Metacosmic-based power dispatching model construction and data interaction method and system, and storage medium

By adopting the meta-universe-based power scheduling model construction and data interaction method in the new power system, the problem of grid scheduling is difficult to cope with external environmental factors and insufficient cross-regional interconnection is achieved, and a more efficient power system scheduling and intelligence level is achieved.

CN119578096APending Publication Date: 2025-03-07STATE GRID TIANJIN ELECTRIC POWER COMPANY +2
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411729014.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-28
Publication Date
2025-03-07

AI Technical Summary

Technical Problem

The existing technology fails to effectively consider the external environment modeling of the new power system, which makes it difficult for power grid scheduling to cope with the influence of external factors such as weather, and fails to make full use of the interconnection and information sharing of cross-regional power grids, resulting in limited flexibility and efficiency of the scheduling system.

Method used

The power scheduling model based on the metaverse is used to construct and interact with data. By establishing a physical entity model of the power grid, a meteorological environment model, a topological abstract entity model and a fault oscillation model, and combining it into a power scheduling model based on the metaverse, the correlation analysis of the model and physical entities and historical data is realized, and finally data interaction with the data center.

Benefits of technology

It improves the intelligence level and scheduling efficiency of the power system, enhances the power grid's adaptability to the external environment, realizes the interconnection and information interaction of cross-regional power grids, and improves the planning, operation and maintenance and disaster response capabilities of the power system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119578096A_ABST
    Figure CN119578096A_ABST
Patent Text Reader

Abstract

The invention relates to a power dispatching model construction and data interaction method and system based on meta universe, and a storage medium. The method comprises the following steps: step 1, establishing a power grid physical entity model; 2, establishing a meteorological environment model; 3, constructing a topological abstract entity model; 4, establishing a fault oscillation model; step 5, performing entity association analysis on the model established in the above step; and step 6, performing interaction based on an entity association analysis result. According to the invention, problems and challenges faced by power grid operation and scheduling management under the background of a novel power system can be solved, the intelligent level and scheduling efficiency of the power system are improved, and the actual demand of a scheduling support mode is remodeled.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of power system dispatching operation and metaverse technology, and relates to a method and system for power dispatching model construction and data interaction, and in particular to a method and system for power dispatching model construction and data interaction based on metaverse. Background Art

[0002] With the rapid development of energy transformation, the current power system is facing the challenges of global energy structure transformation, environmental protection needs and sustainable development. The rapid development of new energy and the construction of new power systems have brought about major changes in power supply composition, grid form, load characteristics, technical foundation, and operating characteristics. In addition, due to its typical interdisciplinary, cross-cutting and deep integration characteristics, it is necessary to organically integrate meteorology, operations research, chaos theory, geological science, artificial intelligence, big data and other disciplines, which greatly increases the difficulty of system cognition and brings great challenges to dispatching in grid operation control, risk identification, and fault emergency defense. Therefore, it is of great significance to conduct in-depth research on power dispatching and data interaction under new power systems.

[0003] In recent years, scholars from various countries have conducted a series of studies on the dispatch of new power systems. However, existing studies have not considered modeling the external environment of new power systems, making it difficult for power grid dispatch to cope with the impact of external factors such as weather, posing a serious threat to power grid security; they have not considered modeling abstract entities such as virtual power plants, and have not considered the interconnectivity and information sharing between cross-regional power grids, which restricts the flexibility and efficiency of the dispatch system. Summary of the invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and propose a method and system for constructing a power dispatching model and interacting with data based on the metaverse, which can solve the problems and challenges faced by power grid operation and dispatching management under the background of new power systems, improve the intelligence level and dispatching efficiency of power systems, and reshape the actual needs of dispatching support models.

[0005] The present invention solves the practical problem by adopting the following technical solutions:

[0006] A method for constructing a power dispatching model and interacting with data based on the metaverse includes the following steps:

[0007] Establish a physical model of the power grid based on the physical characteristics of each component in the power system and the relationship between input and output;

[0008] Establish a meteorological environment model based on the meteorological data of the environment in which the power system is located;

[0009] Construct a topological abstract entity model based on the input-output relationship of one or more controllable resources in adjustable loads, energy storage, microgrids, and distributed power sources in different spaces within the region;

[0010] Establish a fault oscillation model based on the physical mechanisms of faults and oscillations in the power system;

[0011] The above-mentioned power grid physical entity model, meteorological environment model, topological abstract entity model and fault oscillation model are combined to establish a power dispatch model based on the metaverse;

[0012] The established Metaverse-based power dispatch model is physically associated with physical entities and historical data to obtain the model-entity-data association analysis results;

[0013] The entity association analysis results of the model-entity-data are exchanged with the data center.

[0014] Furthermore, the physical entity model of the power grid includes:

[0015] (1) Gas turbine power generation model:

[0016] Q GG =α GG +β GG +γ GG P GG 2

[0017]

[0018] Among them, Q GG is the gas turbine output, where α GG With β GG is a linear constant, γ GG is the coefficient of the quadratic term, Γ GG is the amount of natural gas consumed, L HV The calorific value of natural gas

[0019] (2) Thermal power unit model:

[0020] Output constraints of thermal power units:

[0021]

[0022] in, is the lower limit of thermal power unit output, is the upper limit of thermal power unit output, is the set of thermal power units, u i,t is the state parameter of the thermal power unit;

[0023] Unit climbing constraints:

[0024]

[0025] Among them, SU / SD is the ramp constraint when the thermal power unit is started / stopped, and RU / RD is the power of the thermal power unit climbing up / down;

[0026] Minimum start and stop time constraints:

[0027]

[0028] Among them, MU is the minimum continuous on time, and MD is the minimum continuous down time

[0029] (3) Cogeneration unit model

[0030] Run the model as follows:

[0031]

[0032] 0≤δ n,CHP,t ≤1,n∈(1,2…,N k ),t∈T

[0033] Among them, δ n,CHP,t is the electro-thermal coefficient, P t CHP and They represent the active power output and thermal output of the CHP unit at time t, respectively. k is the number of extreme points in the operating area; when the CHP unit is used as a power source, the relationship between the output reactive power and active power is:

[0034]

[0035] in, is the angle between active power and apparent power;

[0036] (4) Transmission network model, whose constraints include:

[0037]

[0038] PL br,t =B b θ b,t A

[0039] PL br,min ≤PL br,t ≤PL br,max

[0040] Among them, NTG is the system generator set, NTD is the system load node set, For the active power output of the generator set,

[0041] is the node active load; B b is the branch susceptance, θ b,t is the node phase angle at time t; PL br,t is the line L at time t br The power transmitted;

[0042] (5) AC distribution network model:

[0043]

[0044] (6) Photovoltaic power generation model:

[0045]

[0046] Among them, in the formula, and Respectively represent the active and reactive power output of photovoltaic, and Respectively represent the minimum and maximum values ​​of photovoltaic active power output, and Respectively represent the minimum and maximum values ​​of PV reactive power output;

[0047] (7) Power generation cost model:

[0048] The cost of electricity generation from non-CHG units:

[0049]

[0050] Wind power generation cost:

[0051]

[0052] (8) Load model

[0053] The input data of the load model include the induction motor parameters and their proportion in PL, the static load component in P static and Q static The proportion of static load power factor η pfac , the equivalent impedance of the distribution network obtained by taking the equivalent branch initial load capacity as the reference power, and the equivalent active power P output by the generator G and reactive power Q G , the unit parameters and excitation speed control system parameters given by the generator card;

[0054] (9) Substation 3D model

[0055] Moreover, the specific steps of establishing the meteorological environment model according to the meteorological data of the environment in which the power system is located include:

[0056] (1) Meteorological data collection and processing:

[0057] The ordered weighted average operator method is used to preprocess the regional real-time meteorological parameter data. The detailed process is as follows:

[0058] Let represent any real-time meteorological parameter data, and the expression formula is as follows:

[0059]

[0060] Among them, l represents the expected value, Expressed as standard deviation;

[0061] For any real-time meteorological parameter data x 0 To convert, the expression formula is as follows:

[0062]

[0063] Wherein, x represents standard real-time meteorological parameter data; x' represents converted real-time meteorological parameter data;

[0064] make Represents the real-time meteorological parameter data interval, which is divided into k equal parts. The calculation formula of real-time meteorological parameter data fusion support is as follows:

[0065]

[0066] Among them, s(x) represents the support of real-time meteorological parameter data fusion: The i-th equal part of the interval represents the calculation of the weight vector of the ordered weighted average operator of the fusion support of the real-time meteorological parameter data based on the formula result;

[0067] Let A represent the real-time meteorological parameter data fusion support decision support set, R n Represents n-dimensional space, mapping A to R n The real-time meteorological parameter data fusion support degree decision support degree set weighted vector can be expressed as w = (w 1 ,w 2 ,...w n ,) indicates that the n-dimensional ordered weighted average operator A=(a 1 ,a 2 ,...a n ,)The expression formula is as follows:

[0068]

[0069] Based on the above results, the real-time meteorological parameter data is fused and processed. According to the converted support and the weight vector of the ordered weighted average operator of the real-time meteorological parameter data, the data is fused to obtain the final decision result of the fusion decision and store it in the power terminal database.

[0070] (2) Use the improved Elman neural network to establish a disaster weather model:

[0071] The hidden layer expression formula of the Elman neural network model is as follows:

[0072]

[0073] Among them, S(k) represents the output result of the hidden layer; k represents the time, and f(·) represents the transfer function; They represent the weights from the input layer to the hidden layer and the hidden layer feedback respectively; B1 represents the input layer offset value; S(k-1) represents the hidden layer feedback value at the k-1th moment;

[0074] At time k, the comprehensive meteorological forecast result output by the Elman neural network model is expressed as follows:

[0075]

[0076] Among them, y(k) represents the comprehensive meteorological forecast result output by the Elman neural network model. Represents the weight from the hidden layer to the output layer: B2 represents the offset value of the hidden layer;

[0077] The Elman neural network model adopts the principle of gradient descent error back propagation. Assuming that the time interval is [0, T), the overall approximation error Q of the Elman neural network model indicator function in this interval can be expressed as follows:

[0078]

[0079] Among them, y 0 represents the expected output at time k; T represents the total time, and k∈T;

[0080] In the Elman neural network model, the weights The correction method is the same as the forward neural network, but W3 can be corrected by the chain derivative rule, and its expression formula is as follows:

[0081]

[0082] In the above formula, represents the corrected value; f'(·) represents the derivative of the transfer function. After learning the Elman neural network model through the above steps, the comprehensive meteorological forecast results can be obtained;

[0083] (3) A multi-source variable photovoltaic and wind power prediction model based on multi-source heterogeneous meteorological data is constructed to build a meteorological site model:

[0084] SpatioTemporal LSTM (STLSTM) is used to extract the spatiotemporal features of satellite cloud images and ERA5 meteorological data; the formula of the STLSTM network is:

[0085]

[0086] Among them, x t is the input at time t, which is satellite cloud images and ERA5 meteorological data in the model; t is the input gate, f t For the forget gate, t is the output gate; W and b are the weight and bias corresponding to each gate, s is the sigmoid activation function, * is the convolution operation, Θ is the Hadamard product; the output of STLSTM is the spatiotemporal feature h, which reflects the overall movement trend of the cloud layer;

[0087] A nonlinear mapping from 3D meteorological field to satellite cloud image is constructed according to the spatiotemporal characteristics, and the optimization goal of the conditional diffusion model is

[0088]

[0089] Among them, h t , is the meteorological spatiotemporal characteristics extracted by STLSTM from satellite cloud images and ERA5 meteorological elements at time t, n is the number of noise addition steps, is the image after n times of denoising, ∈ is the noise vector randomly sampled from the standard Gaussian distribution during the denoising process, ∈ θ is the predicted noise vector;

[0090] The prediction target is used as the training optimization target, and an improved diffusion model loss function is proposed to make the model converge better during the training process and the generated prediction image is closer to the prediction target; the diffusion process of the diffusion model is as follows:

[0091]

[0092] in, is the loss coefficient;

[0093] The improved diffusion model loss function is further obtained:

[0094]

[0095] Moreover, the specific steps of constructing a topological abstract entity model according to the input-output relationship of one or more controllable resources in adjustable loads, energy storage, microgrids, and distributed power sources in different spaces within the region include:

[0096] (1) Establish an interruptible resource model:

[0097]

[0098]

[0099] Where S DR is a collection of interruptible resources. is the usage status of interruptible resource i, is the startup variable of the resource. Both of the above variables are 0-1 variables. N is the maximum number of responses of the resource. is the actual power of resource i, is the initial power and response power of the resource, c i is the unit power response cost of interruptible resource i;

[0100] The mathematical model for establishing the transferable resources is:

[0101]

[0102] Where S TR is a collection of translatable resources. is the usage status of the portable resource i, is the startup variable of the resource. Represents whether the load is moved in during this period. The above three groups of variables are all 0-1 variables; is the actual power of resource i, is the initial power and response power of the resource;

[0103] The mathematical model for establishing a continuously adjustable resource is:

[0104]

[0105]

[0106] Where S CR is a continuously adjustable resource collection, is the usage status of resource i, is the startup variable of the resource. The above variables are all 0-1 variables; The ability to adjust downward and upward for adjustable resources;

[0107] The mathematical model for establishing graded adjustable resources is:

[0108]

[0109] Among them, S GR A collection of adjustable resources for grading. is the usage status of resource i, is the startup variable of the resource. is the variable indicating the load amount used in the kth level, and the above variables are all 0-1 variables; Indicates the power corresponding to the k-th load;

[0110] The mathematical model of energy storage resources is established as follows:

[0111]

[0112]

[0113] Where S ER A collection of adjustable resources for grading. is a variable indicating whether resource i is in the state of charging or discharging, is a variable representing the conversion of the resource into the charging and discharging state. The above variables are all 0-1 variables. N represents the maximum number of charge and discharge conversions of the resource per day. The maximum and minimum values ​​of the charging / discharging power of the resource. is the maximum and minimum value of the total energy of the resource;

[0114] (2) The power dispatching metaverse includes virtual power plants. The current mainstream VPP model objective function is:

[0115]

[0116] in, and The VPPs profit from the electricity market and internal electricity sales respectively; Compensation for when the actual output of the VPP does not meet the contract requirements; C dg is the comprehensive cost of the gas turbine, including operating cost, start-up and shutdown cost, standby cost, etc.; C wind is the sum of the wind power abandonment cost and operation and maintenance cost; C PV is the sum of the photovoltaic abandonment cost and operation and maintenance cost; C bat is the cost of battery charging and discharging; C DR is the DR cost in VPP;

[0117] At the same time, VPP optimization scheduling must also meet the following constraints;

[0118] Wind power and photovoltaic constraints:

[0119]

[0120] Where: and are the dispatchable powers of the i-th wind power and photovoltaic units at the moment respectively; and They are the predicted dispatchable power of the i-th wind power and photovoltaic unit at time t respectively.

[0121] Energy storage charging and discharging constraints:

[0122]

[0123] in, and are the capacity and minimum energy storage value of the i-th energy storage device respectively; and are the maximum charging power and maximum discharging power of the i-th energy storage device respectively;

[0124] DR constraints:

[0125]

[0126] in: is the interruption load power of the DR user at time t; is the interruption load power of all DR users in the VPP at the moment; N is the number of DR users in the VPP; is the upper limit of the interruption load power of the th DR user at the moment;

[0127] System power balance constraints:

[0128]

[0129] in: is the internal load power of VPP; It is the power provided by VPP to the energy market. A positive value represents power output to the energy market, and a negative value represents power input from the energy market.

[0130] Moreover, the specific steps of establishing the fault oscillation model according to the physical mechanism of faults and oscillations in the power system include:

[0131] (1) Establish a negatively damped oscillation model:

[0132] Negatively damped oscillation corresponds to the state of σ>0; where the capacitor voltage u c The expression can be written as

[0133] u c =e σt (C 1 cosω d t+C 2 sinω d t)

[0134] In addition, since the sinusoidal signal cosω d t, sinω d The envelopes of t are C 1 e σt and C 2 eσt , and the slope of the envelope gradually increases until it reaches infinity;

[0135] (2) Establishing a forced oscillation model

[0136] Here we can assume that the initial state of the system is zero, that is, U c0 =0,I 0 =0;

[0137] When the system inherent damping ratio is 0<ζ<1

[0138] Let β = ω / ω n , then the second-order LC oscillation formula can be transformed into:

[0139]

[0140] The inherent damping ratio of the system ζ=0

[0141] When the external disturbance frequency is not equal to the system frequency, it can be simplified to:

[0142]

[0143] Let ε=(ω n -ω) / 2, we can see

[0144]

[0145] When the external disturbance frequency is not equal to the system frequency, that is, β = 1, it can be simplified to:

[0146]

[0147] (3) Establishing a self-resonance mechanism model

[0148] The state equation of the dual-machine parallel system is:

[0149]

[0150] Among them, U1 and U2 are the DC side voltages of inverter 1 and inverter 2 respectively; L1 and L2 are the bridge arm side filter inductors of inverter 1 and inverter 2 respectively; R1 and R2 are the bridge arm side filter inductor resistance components of inverter 1 and inverter 2 respectively; C11 and C22 are the filter capacitors of inverter 1 and inverter 2 respectively; Lg1 and Lg2 are the grid side filter inductors of inverter 1 and inverter 2 respectively; i1 and i2 are the bridge arm side inductor currents of inverter 1 and inverter 2 respectively; ig1 and ig2 are the grid side inductor currents of inverter 1 and inverter 2 respectively; uc1 and uc2 are the capacitor voltages of inverter 1 and inverter 2 respectively; Ug is the grid voltage; Lg is the grid impedance; ig is the grid current;

[0151] It can be converted into the Jacobian matrix as follows:

[0152]

[0153] in,

[0154] a 1 =L gl L g / (L gl L g +L g2 L g +L g1 L g2 )

[0155] a 2 =L g2 L g / (L g1 L g +L g2 L g +L g1 L g2 )

[0156] a 3 =L g1 L g2 / (L g1 L g +L g2 L g +L g1 L g2 )

[0157] When there is interaction between the two machines, the Jacobian matrix of the system is:

[0158]

[0159] Moreover, the specific steps of physically associating the established Metaverse-based power dispatching model with physical entities and historical data to obtain the model-entity-data association analysis results include:

[0160] (1) The power dispatch model, physical entities and historical data based on the metaverse are represented by sets:

[0161] D={E,M,K}

[0162] Among them, E represents the set of metaverse power dispatch models, M represents the set of physical entities, and K represents the set of historical data;

[0163] (2) Perform data preprocessing, including cleaning, standardization and formatting. This step uses function f pre express:

[0164] D'=fpre (D)

[0165] Where D′ is the preprocessed data set;

[0166] (3) Constructing the power grid knowledge graph using function f graph express:

[0167] G=f graph (D′)

[0168] Among them, G = (V, E) is a graph, V is a node set, and E is an edge set;

[0169] (4) Entity linking function f link express:

[0170] L = f link (T,G)

[0171] Among them, T is the model set, L is the linked entity set;

[0172] (5) Function f for entity association analysis predict express:

[0173] R″=f predict (L,K)

[0174] Among them, K is the historical data set, and R″ is the model-entity-data relationship set after entity association.

[0175] Moreover, the specific expression for performing data interaction between the entity association analysis result of the model-entity-data and the data center is:

[0176] The entity association analysis results of the model-entity-data and the data center interact with each other through file transfer and sharing:

[0177] R″:S×A×N×C→{uploaded,downloaded}

[0178] Among them, R″ is the model-entity-data collection file, S is the data center, A is the data center address, N is the naming rule, and C is the content format.

[0179] A power dispatch model construction and data interaction system based on Metaverse, including:

[0180] The power grid physical entity model building module builds the power grid physical entity model according to the physical characteristics of each component in the power system and the relationship between input and output;

[0181] The meteorological environment model building module builds the meteorological environment model according to the meteorological data of the environment where the power system is located;

[0182] A topological abstract entity model construction module is used to construct a topological abstract entity model based on the input-output relationship of one or more controllable resources such as adjustable loads, energy storage, microgrids, and distributed power sources in different spaces within the region;

[0183] Fault oscillation model building module, which builds fault oscillation model according to the physical mechanism of faults and oscillations in the power system;

[0184] A power dispatch model establishment module based on the metaverse combines the above-mentioned power grid physical entity model, meteorological environment model, topological abstract entity model and fault oscillation model to establish a power dispatch model based on the metaverse;

[0185] The entity association module associates the established power dispatch model based on the metaverse with physical entities and historical data to obtain the association analysis results of the model-entity-data;

[0186] The data interaction module exchanges the entity association analysis results of the model-entity-data with the data center.

[0187] Furthermore, the data interaction module further includes:

[0188] The transmission and sharing module is used to describe the entity association analysis results of the model-entity-data and the data center through file transmission and sharing:

[0189] R″:S×A×N×C→{uploaded,downloaded}

[0190] Among them, R″ is the model-entity-data collection file, S is the data center, A is the data center address, N is the naming rule, and C is the content format.

[0191] A storage medium stores a computer program, which, when executed by a processor, implements a method for constructing a power dispatching model and interacting with data based on a metaverse.

[0192] Advantages and beneficial effects of the present invention:

[0193] The present invention proposes a method for constructing a power dispatching model and data interaction method using metaverse technology, constructing a physical entity model of the power grid, a meteorological environment model, a topological abstract entity model, a conceptual model of faults and oscillations, and a power dispatching metaverse data interaction method in the metaverse. The present invention can break through the difficulty of visual expression of information coupling such as power grid equipment, operating information, and external environment, and can also simulate dispatching in the power dispatching model in the metaverse. By simulating various operations and emergency response scenarios, it can effectively improve the planning, operation and maintenance, and disaster response capabilities of the power system, and realize interconnection and information interaction between cross-regional power grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0194] Figure 1 It is an equivalent circuit diagram of the comprehensive load model of the distribution network considered in the present invention;

[0195] Figure 2 It is a schematic diagram of the Elman neural network structure of the present invention;

[0196] Figure 3 is a classification diagram of the oscillation mechanism of the present invention;

[0197] Figure 4 It is a negative damping oscillation curve diagram without external disturbance of the present invention;

[0198] Figure 5 It is a curve diagram of equal-amplitude undamped beat frequency oscillation when the frequencies are unequal according to the present invention;

[0199] Figure 6 It is the undamped forced oscillation curve diagram of the present invention when the frequencies are equal;

[0200] Figure 7 This is a diagram of a dual-machine parallel system of the present invention;

[0201] Figure 8 It is the root locus diagram of the present invention. DETAILED DESCRIPTION

[0202] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings:

[0203] A method for constructing a power dispatching model and interacting with data based on the metaverse includes the following steps:

[0204] Step 1: Establish a physical model of the power grid based on the physical characteristics of each component in the power system and the relationship between input and output;

[0205] The establishment of the physical entity model of the power grid comprises:

[0206] (1) Gas turbine power generation model:

[0207] Q GG =α GG +β GG +γ GG P GG 2

[0208]

[0209] Among them, Q GG is the gas turbine output, where α GG With β GG is a linear constant, γ GG is the coefficient of the quadratic term, Γ GG is the amount of natural gas consumed, LHV The calorific value of natural gas

[0210] (2) Thermal power unit model:

[0211] Output constraints of thermal power units:

[0212]

[0213] in, is the lower limit of thermal power unit output, is the upper limit of thermal power unit output, is the set of thermal power units, u i,t It is the state parameter of thermal power unit.

[0214] Unit climbing constraints:

[0215]

[0216]

[0217] Among them, SU / SD is the climbing constraint when the thermal power unit is started / stopped, and RU / RD is the power of the thermal power unit climbing up / down.

[0218] Minimum start and stop time constraints:

[0219]

[0220] Among them, MU is the minimum continuous on time, and MD is the minimum continuous down time

[0221] (3) Cogeneration unit model

[0222] Since CHP units can provide both heat and electricity, they are usually driven by fossil fuels such as gas or coal. Their operating area is generally represented by an irregular trapezoid, and the operating model is as follows:

[0223]

[0224] 0≤δ n,CHP,t ≤1,n∈(1,2…,N k ),t∈T

[0225] Among them, δ n,CHP,t is the electro-thermal coefficient, P t CHP and They represent the active power output and thermal output of the CHP unit at time t, respectively. k is the number of extreme points in the operating area. When the CHP unit is used as a power source, the relationship between the output reactive power and active power is:

[0226]

[0227] in, It is the angle between active power and apparent power.

[0228] (4) Transmission network model, whose constraints include:

[0229]

[0230] PL br,t =B b θ b,t A

[0231] PL br,min ≤PL br,t ≤PL br,max

[0232] Among them, NTG is the system generator set, NTD is the system load node set, For the active power output of the generator set,

[0233] is the node active load; B b is the branch susceptance, θ b,t is the node phase angle at time t; PL br,t is the line L at time t br The power transmitted.

[0234] (5) AC distribution network model:

[0235]

[0236]

[0237] (6) Photovoltaic power generation model:

[0238]

[0239] Among them, in the formula, and Respectively represent the active and reactive power output of photovoltaic, and Respectively represent the minimum and maximum values ​​of photovoltaic active power output, and Respectively represent the minimum and maximum values ​​of PV reactive power output;

[0240] (7) Power generation cost model:

[0241] The cost of electricity generation from non-CHG units:

[0242]

[0243] Wind power generation cost:

[0244]

[0245] (8) Load model, such as Figure 1 As shown:

[0246] The input data of the load model include the induction motor parameters and their proportion in PL, the static load component in P static and Q static The proportion of static load power factor η pfac , the equivalent impedance (standard value) of the distribution network obtained by taking the equivalent branch initial load capacity as the reference power, and the active power P output by the equivalent generator G and reactive power Q G , unit parameters and excitation speed control system parameters given by the generator card.

[0247] SLM initialization steps in the program implementation process

[0248] The initial load power is obtained based on the load power obtained by flow calculation, the output of the equivalent generator and the loss of the equivalent impedance of the distribution network.

[0249] P L =P 0 +P G -P D0

[0250] Q L =Q 0 +Q G -Q D0

[0251] Where P D0 and Q D0 They are respectively the active loss and reactive loss of the equivalent impedance of the distribution network

[0252] The admittance matrix is ​​initialized according to the induction motor parameters, static load and generator parameters respectively.

[0253] Calculate the load bus voltage

[0254]

[0255] Where: is the system bus voltage; is the conjugate of the system bus voltage. The system gives a warning message.

[0256] According to some initialization parameters of the induction motor, Q IM0 .

[0257] Calculate the load static active power and load static reactive power

[0258] P static =P L -P IM0 =P Z0 +P I0 +P P0

[0259]

[0260] In the formula is the power factor angle of the static load,

[0261] Check the active power balance according to the load power balance formula (6) and calculate the compensation capacity Q of static reactive power compensation SC0

[0262] 0=P 0 -P D0 -(P IM0 +P Z0 +P I0 +P P0 -P G0 )

[0263] -Q SC0 =Q 0 -Q D0 -(Q IM0 +Q Z0 +Q I0 +Q P0 -Q G0 )

[0264] (9) Substation 3D model

[0265] First, detailed data collection of the substation is required, including the location, size, and operating status information of the equipment;

[0266] Use 3D modeling software to build a 3D model of the substation based on the collected data;

[0267] Import the model into virtual reality software to achieve 3D visualization of the substation;

[0268] Step 2: Establish a meteorological environment model based on meteorological data of the environment in which the power system is located;

[0269] The specific steps of establishing the meteorological environment model include:

[0270] (1) Meteorological data collection and processing:

[0271] The ordered weighted average operator method is used to preprocess the regional real-time meteorological parameter data. The detailed process is as follows:

[0272] Let represent any real-time meteorological parameter data, and the expression formula is as follows:

[0273]

[0274] Among them, l represents the expected value, Expressed as standard deviation.

[0275] For any real-time meteorological parameter data x 0 To convert, the expression formula is as follows:

[0276]

[0277] Wherein, x represents standard real-time meteorological parameter data; x' represents converted real-time meteorological parameter data.

[0278] make Represents the real-time meteorological parameter data interval, which is divided into k equal parts. The calculation formula of real-time meteorological parameter data fusion support is as follows:

[0279]

[0280] Among them, s(x) represents the support of real-time meteorological parameter data fusion: The i-th equal part of the interval represents the calculation of the weight vector of the ordered weighted average operator of the fusion support of real-time meteorological parameter data based on the formula result.

[0281] Let A represent the real-time meteorological parameter data fusion support decision support set, R n Represents n-dimensional space, mapping A to R n The real-time meteorological parameter data fusion support degree decision support degree set weighted vector can be expressed as w = (w 1 ,w 2 ,...w n ,) indicates that the n-dimensional ordered weighted average operator A=(a 1 ,a 2 ,...a n ,)The expression formula is as follows:

[0282]

[0283] Based on the above results, the real-time meteorological parameter data are fused and processed. According to the converted support and the ordered weighted average operator weight vector of the real-time meteorological parameter data, they are fused to obtain the final decision result of the fusion decision and store it in the power end database.

[0284] (2) Use the improved Elman neural network to establish a disaster weather model:

[0285] The hidden layer expression formula of the Elman neural network model is as follows:

[0286]

[0287] Among them, S(k) represents the output result of the hidden layer; k represents the time, and f(·) represents the transfer function; They represent the weights from the input layer to the hidden layer and the hidden layer feedback respectively; B1 represents the input layer offset value; S(k-1) represents the hidden layer feedback value at the k-1th moment.

[0288] At time k, the comprehensive meteorological forecast result output by the Elman neural network model is expressed as follows:

[0289]

[0290] Among them, y(k) represents the comprehensive meteorological forecast result output by the Elman neural network model. Represents the weight from the hidden layer to the output layer: B2 represents the offset value of the hidden layer.

[0291] The Elman neural network model adopts the principle of gradient descent error back propagation. Assuming that the time interval is [0, T), the overall approximation error Q of the Elman neural network model indicator function in this interval can be expressed as follows:

[0292]

[0293] Among them, y 0 represents the expected output at time k; T represents the total time, and k∈T.

[0294] like Figure 2 As shown, in the Elman neural network model, the weights The correction method is the same as the forward neural network, but W3 can be corrected by the chain derivative rule, and its expression formula is as follows:

[0295]

[0296] In the above formula, represents the corrected value; f'(·) represents the derivative of the transfer function. After learning the Elman neural network model through the above steps, the comprehensive meteorological forecast results can be obtained.

[0297] (3) A multi-source variable photovoltaic and wind power prediction model based on multi-source heterogeneous meteorological data is constructed to build a meteorological site model:

[0298] SpatioTemporal LSTM (STLSTM) is used to extract the spatiotemporal features of satellite cloud images and ERA5 meteorological data. The formula of the STLSTM network is:

[0299]

[0300] Among them, x t is the input at time t, which is satellite cloud images and ERA5 meteorological data in the model. t is the input gate, f t For the forget gate, t is the output gate. W and b are the weights and biases corresponding to each gate, s is the sigmoid activation function, * is the convolution operation, and Θ is the Hadamard product. The output of STLSTM is the spatiotemporal feature h, which reflects the overall motion trend of the cloud layer.

[0301] A nonlinear mapping from 3D meteorological field to satellite cloud image is constructed according to the spatiotemporal characteristics, and the optimization goal of the conditional diffusion model is

[0302]

[0303] Among them, h t , is the meteorological spatiotemporal characteristics extracted by STLSTM from satellite cloud images and ERA5 meteorological elements at time t, n is the number of noise addition steps, is the image after n times of denoising, ∈ is the noise vector randomly sampled from the standard Gaussian distribution during the denoising process, ∈ θ is the prediction noise vector.

[0304] The prediction target is used as the training optimization target, and an improved diffusion model loss function is proposed to make the model converge better during the training process and the generated prediction image is closer to the prediction target. The diffusion process of the diffusion model is as follows:

[0305]

[0306] in, is the loss coefficient.

[0307] The improved diffusion model loss function is further obtained:

[0308]

[0309] Step 3: Construct a topological abstract entity model based on the input-output relationship of one or more controllable resources in adjustable loads, energy storage, microgrids, and distributed power sources in different spaces within the region:

[0310] The specific steps of constructing the topological abstract entity model include:

[0311] (1) Modeling distributed resources is the basis for realizing the operation and control of the power dispatching metaverse. A standardized modeling method is proposed to model distributed resources into five general resource models: interruptible resources, continuous regulation resources, tiered regulation resources, transferable resources, and energy storage-like resources. These five types of resources can basically cover most distributed resources and realize the generalization of model interaction and data structure. The characteristics of each resource are as follows:

[0312] Modeling interruptible resources:

[0313]

[0314] Where S DR is a collection of interruptible resources. is the usage status of interruptible resource i, is the startup variable of the resource. Both of the above variables are 0-1 variables. N is the maximum number of responses of the resource. is the actual power of resource i, is the initial power and response power of the resource, c i is the unit power response cost of interruptible resource i.

[0315] The mathematical model for establishing the transferable resources is:

[0316]

[0317]

[0318] Where S TR is a collection of translatable resources. is the usage status of the portable resource i, is the startup variable of the resource. Represents whether the load is moved in during this period. The above three groups of variables are all 0-1 variables. is the actual power of resource i, are the initial power and response power of the resource.

[0319] The mathematical model for establishing a continuously adjustable resource is:

[0320]

[0321] Where S CR is a continuously adjustable resource collection, is the usage status of resource i, It is the startup variable of the resource. The above variables are all 0-1 variables. The downward and upward adjustment capabilities of adjustable resources.

[0322] The mathematical model for establishing graded adjustable resources is:

[0323]

[0324] Among them, S GR A collection of adjustable resources for grading. is the usage status of resource i, is the startup variable of the resource. is a variable representing the load amount used in the kth level, and the above variables are all 0-1 variables. Indicates the power corresponding to the k-th load.

[0325] The mathematical model of energy storage resources is established as follows:

[0326]

[0327]

[0328] Where S ER A collection of adjustable resources for grading. is a variable indicating whether resource i is in the state of charging or discharging, is a variable representing the conversion of the resource to the charging and discharging state. The above variables are all 0-1 variables. N represents the maximum number of charge and discharge conversions of the resource per day. The maximum and minimum values ​​of the charging / discharging power of the resource. is the maximum and minimum value of the total energy of the resource.

[0329] (2) The power dispatching metaverse includes virtual power plants. The current mainstream VPP model objective function is:

[0330]

[0331] in, and The VPPs profit from the electricity market and internal electricity sales respectively; Compensation for when the actual output of the VPP does not meet the contract requirements; C dg is the comprehensive cost of the gas turbine, including operating cost, start-up and shutdown cost, standby cost, etc.; C wind is the sum of the wind power abandonment cost and operation and maintenance cost; C PV is the sum of the photovoltaic abandonment cost and operation and maintenance cost; C bat is the cost of battery charging and discharging; C DR It is the DR cost in VPP.

[0332] At the same time, VPP optimization scheduling must also meet the following constraints.

[0333] Wind power and photovoltaic constraints:

[0334]

[0335] Where: and are the dispatchable powers of the i-th wind power and photovoltaic units at the moment respectively; and They are the predicted dispatchable power of the i-th wind power and photovoltaic unit at time t respectively.

[0336] Energy storage charging and discharging constraints:

[0337]

[0338] in, and are the capacity and minimum energy storage value of the i-th energy storage device respectively; and are the maximum charging power and maximum discharging power of the i-th energy storage device respectively.

[0339] DR constraints:

[0340]

[0341] in: is the interruption load power of the DR user at time t; is the interruption load power of all DR users in VPP at the moment; N is the number of DR users in VPP; is the upper limit of the interruption load power of the th DR user at the moment.

[0342] System power balance constraints:

[0343]

[0344] in: is the internal load power of VPP; It is the power provided by VPP to the energy market. A positive value represents power output to the energy market, and a negative value represents power input from the energy market.

[0345] Step 4: Establish a fault oscillation model based on the physical mechanism of faults and oscillations in the power system

[0346] like Figure 3 As shown, the specific steps of establishing the fault and oscillation model include:

[0347] Taking the LC oscillation circuit as an example, combined with the knowledge of second-order constant coefficient non-homogeneous differential equations, the expression for capacitor voltage is obtained:

[0348]

[0349] in,

[0350]

[0351] A series of oscillation mechanisms are derived from this formula.

[0352] (1) Establish a negatively damped oscillation model:

[0353] Negative damping oscillation corresponds to the state where σ>0 in the above equation (2). Therefore, regardless of whether external disturbances are considered, the capacitor voltage waveform will be at the initial value C 1 and C 2 The capacitor voltage u c The expression can be written as

[0354] u c =e σt (C 1 cosω d t+C 2 sinω d t)

[0355] In addition, since the sinusoidal signal cosω d t, sinω d The envelopes of t are C 1 e σt and C 2 e σt , and the slope of the envelope gradually increases until it reaches infinity, the oscillation waveform, such as Figure 4 shown.

[0356] (2) Establishing a forced oscillation model

[0357] Taking the LC second-order oscillation circuit as an example, since this part focuses on analyzing the impact of external periodic disturbances on the system, it can be assumed that the initial state of the system is zero, that is, U c0 =0,I 0 =0.

[0358] When the system inherent damping ratio is 0<ζ<1

[0359] Let β = ω / ω n , then the second-order LC oscillation formula can be transformed into:

[0360]

[0361] The inherent damping ratio of the system ζ=0

[0362] like Figure 5 As shown, when the external disturbance frequency and the system frequency are not equal, it can be simplified to:

[0363]

[0364] Let ε=(ω n -ω) / 2, we can see

[0365]

[0366] like Figure 6 As shown, when the external disturbance frequency is equal to the system frequency, that is, β = 1, it can be simplified to:

[0367]

[0368] like Figure 6 As shown, it is not difficult to find from the oscillation curve that the capacitor voltage u c A resonance with a continuously diverging amplitude occurs, and the amplitude amplification factor is much greater than the undamped beat frequency oscillation when the frequencies are not equal. This resonance will bring great harm to the stability of the power system.

[0369] (3) Establishing a self-resonance mechanism model

[0370] Take the example of a dual-machine system connected to the power grid to illustrate the strong and weak resonance phenomenon in internal resonance. Figure 7 As shown, it is assumed that the external disturbance of the two-machine system is zero, the initial state is exactly the same, and they are connected to the large power grid through output inductors Lg1 and Lg2. The equivalent impedance of the large power grid is Lg.

[0371] The state equation of the dual-machine parallel system is:

[0372]

[0373] Among them, U1 and U2 are the DC side voltages of inverter 1 and inverter 2 respectively; L1 and L2 are the bridge arm side filter inductors of inverter 1 and inverter 2 respectively; R1 and R2 are the bridge arm side filter inductor resistance components of inverter 1 and inverter 2 respectively; C11 and C22 are the filter capacitors of inverter 1 and inverter 2 respectively; Lg1 and Lg2 are the grid side filter inductors of inverter 1 and inverter 2 respectively; i1 and i2 are the bridge arm side inductor currents of inverter 1 and inverter 2 respectively; ig1 and ig2 are the grid side inductor currents of inverter 1 and inverter 2 respectively; uc1 and uc2 are the capacitor voltages of inverter 1 and inverter 2 respectively; Ug is the grid voltage; Lg is the grid impedance; ig is the grid current.

[0374] It can be converted into the Jacobian matrix as follows:

[0375]

[0376] in,

[0377] a 1 =L gl L g / (L gl Lg +L g2 L g +L g1 L g2 )

[0378] a 2 =L g2 L g / (L g1 L g +L g2 L g +L g1 L g2 )

[0379] a 3 =L g1 L g2 / (L g1 L g +L g2 L g +L g1 L g2 )

[0380] When there is interaction between the two machines, the Jacobian matrix of the system is:

[0381]

[0382] When the parameter η is changed, the corresponding characteristic root trajectory is as follows Figure 8 shown.

[0383] Step 5: Combine the above-mentioned power grid physical entity model, meteorological environment model, topological abstract entity model and fault oscillation model to establish a power dispatching model based on the metaverse;

[0384] Step 6: Associate the established Metaverse-based power dispatch model with physical entities and historical data to obtain the model-entity-data association analysis results;

[0385] The specific steps of performing entity association analysis on the model include:

[0386] (1) The collected power dispatch model, physical entities and historical data are expressed as a set:

[0387] D={E,M,K}

[0388] Among them, E represents the set of metaverse power dispatch models, M represents the set of physical entities, and K represents the set of historical data;

[0389] (2) Perform data preprocessing, including cleaning, standardization and formatting. This step uses function f pre express:

[0390] D'=f pre (D)

[0391] Where D′ is the preprocessed data set.

[0392] (3) Constructing the power grid knowledge graph using function f graph express:

[0393] G=f graph (D′)

[0394] Among them, G = (V, E) is the graph, V is the node set, and E is the edge set.

[0395] (4) Entity linking technology uses function f link express:

[0396] L = f link (T,G)

[0397] Among them, T is the model set and L is the linked entity set.

[0398] (6) Function f for entity association analysis predict express:

[0399] R″=f predict (L,K)

[0400] Among them, K is the historical data set, and R″ is the model-entity-data relationship set after entity association.

[0401] Step 7: Exchange the entity association analysis results of the model-entity-data with the data center

[0402] The method of exchanging data between the entity association analysis result and the data center is described by file transmission and sharing, and the specific expression is:

[0403] R″:S×A×N×C→{uploaded,downloaded}

[0404] Among them, R″ is the model-entity-data collection file, S is the data center, A is the data center address, N is the naming rule, and C is the content format.

[0405] The working principle of the present invention is:

[0406] The present invention proposes to use the metaverse as a carrier to study the modeling technology of the power dispatching metaverse, which is the key to solving the difficulty of modeling complex systems. First, the physical entities of the metaverse power grid are classified, derived and modeled. According to the needs of the physical entity objects of the power grid in each link of the whole process of "generation-transmission-transformation-distribution-use" in business use, the physical objects such as power grid containers (such as power plants, substations) and power grid equipment (such as transmission lines, transformers, circuit breakers, disconnectors, etc.) are modeled to form a full-factor decomposition model of the physical entity objects of the power grid in a digital environment. Secondly, according to the influencing factors of the external environment of the power grid, the related entities affected by the meteorological environment are derived into meteorological sites, disaster weather (typhoons, wildfires, ice cover, lightning, sandstorms), and regional resource endowment (irradiation, wind power density) models, and historical and predicted weather, disaster weather, wind and light resources and other meteorological environment information are obtained according to different modes such as sites and grids. Then, for the abstract entity virtual power plant in the metaverse, the distributed resources therein are established into five general resource models (interruptible resources, continuous regulation resources, graded regulation resources, shiftable resources and energy storage-like resources) through standardized modeling methods, and optimized scheduling is performed using advanced communication technology and control strategies. At the same time, in view of the oscillation stability problem of high-proportion new energy power systems, four types of oscillation models (negative damping oscillation model, forced oscillation model, self-resonance mechanism) are taken into account in the actual operation of the power system through the classification and modeling analysis of the oscillation mechanism. Finally, based on the digital power grid knowledge graph, a relationship association index is established, and the multi-round weighted entity linking technology of semantic similarity and convolutional neural network is used to realize the evolution of entity relationship reasoning, perform extended queries based on the input entity information, and efficiently feedback the associated models and mapping relationships to achieve modeling and data interaction.

[0407] It should be emphasized that the embodiments of the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific implementation modes. Any other implementation modes derived by those skilled in the art based on the technical solutions of the present invention also fall within the scope of protection of the present invention.

Claims

1. A method for constructing a power dispatching model and interacting with data based on the Metaverse, characterized in that: The following steps are involved: Establish a physical model of the power grid based on the physical characteristics of each component in the power system and the relationship between input and output; Establish a meteorological environment model based on the meteorological data of the environment in which the power system is located; Construct a topological abstract entity model based on the input-output relationship of one or more controllable resources in adjustable loads, energy storage, microgrids, and distributed power sources in different spaces within the region; Establish a fault oscillation model based on the physical mechanisms of faults and oscillations in the power system; The above-mentioned power grid physical entity model, meteorological environment model, topological abstract entity model and fault oscillation model are combined to establish a power dispatch model based on the metaverse; The established Metaverse-based power dispatch model is physically associated with physical entities and historical data to obtain the model-entity-data association analysis results; The entity association analysis results of the model-entity-data are exchanged with the data center.

2. According to the method for constructing a power dispatching model and interacting with data based on the Metaverse according to claim 1, it is characterized by: The power grid physical entity model includes: (1) Gas turbine power generation model: Q GG =a GG +b GG +g GG P GG 2 Among them, Q GG is the gas turbine output, where α GG With β GG is a linear constant, γ GG is the coefficient of the quadratic term, Γ GG is the amount of natural gas consumed, L HV The calorific value of natural gas (2) Thermal power unit model: Output constraints of thermal power units: in, is the lower limit of thermal power unit output, is the upper limit of thermal power unit output, is the set of thermal power units, u i,t is the state parameter of the thermal power unit; Unit climbing constraints: Among them, SU / SD is the ramp constraint when the thermal power unit is started / stopped, and RU / RD is the power of the thermal power unit climbing up / down; Minimum start and stop time constraints: Among them, MU is the minimum continuous on time, and MD is the minimum continuous down time (3) Cogeneration unit model Run the model as follows: 0≤δ n,CHP,t ≤1,n∈(1,2…,N k ),t∈T Among them, δ n,CHP,t is the electro-thermal coefficient, and They represent the active power output and thermal output of the CHP unit at time t, respectively. k is the number of extreme points in the operating area; when the CHP unit is used as a power source, the relationship between the output reactive power and active power is: in, is the angle between active power and apparent power; (4) Transmission network model, whose constraints include: PL br,t =B b θ b,t A PL br,min ≤PL br,t ≤PL br,max Among them, NTG is the system generator set, NTD is the system load node set, For the active power output of the generator set, is the node active load; B b is the branch susceptance, θ b,t is the node phase angle at time t; PL br,t is the line L at time t br The power transmitted; (5) AC distribution network model: (6) Photovoltaic power generation model: Among them, in the formula, and Respectively represent the active and reactive power output of photovoltaic, and Respectively represent the minimum and maximum values ​​of photovoltaic active power output, and Respectively represent the minimum and maximum values ​​of PV reactive power output; (7) Power generation cost model: The cost of electricity generation from non-CHG units: Wind power generation cost: (8) Load model The input data of the load model include the induction motor parameters and their proportion in PL, the static load component in P static and Q static The proportion of static load power factor η pfac , the equivalent impedance of the distribution network obtained by taking the equivalent branch initial load capacity as the reference power, and the equivalent active power P output by the generator G and reactive power Q G , the unit parameters and excitation speed control system parameters given by the generator card; (9) Three-dimensional model of substation.

3. The method for constructing a power dispatching model and interacting with data based on the metaverse according to claim 1, characterized in that: The specific steps of establishing the meteorological environment model according to the meteorological data of the environment in which the power system is located include: (1) Meteorological data collection and processing: The ordered weighted average operator method is used to preprocess the regional real-time meteorological parameter data. The detailed process is as follows: Let represent any real-time meteorological parameter data, and the expression formula is as follows: Among them, l represents the expected value, Expressed as standard deviation; To convert any real-time meteorological parameter data x0, the expression formula is as follows: Wherein, x represents standard real-time meteorological parameter data; x' represents converted real-time meteorological parameter data; make Represents the real-time meteorological parameter data interval, which is divided into k equal parts. The calculation formula of real-time meteorological parameter data fusion support is as follows: Among them, s(x) represents the support of real-time meteorological parameter data fusion: The i-th equal part of the interval represents the calculation of the weight vector of the ordered weighted average operator of the fusion support of the real-time meteorological parameter data based on the formula result; Let A represent the real-time meteorological parameter data fusion support decision support set, R n Represents n-dimensional space, mapping A to R n The real-time meteorological parameter data fusion support and decision support set weighted vector can be expressed as w = (w1, w2, ... w n ,) indicates that the n-dimensional ordered weighted average operator A=(a1,a2,...a n ,)The expression formula is as follows: Based on the above results, the real-time meteorological parameter data is fused and processed. According to the converted support and the weight vector of the ordered weighted average operator of the real-time meteorological parameter data, it is fused to obtain the final decision result of the fusion decision and store it in the power terminal database. (2) Use the improved Elman neural network to establish a disaster weather model: The hidden layer expression formula of the Elman neural network model is as follows: Among them, S(k) represents the output result of the hidden layer; k represents the time, and f(·) represents the transfer function; They represent the weights from the input layer to the hidden layer and the hidden layer feedback respectively; B1 represents the input layer offset value; S(k-1) represents the hidden layer feedback value at the k-1th moment; At time k, the comprehensive meteorological forecast result output by the Elman neural network model is expressed as follows: Among them, y(k) represents the comprehensive meteorological forecast result output by the Elman neural network model. Represents the weight from the hidden layer to the output layer: B2 represents the offset value of the hidden layer; The Elman neural network model adopts the principle of gradient descent error back propagation. Assuming that the time interval is [0, T), the overall approximation error Q of the Elman neural network model indicator function in this interval can be expressed as follows: Among them, y 0 represents the expected output at time k; T represents the total time, and k∈T; In the Elman neural network model, the weights The correction method is the same as the forward neural network, but W3 can be corrected by the chain derivative rule, and its expression formula is as follows: In the above formula, represents the corrected value; f'(·) represents the derivative of the transfer function. After learning the Elman neural network model through the above steps, the comprehensive meteorological forecast results can be obtained; (3) A multi-source variable photovoltaic and wind power prediction model based on multi-source heterogeneous meteorological data is constructed to build a meteorological site model: SpatioTemporal LSTM (STLSTM) is used to extract the spatiotemporal features of satellite cloud images and ERA5 meteorological data; the formula of the STLSTM network is: Among them, x t is the input at time t, which is satellite cloud images and ERA5 meteorological data in the model; t is the input gate, f t For the forget gate, t is the output gate; W and b are the weight and bias corresponding to each gate, s is the sigmoid activation function, * is the convolution operation, Θ is the Hadamard product; the output of STLSTM is the spatiotemporal feature h, which reflects the overall movement trend of the cloud layer; A nonlinear mapping from 3D meteorological field to satellite cloud image is constructed according to the spatiotemporal characteristics, and the optimization goal of the conditional diffusion model is Among them, h t , is the meteorological spatiotemporal characteristics extracted by STLSTM from satellite cloud images and ERA5 meteorological elements at time t, n is the number of noise addition steps, is the image after n times of denoising, ∈ is the noise vector randomly sampled from the standard Gaussian distribution during the denoising process, ∈ θ is the predicted noise vector; The prediction target is used as the training optimization target, and an improved diffusion model loss function is proposed to make the model converge better during the training process and the generated prediction image is closer to the prediction target; the diffusion process of the diffusion model is as follows: in, is the loss coefficient; The improved diffusion model loss function is further obtained:

4. The method for constructing a power dispatching model and interacting with data based on the Metaverse according to claim 1, characterized in that: The specific steps of constructing a topological abstract entity model according to the input-output relationship of one or more controllable resources in adjustable loads, energy storage, microgrids, and distributed power sources in different spaces in the region include: (1) Establish an interruptible resource model: Where S DR is a collection of interruptible resources. is the usage status of interruptible resource i, is the startup variable of the resource. Both of the above variables are 0-1 variables. N is the maximum number of responses of the resource. is the actual power of resource i, is the initial power and response power of the resource, c i is the unit power response cost of interruptible resource i; The mathematical model for establishing the transferable resources is: Where S TR is a collection of translatable resources. is the usage status of the portable resource i, is the startup variable of the resource. Represents whether the load is moved in during this period. The above three groups of variables are all 0-1 variables; is the actual power of resource i, is the initial power and response power of the resource; The mathematical model for establishing a continuously adjustable resource is: Where S CR is a continuously adjustable resource collection, is the usage status of resource i, is the startup variable of the resource. The above variables are all 0-1 variables; The ability to adjust downward and upward for adjustable resources; The mathematical model for establishing graded adjustable resources is: Among them, S GR A collection of adjustable resources for grading. is the usage status of resource i, is the startup variable of the resource. is the variable indicating the load amount used in the kth level, and the above variables are all 0-1 variables; Indicates the power corresponding to the k-th load; The mathematical model of energy storage resources is established as follows: Where S ER A collection of adjustable resources for grading. is a variable indicating whether resource i is in the state of charging or discharging, is a variable representing the conversion of the resource into the charging and discharging state. The above variables are all 0-1 variables. N represents the maximum number of charge and discharge conversions of the resource per day. The maximum and minimum values ​​of the charging / discharging power of the resource. is the maximum and minimum value of the total energy of the resource; (2) The power dispatching metaverse includes virtual power plants. The current mainstream VPP model objective function is: in, and The VPPs profit from the electricity market and internal electricity sales respectively; Compensation for when the actual output of the VPP does not meet the contract requirements; C dg is the comprehensive cost of the gas turbine, including operating cost, start-up and shutdown cost, standby cost, etc.; C wind is the sum of the wind power abandonment cost and operation and maintenance cost; C PV is the sum of the photovoltaic abandonment cost and operation and maintenance cost; C bat is the cost of battery charging and discharging; C DR is the DR cost in VPP; At the same time, VPP optimization scheduling must also meet the following constraints; Wind power and photovoltaic constraints: Where: and are the dispatchable powers of the i-th wind power and photovoltaic units at the moment respectively; and They are the predicted dispatchable power of the i-th wind power and photovoltaic unit at time t respectively. Energy storage charging and discharging constraints: in, and are the capacity and minimum energy storage value of the i-th energy storage device respectively; and are the maximum charging power and maximum discharging power of the i-th energy storage device respectively; DR constraints: in: is the interruption load power of the DR user at time t; is the interruption load power of all DR users in the VPP at the moment; N is the number of DR users in the VPP; is the upper limit of the interruption load power of the th DR user at the moment; System power balance constraints: in: is the internal load power of VPP; It is the power provided by VPP to the energy market. A positive value represents power output to the energy market, and a negative value represents power input from the energy market.

5. The method for constructing a power dispatching model and interacting with data based on the Metaverse according to claim 1, characterized in that: The specific steps of establishing the fault oscillation model according to the physical mechanism of faults and oscillations in the power system include: (1) Establish a negatively damped oscillation model: Negatively damped oscillation corresponds to the state of σ>0; where the capacitor voltage u c The expression can be written as u c =e σt (C1cosω d t+C2sinω d t) In addition, since the sinusoidal signal cosω d t, sinω d The envelopes of t are and And the slope of the envelope gradually increases until it reaches infinity; (2) Establishing a forced oscillation model Here we can assume that the initial state of the system is zero, that is, U c0 =0,I0=0; When the system inherent damping ratio is 0<ζ<1 Let β = ω / ω n , then the second-order LC oscillation formula can be transformed into: The inherent damping ratio of the system ζ=0 When the external disturbance frequency is not equal to the system frequency, it can be simplified to: Let ε=(ω n -ω) / 2, we can see When the external disturbance frequency is not equal to the system frequency, that is, β = 1, it can be simplified to: (3) Establishing a self-resonance mechanism model The state equation of the dual-machine parallel system is: Among them, U1 and U2 are the DC side voltages of inverter 1 and inverter 2 respectively; L1 and L2 are the bridge arm side filter inductors of inverter 1 and inverter 2 respectively; R1 and R2 are the bridge arm side filter inductor resistance components of inverter 1 and inverter 2 respectively; C11 and C22 are the filter capacitors of inverter 1 and inverter 2 respectively; Lg1 and Lg2 are the grid side filter inductors of inverter 1 and inverter 2 respectively; i1 and i2 are the bridge arm side inductor currents of inverter 1 and inverter 2 respectively; ig1 and ig2 are the grid side inductor currents of inverter 1 and inverter 2 respectively; uc1 and uc2 are the capacitor voltages of inverter 1 and inverter 2 respectively; Ug is the grid voltage; Lg is the grid impedance; ig is the grid current; It can be converted into the Jacobian matrix as follows: in, a1=L gl L g / (L gl L g +L g2 L g +L g1 L g2 ) a2=L g2 L g / (L g1 L g +L g2 L g +L g1 L g2 ) a3=L g1 L g2 / (L g1 L g +L g2 L g +L g1 L g2 ) When there is interaction between the two machines, the Jacobian matrix of the system is:

6. The method for constructing a power dispatching model and interacting with data based on the Metaverse according to claim 1, characterized in that: The specific steps of physically associating the established Metaverse-based power dispatching model with physical entities and historical data to obtain the model-entity-data association analysis results include: (1) The power dispatch model, physical entities and historical data based on the metaverse are represented by sets: D={E,M,K} Among them, E represents the set of metaverse power dispatch models, M represents the set of physical entities, and K represents the set of historical data; (2) Perform data preprocessing, including cleaning, standardization and formatting. This step uses function f pre express: D'=f pre (D) Where D′ is the preprocessed data set; (3) Constructing the power grid knowledge graph using function f graph express: G=f graph (D′) Among them, G = (V, E) is a graph, V is a node set, and E is an edge set; (4) Entity linking function f link express: L=f link (T,G) Among them, T is the model set, L is the linked entity set; (5) Function f for entity association analysis pre dict means: R″=f predict (L,K) Among them, K is the historical data set, and R″ is the model-entity-data relationship set after entity association.

7. The method for constructing a power dispatching model and interacting with data based on the Metaverse according to claim 1, characterized in that: The specific expression for performing data interaction between the entity association analysis result of the model-entity-data and the data center is: The entity association analysis results of the model-entity-data and the data center interact with each other through file transfer and sharing: R″:S×A×N×C→{uploaded,downloaded} Among them, R″ is the model-entity-data collection file, S is the data center, A is the data center address, N is the naming rule, and C is the content format.

8. A power dispatch model construction and data interaction system based on Metaverse, characterized by: include: The power grid physical entity model building module builds the power grid physical entity model according to the physical characteristics of each component in the power system and the relationship between input and output; The meteorological environment model building module builds the meteorological environment model according to the meteorological data of the environment where the power system is located; A topological abstract entity model construction module is used to construct a topological abstract entity model based on the input-output relationship of one or more controllable resources such as adjustable loads, energy storage, microgrids, and distributed power sources in different spaces within the region; Fault oscillation model building module, which builds fault oscillation model according to the physical mechanism of faults and oscillations in the power system; A power dispatch model establishment module based on the metaverse combines the above-mentioned power grid physical entity model, meteorological environment model, topological abstract entity model and fault oscillation model to establish a power dispatch model based on the metaverse; The entity association module associates the established power dispatch model based on the metaverse with physical entities and historical data to obtain the association analysis results of the model-entity-data; The data interaction module exchanges the entity association analysis results of the model-entity-data with the data center.

9. The power dispatch model construction and data interaction system based on Metaverse according to claim 8, characterized in that: The data interaction module further includes: The transmission and sharing module is used to describe the entity association analysis results of the model-entity-data and the data center through file transmission and sharing: R″:S×A×N×C→{uploaded,downloaded} Among them, R″ is the model-entity-data collection file, S is the data center, A is the data center address, N is the naming rule, and C is the content format.

10. A storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for constructing a power dispatching model and interacting with data based on the metaverse as claimed in any one of claims 1 to 7.

Citation Information

Cited By

  • Photovoltaic dust thickness real-time inversion method and system

    CN121520988A