algorithm multipliers, consistent with the day-ahead algorithm multiplier parameters.

[0109] The beneficial effects of the present application are:

[0110] 1. The application of the demand response in the multi-time scale transmission and distribution network collaborative optimization dispatching improves the economic operation of the system, effectively improves the new energy consumption capacity, and achieves the purpose of peak clipping and valley filling and smoothing the load curve.

[0111] 2. The massive heterogeneous flexible load partition clustering method considers the massive characteristics of heterogeneous resources, and through the two-stage load aggregation of partitioning first and then clustering, can effectively improve the efficiency of the coordinated dispatching of massive heterogeneous resources of the power grid.

[0112] 3. The massive heterogeneous flexible load partition clustering method determines the demand response time sequence dynamic price according to the proposed regulation potential evaluation index, and when the flexible load calling amount is less than 60%, the system dispatching can select the flexible load with high energy density, short spatial distance and good economy. BRIEF DESCRIPTION OF DRAWINGS

[0113] Figure 1 It is a schematic diagram of the transmission and distribution network collaborative dispatching for heterogeneous resource partition clustering;

[0114] Figure 2 It is a schematic diagram of the flexible load adjustable potential evaluation method and index of the distribution network;

[0115] Figure 3 It is a block diagram of the day-ahead-intra-day multi-time scale transmission and distribution network dispatching strategy;

[0116] Figure 4 It is a two-stage resource schematic diagram of the transmission network;

[0117] Figure 5 It is a time sequence dynamic demand response price. DETAILED DESCRIPTION

[0118] The present application will be further described below in conjunction with the drawings and specific embodiments, so that those skilled in the art can better understand the present application and implement it, and the embodiments of the present application are not limited thereto.

[0119] Example 1

[0120] As Figure 1The flow chart of the application is shown, first, the high load energy load model and the civil load model considering the differentiation of flexible load are established; the flexible load partition clustering model based on digital model fusion is established, then the adjustment potential evaluation index is established according to the flexible load aggregation method; finally, the day-ahead-day-in multi-time scale power transmission and distribution network demand response strategy model is established, which contains the power flow constraint and the coupling constraint of power transmission and distribution network, forming the demand response scheduling model of power transmission and distribution network.

[0121] A power transmission and distribution network collaborative scheduling strategy considering mass heterogeneous resource partition clustering, comprising the following steps:

[0122] There are many kinds of mass heterogeneous resources in power transmission and distribution network, and the response characteristics have great difference. The flexible load in power transmission network mainly includes high load energy load such as electrolytic aluminum, ferroalloy and silicon carbide; the flexible load in power distribution network mainly includes civil load such as heat storage electric boiler, electric vehicle and air conditioner.

[0123] In terms of spatial scale, high load energy load has the characteristics of aggregation distribution, and the individual capacity is large and the number is limited. In terms of time scale, electrolytic aluminum and silicon carbide have the characteristics of discrete adjustment, and can only participate in system optimization scheduling in day-ahead time scale; ferroalloy load has the characteristics of continuous adjustment, and is suitable for participating in system optimization scheduling in day-in time scale.

[0124] The load such as electrolytic aluminum has fast adjustment rate and large ramp rate in the actual industrial production process. Ignoring the ramp constraint, considering the upper and lower limit constraint, the operation time constraint and the adjustment times constraint, the discrete adjustment load participating in the scheduling model is shown in the following formula:

[0125]

[0126] In the formula, the first sub-formula is the upper and lower limit constraint of the discrete adjustment load, wherein represents the minimum and maximum values of the discrete adjustment load, and t is the power of the discrete adjustment load at the time period; the second sub-formula is the minimum operation time constraint of the discrete adjustment load, wherein represents the power adjustment flag of the discrete adjustment load at the time period t; represents the minimum operation time of the load power; the third sub-formula is the definition of the adjustment flag, wherein is the participating adjustment power of the discrete adjustment load at the time period t; the fourth sub-formula is the adjustment times limit of the discrete adjustment load, wherein represents the maximum adjustment times of the discrete adjustment load.

[0127] The continuous adjustment load scheduling model includes the upper and lower limit constraint of the continuous adjustment load power and the adjustment power ramp limit, and is shown in the following formula:

[0128]

[0129] wherein, represents the load power of the continuously adjustable load at time t, respectively represent the upper and lower limits of the continuously adjustable load power, represents the limit value of the continuously adjustable load power adjustment rate.

[0130] In terms of spatial scale, civil loads have the characteristics of wide-area dispersion, small individual capacity and large quantity. In terms of time scale, according to the response speed, the response speed of the heat storage electric boiler and the electric vehicle is slow, and they are considered to participate in day-ahead scheduling, and the air conditioning load has a fast adjustment rate and participates in intra-day scheduling.

[0131] The heat storage electric boiler load adjustment model includes the power conservation constraint of the heat storage boiler within one day, the actual operation constraint, the adjustment rate constraint and the heat storage quantity constraint, and is specifically as follows:

[0132]

[0133] wherein, the first sub-formula is the power conservation constraint of the heat storage boiler within one day, wherein represents the actual heating electric power and the predicted heating electric power of the heat storage electric boiler load at time t; the second sub-formula is the actual operation constraint of the heat storage boiler, wherein represents the upper and lower limit values of the heating power of the heat storage electric boiler load at time t; the third sub-formula is the adjustment rate constraint of the heat storage boiler, wherein represents the maximum value of the adjustment rate of the heat storage electric boiler load power; the fourth sub-formula is the heat storage quantity constraint of the heat storage tank, wherein represents the maximum and minimum heat storage quantities of the heat storage tank.

[0134] The establishment of the electric vehicle load adjustment model is similar to that of the heat storage electric boiler, and includes the power conservation constraint of the electric vehicle within one day, the charging and discharging identification and the power operation constraint, and is as follows:

[0135]

[0136] wherein, the first sub-formula is the power conservation constraint of the electric vehicle within one day, wherein represents the actual value and the predicted value of the charging and discharging power of the electric vehicle load at time t; the second sub-formula is the definition formula of the charging and discharging identification of the electric vehicle, wherein represents the charging and discharging identification of the electric vehicle, 1, -1 and 0 respectively represent charging, discharging and non-grid-connected state; and are respectively the rated charging and discharging power; T in and T out respectively represent the time when the electric vehicle accesses the power grid and leaves the power grid; the third and fourth sub-formulas represent the power operation constraint of the electric vehicle, wherein Indicates the minimum and maximum storage capacity of the power battery; This indicates the amount of electricity stored when an electric vehicle is connected to the grid. This represents the user's expected energy storage capacity when the device is off-grid.

[0137] The thermal principle of an air-conditioned room is described using a first-order equivalent thermal parameter model. The relationship between the real-time electrical power and cooling capacity (or heating capacity) of a variable frequency air conditioner can be described as follows:

[0138]

[0139] In the formula, Q(t) is the real-time cooling capacity (or heating capacity) of the inverter air conditioner; P(t) is the real-time electrical power of the inverter air conditioner; f is the operating frequency of the compressor; a and b are constant coefficients for cooling capacity (heating capacity); m and n are constant coefficients for electrical power. Assume the operating frequency of the inverter air conditioner varies within the range of [f...]. min ,f max ]; Room temperature is maintained at T set The vicinity, let's assume that the range is [T] min ,T max ];

[0140] The user sets the temperature T. set When the operating frequency f of the inverter air conditioner remains constant, it is related to the indoor temperature T. i The relationship is:

[0141]

[0142] In the formula, e represents the controller accuracy, which is usually taken as 1.

[0143] Although high-energy-consuming loads are not scheduled on the same time scale, their distribution is concentrated and their power adjustment is convenient. Civilian flexible loads, on the other hand, have smaller power, are widely distributed, and their operation is random, making them impossible to directly call upon by the system. Therefore, this invention proposes a method and index for evaluating the adjustable potential of civil loads in a distribution network based on digital-analog fusion and zonal clustering.

[0144] The vast distribution of resources in the power distribution network necessitates partitioning these resources before load clustering to reduce the complexity of subsequent clustering tasks. The model employs the Fast Search and Find of Density Peaks (FSFDP) algorithm to partition the civilian loads of the power distribution network. FSFDP divides data points into several clusters using a similarity matrix. This algorithm is based on two fundamental assumptions: ① the existence of large local densities within clusters; and ② the significant distance between different cluster centers. Consequently, each data point derives two features: local density ρ. i and the distance δ between nodes with higher local densityi The cluster center is determined, and the remaining nodes are divided into the class to which the node with the nearest distance and higher local density belongs.

[0145] The electrical distance expectation matrix D * As a similarity matrix, the local density of node i is defined as:

[0146]

[0147] In the formula, d c is the cut-off distance, which needs to be set artificially. The i-th row and j-th column element of D * represents the electrical distance expectation value between node i and node j. χ(·) is a cut-off function.

[0148]

[0149] After determining the local density of each node, the distance δ i of each node is further calculated. i First, each node is arranged in ascending order according to the local density value. The δ i of node i is defined as the electrical distance between the i+1th node.

[0150]

[0151] For the node with the highest local density, the distance is defined as the distance between the node with the farthest electrical distance:

[0152]

[0153] According to the core idea of the algorithm, the cluster center should have a high local density ρ i and a far distance δ i . The binary feature quantity (ρ i , δ i ) of each node is taken as the two-dimensional coordinate point (ρ i is the horizontal coordinate and δ * is the vertical coordinate) of the node. The decision graph is drawn, and the cluster center is identified through the decision graph.

[0154] The data processing stage is to solve the electrical distance D * . The power flow equation of the Newton-Raphson method in polar coordinates is as follows:

[0155]

[0156] In the formula, Δθ, ΔU, ΔP and ΔQ are the transformation vectors of the node phase angle, voltage, active power and reactive power, respectively, respectively, are the change of phase angle of node i caused by the change of active power injection and reactive power injection of node i respectively, and respectively, are the active voltage sensitivity and reactive voltage sensitivity.

[0157]

[0158] The coupling strength between nodes should be reflected by the influence of state variables between nodes. The voltage sensitivity α i,j between node i and node j is defined as:

[0159] ΔU i = α i,j ΔU j (44)

[0160] In the formula, ΔU i and ΔU j are the voltage change of node i and node j respectively. Since the resistance and reactance of distribution network are comparable in magnitude, the influence of active power flow on voltage cannot be ignored, so the comprehensive sensitivity of active and reactive power is used to reflect the relationship between voltages.

[0161]

[0162] In the formula, and are the voltage sensitivity between node i and node j caused by the change of unit active power and reactive power respectively.

[0163] Further, the electrical distance is defined by the voltage sensitivity. The mutual influence between nodes with short electrical distance is large, and the mutual influence between nodes with long electrical distance is small. Since the power flow has directionality, α i,j and α j,i are often not equal. In order to eliminate this asymmetry, and considering that the electrical distance of a node to itself is 0, the electrical distance between node i and node j is defined as:

[0164] d i,j = d j,i = log(α i,j · α j,i ) (46)

[0165] After the above-mentioned partitioning of the massive resources of the power distribution network, the regional aggregator needs to aggregate the widely distributed, large in number, and small in power domestic loads in the region to participate in system dispatch. A density-based clustering algorithm (Density-based spatial clustering of applications with noise, DBSCAN) is used to cluster the regional loads, eliminate single long-distance noise point loads with high calling cost, and realize the aggregation of domestic loads in the region and the assessment of dispatch potential.

[0166] The DBSCAN algorithm has fast clustering speed and can effectively process spatial clustering of any shape. The algorithm expresses the density between sample data through a set of neighborhood parameters, radius parameter Eps and minimum density value MinPts of points in the neighborhood of the core point, and divides the sample data into clusters. The general steps of the algorithm are as follows:

[0167] Step 1: Randomly select an unprocessed point x in the given sample data set, calculate the corresponding Eps and MinPts of x, and determine whether x is a core point or a noise point;

[0168] Step 2: If x is a core point, calculate all density-reachable data points in the neighborhood of x, and establish a new class cluster;

[0169] Step 3: If x is a noise point, do not perform any processing;

[0170] Step 4: Repeat steps 1 to 3 until all data points are traversed.

[0171] Although the DBSCAN algorithm has fast clustering speed, since the algorithm directly processes all sample data during clustering, the time complexity of the algorithm is O(n 2 ), so when the amount of data to be processed is large, a large amount of memory will be consumed, and the parameters used in the clustering process are global, so there is a certain limitation when processing large-scale data. On the basis of the above-mentioned partitioning of the massive resources of the power distribution network, the domestic loads in each region are clustered, and the number of resource samples in the region is greatly reduced. The use of this algorithm can reduce the time complexity and quickly complete the aggregation and potential assessment of domestic loads in the region.

[0172] According to the above-mentioned partitioning and aggregation method, a domestic load aggregation model of the power distribution network can be obtained:

[0173]

[0174] In the formula, C N is the number of clusters of flexible loads in the region; N k,c is the total number of flexible loads in the kth cluster after clustering; and P k,iis the ith flexible load adjustable capacity in the kth cluster; N NP is the total number of noise point loads in the region; P j is the capacity of the jth noise point load in the region;

[0175] Adjustable potential refers to the potential of dispatchable resources participating in system optimization scheduling, and its evaluation can provide a reference standard for incentive compensation issued by system scheduling. According to the above-mentioned power distribution network residential load partition clustering method based on digital-analog fusion of the present application, the corresponding cluster peak clipping rate, cluster distance and cluster energy density are proposed, and the dispatchable potential evaluation index considering the difference in adjustment characteristics of massive heterogeneous resources is considered.

[0176] 1) Cluster peak clipping rate: The peak clipping rate refers to the change rate of the load peak value before and after the load participates in the grid scheduling. Peak clipping can smooth the load curve and relieve the peak regulation pressure of thermal power units after a high proportion of new energy is accessed. The cluster peak clipping rate proposed in the present application is the peak clipping rate of the smallest scheduling object cluster, and is defined as follows:

[0177]

[0178] In the formula, γ C is the cluster peak clipping rate; P C,max represents the maximum power of each cluster flexible load before scheduling; P' C,max represents the maximum power of each cluster flexible load after scheduling.

[0179] 2) Cluster distance: The cluster distance is defined as the average electrical distance expectation value of each flexible load in the cluster to the scheduling center, and the larger the value, the higher the scheduling cost, and the definition formula is as follows:

[0180]

[0181] In the formula, δ C is the cluster distance; C N is the total number of flexible loads in the cluster; is the electrical distance expectation value of the ith flexible load in the cluster to the scheduling center; C Z is the total number of flexible loads in the region; is the electrical distance expectation value of the flexible load j in the region to the scheduling center O.

[0182] 3) Cluster energy density: The cluster energy density refers to the dispatchable energy possessed by each cluster flexible load in unit space area, and the larger the cluster energy density, the better the characteristics of the cluster flexible load. The definition formula is as follows:

[0183]

[0184] ρ CP is the cluster energy density.

[0185] As Figure 2 shown in the figure is a method and index diagram for evaluating the adjustable potential of flexible load in power distribution network based on digital-analog fusion. Considering the large difference and wide distribution of flexible load characteristics in power distribution network, FSFDP algorithm is used to quickly partition large-scale power distribution network by taking electrical distance as similarity matrix; load partitioning solves the problem of time complexity explosion of DBSCAN algorithm for massive data, and clusters regional load in any shape; for each cluster of flexible load after partitioning and clustering, three levels of cluster peak shaving rate, cluster distance and cluster energy density are proposed to consider the difference characteristics of flexible load, and then the flexible load is selectively called to obtain time sequence dynamic demand response price.

[0186] The day-ahead transmission system takes the minimum economic cost as the target, and takes power balance, line transmission power, generator output, minimum operation time, minimum downtime, unit ramping, renewable energy output, energy storage operation and flexible load calling restriction as constraint conditions. The day-ahead dispatching plan model can be expressed in the following form:

[0187]

[0188] The objective function is specifically expressed as follows:

[0189] F1=C G +C H +C J +C WT +C PV +C ESS +C DR (52)

[0190] In the formula, C G is the operation cost of thermal power unit; C H and C J are start-stop costs; C WT and C PV are wind and light abandonment costs; C ESS is the energy storage maintenance cost; C DR is the incentive DR calling cost. Wherein C DR is expressed as follows:

[0191]

[0192] In formula (21), ξ1=[ζ1,ζ2,…,ζ n ] is the day-ahead decision variable, including thermal power unit output, wind and light output, flexible load scheduling quantity, energy storage charging and discharging power, and a large number of 0-1 decision variables, including thermal power unit start-stop state, flexible load calling state and energy storage charging and discharging state; Φ1=[ψ1,ψ2,…,ψ nThe day-ahead transmission and distribution network mass resource known parameters include thermal power unit cost coefficient, output upper and lower limits, minimum allowable operating time and downtime, ramp rate, wind and solar power prediction values, load prediction values, line transmission power limit values, energy storage charging and discharging efficiency, maximum and minimum state of charge.

[0193] The day-ahead transmission and distribution network mass resource known parameters include thermal power unit cost coefficient, output upper and lower limits, minimum allowable operating time and downtime, ramp rate, wind and solar power prediction values, load prediction values, line transmission power limit values, energy storage charging and discharging efficiency, maximum and minimum state of charge.

[0194]

[0195] The objective function is specifically represented as follows:

[0196] F2=C G +C WT +C PV +C ESS +C DR (55)

[0197] In the formula, C G is the operating cost of the thermal power unit; C H and C J are the start-up and shutdown costs; C WT and C PV are the abandoned wind and solar power costs; C ESS is the energy storage maintenance cost; and C DR is the incentive DR calling cost. C DR is represented as follows:

[0198]

[0199] In formula (24), ξ2=[ζ1,ζ2,…,ζ n ] is the day-ahead decision variable, including the thermal power unit output, wind and solar power, flexible load scheduling quantity, and energy storage charging and discharging power; Φ2=[ψ1,ψ2,…,ψ n ] is the day-ahead transmission and distribution network mass resource known parameter, including the thermal power unit cost coefficient, output upper and lower limits, minimum allowable operating time and downtime, ramp rate, wind and solar power prediction values, load prediction values, line transmission power limit values, energy storage charging and discharging efficiency, maximum and minimum state of charge, and the day-ahead determined thermal power unit start-up and shutdown state, flexible load calling state, and energy storage charging and discharging state.

[0200] Figure 3 The day-ahead-day-ahead multi-time scale transmission and distribution network coordinated scheduling strategy block diagram mainly includes two coordinated processes of day-ahead planning and day-ahead correction.

[0201] Day-ahead 24h load dispatching: it is executed once every 24h, and the power of each load agent in the future 1 day is planned in 24 time periods, and the dispatching object is the load in the load agent with slow response speed and long advance notice time. The day-ahead dispatching determines the flexible load aluminum electrolysis, silicon carbide load dispatching plan that can be transferred in the future 24h of the power grid, the flexible load heat storage electric boiler, electric vehicle load dispatching plan that can be transferred in the future 24h of the distribution network, and the start-stop of the conventional unit of the power grid and distribution network should be determined, and the conventional unit output plan is made for guiding the conventional unit power generation.

[0202] Intra-day 4h load dispatching: it is executed once every 4h, and the resolution is 15 minutes, on the basis of considering the effect of day-ahead load dispatching, the latest new energy output and load prediction data are used to correct the conventional unit output result in the dispatching period, at the same time, the flexible load ferroalloy load dispatching plan that can be reduced in the future 4h of the power grid, the flexible load air conditioning load dispatching plan that can be reduced in the future 4h of the distribution network, and the conventional unit day-ahead output plan are corrected and optimized.

[0203] The solving model of the application is a mixed integer linear programming problem (Mised Integer Linear Programming, MILP), and the decision variable is composed of a large number of discrete variables and continuous variables. As a typical NP difficult problem, the number of feasible solutions of the MILP problem increases exponentially with the scale of the discrete variable, and the optimal solution of the problem cannot be obtained by exhaustion in a polynomial. The solving is directly called by calling different algorithm solvers, which is simple and feasible, and has a certain universality, but for large-scale problems, it still faces the actual bottleneck of slow convergence or difficult convergence.

[0204] The discrete information in the model of the application comes from the day-ahead unit start-stop information of the power grid and the distribution network and the information whether the flexible load participates in the dispatching. Considering the solving difficulty brought by a large number of discrete variables in the model of the application, the target cascade method is used to convert the model of the application into a multi-level, multi-agent coordinated optimization problem. The above model is modified as follows:

[0205] The day-ahead power grid target function is modified as:

[0206]

[0207] The day-ahead distribution network target function is modified as:

[0208]

[0209] In the formula, ν k,t And ω k,t Are algorithm multipliers; And For the coupling variable obtained from the adjacent area, the superscript "-" indicates that the item is a known quantity. Through the setting of the penalty function, the coupling variable is made to approach the boundary power value transmitted by the adjacent area as much as possible during the calculation process, and finally reaches consistency. The intra-day and day-ahead modification is consistent.

[0210] The model solution of the intra-day dispatching plan model includes:

[0211] The intra-day power grid objective function is modified as:

[0212]

[0213] The day-ahead distribution network objective function is modified as:

[0214]

[0215] In the formula, ν k,t and ω k,t are algorithm multipliers, which are consistent with the day-ahead algorithm multiplier parameters.

[0216] Based on the above analysis, the application adopts one six-node power grid and three distribution network system data as examples for simulation analysis. The three distribution networks are improved IEEE14-node, 33-node and 39-node distribution network systems. Combined with the 24h wind and light power plant and original load data in a certain region, simulation verification is carried out. The wind and light output prediction error considers two time scales of day-ahead and intra-day, and the demand response mainly considers the massive heterogeneous resources of the power transmission and distribution network.

[0217] In order to analyze the advantages of demand response in multi-time scale power transmission and distribution network coordinated dispatching, four dispatching optimization strategies are designed and compared under the condition that other parameters are the same. Strategy 1: Separate dispatching of power transmission and distribution network, without IDR resource participation; Strategy 2: Power transmission and distribution network coordinated dispatching, without IDR resource participation; Strategy 3: Power transmission and distribution network coordinated dispatching, IDR resources are all in day-ahead dispatching; Strategy 4: The dispatching strategy of the present application. The comparison of various effects under different strategies is shown in Table 1:

[0218] Table 1 Comparison of various effects under different strategies

[0219]

[0220] Compared to Strategy 1, under coordinated dispatch of transmission and distribution networks, renewable energy that cannot be absorbed by a single grid can be absorbed by the remaining grid, resulting in a decrease in the total system operating cost and the curtailment rate of wind and solar power. However, since the peak periods of renewable energy generation often coincide with off-peak periods, and the current-day renewable energy forecasting error is relatively large, the decrease in the curtailment rate of wind and solar power is not significant. Comparing Strategy 2 and Strategy 3, system dispatch allows transferable loads (such as silicon carbide and electric vehicles) to be shifted from peak to off-peak periods, while loads that can be reduced (such as ferroalloys and air conditioners) are reduced during peak periods. The coordinated dispatch of massive heterogeneous resources in the transmission and distribution networks achieves the goals of peak shaving and valley filling, and smoothing the load curve. After flexible loads participate in system dispatch, the large-scale dispatch of flexible loads in the distribution network slightly increases the total cost, but the overall economy of the transmission and distribution network is slightly lower than under Strategy 2, and the peak shaving rate reaches 34.3%. Strategy 4 adopts a two-stage flexible load participation in system scheduling, which fully utilizes the adjustment potential of flexible loads. The rapid adjustment of loads that can be reduced within the day is combined with the high-precision forecasting of renewable energy within the day, so that the renewable energy in the system is almost completely absorbed (extreme cases may cause large forecast deviations), and the peak shaving rate is further improved.

[0221] Taking the results of optimizing the massive internal resources of the power transmission network in two phases from the day-to-day as an example, by Figure 4 It can be seen that during the daytime phase, the load response of transferable high-energy-consuming loads is to reduce industrial production during peak load periods and to increase industrial production during off-peak periods. This reduces the system's peak-to-valley difference without affecting product quality and output, thus benefiting the system and smoothing the load curve. During the intraday phase, wind and solar forecasting errors still exist but are relatively small, allowing for faster load response and adjustment. This can, to some extent, address the curtailment of wind and solar power caused by forecasting errors, reducing total load during peak periods and achieving the goal of absorbing new energy sources and smoothing out peak loads.

[0222] 1) Analysis of the computational speed of partitioned clustering method for load scheduling: To analyze the impact of partitioned clustering of massive heterogeneous resources on computational speed, four heterogeneous resource models of different scales were designed and compared under the same parameter conditions. Scale 1: IEEE 33-node load model; Scale 2: IEEE 145-node load model; Scale 3: Monte Carlo simulation generating 1000-node load model; Scale 4: Monte Carlo simulation generating 10000-node load model. The computational speed comparison of heterogeneous resource models of different scales is shown in Table 2.

[0223] Table 2 Comparison of computation speeds for heterogeneous resource models of different scales

[0224]

[0225] Although the DBSCAN algorithm has fast clustering speed, the algorithm complexity is O(n 2 ) because the algorithm directly operates all samples in the clustering process. The partition clustering method based on digital-analog fusion proposed in the present application has a longer calculation time than direct clustering when the scale is small due to the complexity of the method; but the superiority of the partition clustering method is embodied when mass resources are clustered, the calculation time is greatly shortened, and the calculation effect is effectively improved.

[0226] 2) Analysis of the calling cost of the partition clustering method for load participation in dispatching: In order to analyze the influence of the regulation potential on demand response, two demand response participation dispatching strategies are designed and compared under the condition that other parameters are the same. Strategy 1: without considering the regulation potential, the uniform price of incentive demand response; Strategy 2: considering the regulation potential of flexible load, the demand response time sequence dynamic price is determined according to the regulation potential evaluation index. The strategy comparison under different flexible load calling amounts is shown in Table 3:

[0227] Table 3 Strategy comparison under different flexible load calling amounts

[0228]

[0229] When the uniform price of incentive demand response is used, the difference characteristics of mass flexible load are not fully considered, and the domestic load with large distribution area, small individual capacity and large quantity cannot be deeply excavated. The regulation potential evaluation index proposed in the present application is used to determine the demand response time sequence dynamic price. After the domestic load is partitioned and clustered, the regulation potential of each cluster of domestic load is comprehensively evaluated from the cluster peak shaving rate, cluster distance and cluster energy density. The three evaluation indexes consider the difference characteristics of mass domestic load from the dispatching capacity, dispatching space and dispatching efficiency, and evaluate the advantages and disadvantages of the regulation potential of each cluster of domestic load. It can be seen from Table 3 that when the calling amount of flexible load is less than 60%, the system will consider the difference characteristics of the domestic load, select the high-quality cluster of domestic load evaluated by the index, and reduce the economy thereof; when the calling amount of flexible load is higher than 70%, the system has to call the domestic load with poor economy and poor quality, resulting in a higher economy than the price strategy considering the regulation potential of the load.

[0230] As shown in Figure 5 , the three curves are respectively the time sequence dynamic demand response price unit value of three clusters of flexible load after partition clustering. The time sequence dynamic demand response price comprehensively considers the three regulation potential evaluation indexes of cluster peak shaving rate, cluster distance and cluster energy density, and can provide more economical and high-quality flexible load for system dispatching at different demand amounts, which is different from the uniform demand response price, which only considers the demand amount and ignores the difference in geographical position of flexible load and the energy density.

[0231] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for coordinated scheduling of power transmission and distribution networks based on heterogeneous resource partitioning and clustering, characterized in that, include: Step S1: Construct a high-energy-consuming load model and a residential load model; Step S2: Construct a two-stage partitioned clustering model for flexible loads; Step S3: Construct a flexible load adjustment potential assessment index based on the partition clustering method; Step S4: Construct a multi-timescale collaborative scheduling model for the transmission and distribution network from the day-ahead to intraday. Step S5 involves constructing and solving a multi-timescale coordinated scheduling strategy for the transmission and distribution network from the day-ahead to intraday period; among which, In step S1, the high-energy-consuming load model includes a discretely adjustable load dispatching model and a continuously adjustable load dispatching model; the residential load model includes a thermal storage electric boiler load regulation model, an electric vehicle load regulation model, and an air conditioning load regulation model; among them, The discretely adjustable load participation scheduling model is shown in the following equation: In equation (1), the first sub-equation represents the upper and lower limits of the output of the discretely adjustable load, where... Let represent the minimum and maximum output values ​​of the discretely adjustable load, respectively, and represent the power consumption of the discretely adjustable load during time period t; the second sub-formula represents the minimum operating time constraint of the discretely adjustable load, where... Indicates the adjustable load power regulation flag for time period t; The third formula represents the minimum operating time of the load power; the third formula defines the adjustment flag, where... The fourth formula represents the power of the discretely adjustable load participating in regulation during time period t; the fifth formula is the limit on the number of times the discretely adjustable load can be regulated. This indicates the maximum number of times the discretely adjustable load can be adjusted; The continuously adjustable load dispatch model includes upper and lower limits for continuously adjustable load power and ramp-up limits for adjustable power, as shown in the following equation: In equation (2), This indicates the load power that can be continuously adjusted during time period t. These represent the upper and lower limits of the continuously adjustable load power, respectively. This indicates the limit of the continuously adjustable load power regulation rate; The load regulation model for thermal storage electric boilers includes the boiler's power conservation constraints, actual operating constraints, regulation rate constraints, and heat storage constraints within a day, as detailed below: In equation (3), the first sub-equation is the power conservation constraint of the thermal storage electric boiler within one day, where The first formula represents the actual heating power and the predicted heating power of the thermal storage electric boiler during time period t; the second formula represents the actual operating constraints of the thermal storage electric boiler, where... The third formula represents the upper and lower limits of the heating power of the thermal storage electric boiler during time period t; the third formula is the regulation rate constraint of the thermal storage electric boiler, where... The fourth formula represents the maximum load power regulation rate of the thermal storage electric boiler; the fifth formula represents the heat storage capacity constraint of the thermal storage tank. Indicates the maximum and minimum heat storage capacity of the thermal storage tank; The electric vehicle load regulation model includes the power conservation constraints, charge / discharge indicators, and power operation constraints of the electric vehicle throughout the day, as shown in the following formula: In equation (4), the first sub-equation is the power conservation constraint for electric vehicles within a day, where The first formula represents the actual and predicted values ​​of the charging and discharging power of the electric vehicle load during time period t; the second formula is the definition formula for the charging and discharging of electric vehicles, where... This indicates the charging and discharging status of electric vehicles. 1, -1, and 0 represent charging, discharging, and off-grid status, respectively. and These are the rated charge and discharge power, respectively; T in T out These represent the times when the electric vehicle connects to and disconnects from the power grid, respectively; the third and fourth sub-formulas represent the power operation constraints of the electric vehicle, where... Indicates the minimum and maximum storage capacity of the power battery; This indicates the amount of electricity stored when an electric vehicle is connected to the grid. This represents the user's expected energy storage capacity when the device is off-grid. In the air conditioning load regulation model, the relationship between real-time electrical power and cooling capacity (or heating capacity) can be described as follows: In equation (5), Q(t) is the real-time cooling capacity (or heating capacity) of the variable frequency air conditioner; P(t) is the real-time electrical power of the variable frequency air conditioner; f is the operating frequency of the compressor; a and b are constant coefficients for cooling capacity (heating capacity); m and n are constant coefficients for electrical power. Assume the operating frequency of the variable frequency air conditioner varies within the range of [f...]. min ,f max ]; Room temperature is maintained at T set The vicinity, let's assume that the range is [T] min ,T max ]; The user sets the temperature T. set When the operating frequency f of the inverter air conditioner remains constant, it is related to the indoor temperature T. i The relationship is: In equation (6), e represents the controller accuracy, which is usually taken as 1; Step S2 includes: load partitioning based on a fast search and density peak detection algorithm and density-based load clustering; wherein, The load partitioning based on the fast search and density peak detection algorithm is as follows: Based on two fundamental assumptions: large local densities exist within clusters; and the distances between different cluster centers are large. Each data point derives two features: local density ρ. i and the distance δ between nodes with higher local density i Determine the cluster center, and then assign the remaining nodes to the class of the nearest node with the highest local density; Using the electrical distance expectation matrix D * As a similarity matrix, considering a distribution network with n nodes, when n is small and the spatial density distribution is relatively dense, the local density of node i is defined as: In equation (8), d c To cut off the distance, it needs to be set manually. D * The element in the i-th row and j-th column represents the expected electrical distance between node i and node j; χ(·) is the cutoff function: After determining the local density of each node, the distance δ of each node is further calculated. i ; First, sort each node in ascending order of its local density value, and then consider the δ value of node i. i Defined as the electrical distance to the (i+1)th node: For the node with the highest local density, its distance is defined as the distance between it and the node with the greatest electrical distance from it: Solve for electrical distance D * The power flow equations for the Newton-Raphson method in polar coordinates are shown below: In equation (12), Δθ, ΔU, ΔP, and ΔQ are the vectors representing the transformed quantities of node phase angle, voltage, active power, and reactive power, respectively. These represent the changes in node phase angle caused by changes in unit active power injection and reactive power injection, respectively. and These are active voltage sensitivity and reactive voltage sensitivity, respectively: The coupling strength between nodes should be reflected by the influence of the state variables between nodes, such as the voltage sensitivity α between node i and node j. i,j Represented as: D.U. i =a i,j D.U. j (14) In equation (14), ΔU i and ΔU j These are represented as the voltage changes at node i and node j, respectively. The relationship between voltages is reflected by the combined active and reactive power sensitivity: In equation (15), and These are the voltage sensitivities between node i and node j caused by unit active and reactive power changes, respectively. The electrical distance is defined by voltage sensitivity, and considering that the electrical distance between a node and itself is 0, the electrical distance between node i and node j is defined as follows: d i,j =d j,i =log(α i,j ·a j,i ) (16); Density-based load clustering uses a set of neighborhood parameters (radius parameter Eps) and the minimum density value (MinPts) of points within the neighborhood of the core point to represent the density between sample data, and divides the sample data into clusters; specifically including: Step 1: Randomly select an unprocessed point x in the given sample dataset, calculate its corresponding Eps and MinPts, and determine whether it is a core point or a noise point. Step 2: If x is the core point, calculate all data points with reachable density in the neighborhood of x and establish a new cluster; Step 3: If x is a noise point, no processing is performed; Step 4: Repeat steps 1 through 3 until all data points have been traversed; Constructing a power distribution network residential load aggregation model: In the formula, C N N represents the number of flexible load clusters within the region. k,c P is the total number of flexible loads within the k-th cluster after clustering; k,i It is the adjustable capacity of the i-th flexible load within the k-th cluster; N NP It is the total number of noise point loads within the area; P j Let be the load capacity of the j-th noise point within the region; The evaluation indicators for the adjustable potential of flexible load in step S3 include cluster clipping rate, cluster distance, and cluster energy density; specifically: Cluster clipping rate: The cluster clipping rate is the rate at which the minimum scheduled object clusters are clipped, and is defined as follows: In equation (18), γ C For cluster clipping peak ratio; P C,max P' represents the maximum power of each cluster of flexible loads before scheduling; C,max This indicates the maximum power of each cluster of flexible loads after scheduling; Cluster distance: Cluster distance is defined as the expected average electrical distance from each flexible load within a cluster to the dispatch center. A larger value indicates higher dispatch costs. The definition is as follows: In equation (19), δ C C is the cluster distance; N This represents the total number of flexible loads within the cluster. Let C be the expected electrical distance from the i-th flexible load within the cluster to the dispatch center; Z This represents the total number of flexible loads in the area. Let J be the expected electrical distance between the flexible load j and the dispatch center O within this area; Cluster energy density: Cluster energy density refers to the dispatchable energy possessed by each cluster of flexible loads per unit area. A high cluster energy density indicates superior characteristics of the flexible load cluster. The definition is as follows: In equation (20), ρ CP The energy density of the cluster.

2. The method for coordinated scheduling of power transmission and distribution networks based on heterogeneous resource partitioning and clustering as described in claim 1, characterized in that, Step S4 includes: day-ahead scheduling model and intraday scheduling model; specifically: The day-ahead scheduling model can be represented in the following form: The objective function in equation (21) is specifically expressed as follows: F1=C G +C H +C J +C WT +C PV +C ESS +C DR (22) In equation (22), C G C is the operating cost of thermal power units; H and C J For start-stop costs; C WT and C PV Costs associated with wind and solar power curtailment; C ESS For energy storage maintenance costs; C DR Cost of incentivized DR calls; Where C DR It is expressed as follows: In formula (21), ξ1=[ζ1,ζ2,···,ζ n [ψ1, ψ2, ..., ψ3] represents the day-ahead decision variables, including thermal power unit output, wind and solar power output, flexible load dispatching, energy storage charging and discharging power, and 0-1 decision variables. The 0-1 decision variables include the start-up and shutdown status of thermal power units, the flexible load dispatching status, and the energy storage charging and discharging status. n The parameters of the massive resources of the power transmission and distribution network are known at present, including the cost coefficient of thermal power units, upper and lower limits of output, minimum allowable operating time and downtime, ramp rate, wind and solar power output forecast, load forecast, line transmission power limit, energy storage charging and discharging efficiency, and maximum and minimum state of charge. The intraday scheduling model can be represented in the following form: The objective function in equation (24) is specifically expressed as follows: F2=C G +C WT +C PV +C ESS +C DR (25) In equation (25), C G C is the operating cost of thermal power units; H and C J For start-stop costs; C WT and C PV Costs associated with wind and solar power curtailment; C ESS For energy storage maintenance costs; C DR Cost of incentivized DR calls; Where C DR It is expressed as follows: In formula (24), ξ2=[ζ1,ζ2,···,ζ n ] represents intraday decision variables, including thermal power unit output, wind and solar power output, flexible load dispatching, and energy storage charging and discharging power; Φ2=[ψ1,ψ2,···,ψ n The parameters of the massive resources of the power transmission and distribution network during the day include the cost coefficient of thermal power units, upper and lower limits of output, minimum allowable operating time and downtime, ramp rate, wind and solar power output forecast, load forecast, line transmission power limit, energy storage charging and discharging efficiency, maximum and minimum state of charge, as well as the start-up and shutdown status of thermal power units, flexible load dispatch status and energy storage charging and discharging status determined before the day.

3. The method for coordinated scheduling of power transmission and distribution networks based on heterogeneous resource partitioning and clustering as described in claim 1, characterized in that, In step S5, The day-ahead and intraday multi-timescale coordinated dispatch strategy for transmission and distribution networks includes day-ahead 24-hour load dispatch and intraday 4-hour load dispatch; specifically: The day-ahead 24-hour load dispatch includes: executing once every 24 hours, dividing into 24 time periods to plan the power of each load agent for the next day. The dispatch targets loads with slower response times and longer advance notice times. The day-ahead dispatch determines the dispatch plan for flexible loads (electrolytic aluminum and silicon carbide) that can be transferred within the next 24 hours in the transmission network, and the dispatch plan for flexible loads (thermal storage electric boilers and electric vehicles) that can be transferred within the next 24 hours in the distribution network. It should also determine the start-up and shutdown of conventional generating units in the transmission and distribution networks and formulate output plans for conventional generating units to guide their power generation. 4-hour load dispatching: Executed every 4 hours with a resolution of 15 minutes. Based on the effect of day-ahead load dispatching, the output results of conventional units during the dispatching period are corrected based on the latest new energy output and load forecast data. At the same time, the dispatching plan for flexible loads such as ferroalloy loads that can be reduced in the transmission network in the next 4 hours and the dispatching plan for flexible loads such as air conditioning loads that can be reduced in the distribution network in the next 4 hours are determined. The day-ahead output plan of conventional units is rolled out and optimized. The solution to the current scheduling plan model includes: using the objective cascading method to transform the model into a multi-level, multi-agent coordination optimization problem; The objective function of the power transmission network has recently been modified as follows: The objective function for the distribution network has recently been modified as follows: In the formula, ν k,t and ω k,t For algorithm multipliers; and For coupling variables obtained from adjacent regions, the superscript "-" indicates that the term is a known quantity; By setting the penalty function, the coupling variables are made to be as close as possible to the boundary power values ​​transmitted from adjacent regions during the calculation process, and finally achieve consistency. The solution to the intraday scheduling model includes: The objective function for the intraday power transmission network is modified as follows: The objective function for the distribution network has recently been modified as follows: In the formula, ν k,t and ω k,t The multiplier is the same as the multiplier parameter of the previous algorithm.

Images