Power market flexibility resource configuration method, device and equipment and storage medium

Abstract

The invention provides a power market flexibility resource configuration method and device, and the method comprises the steps: obtaining the real-time measurement, weather forecast and equipment state data of a power grid node, and extracting the spatial-temporal characteristics of a power grid topology; calculating new energy output and equipment availability rate based on meteorological and equipment data; generating a dynamic adjustment margin matrix in combination with the spatial-temporal characteristics, the new energy output and the equipment availability rate; predicting a quantile map of the short-term flexibility demand; global optimization is decomposed into regional sub-problems and solved, power adjusting instructions of the power generation side and the load side are generated, and resource dynamic configuration is achieved. The method improves the evaluation precision and response speed, and guarantees the stability of the power grid.

Description

Technical Field

[0001] The present application relates to the field of resource allocation, and in particular to a method, device, equipment and storage medium for allocating flexibility resources in an electricity market. Background Art

[0002] In the operation of modern power systems, traditional methods mainly rely on static evaluation based on historical load curves and centralized optimization models of the day-ahead market to evaluate and configure flexibility resources. This method is difficult to fully cover the complex and changeable actual operating scenarios after the access of new energy, resulting in the evaluation results being out of touch with actual needs. It is unable to respond to sudden changes in new energy output and abnormal load fluctuations in a timely manner, making it difficult to meet the power system's requirements for rapid response to flexibility resources, threatening the safety and stability of power grid operation. Summary of the Invention

[0003] The purpose of this application is to overcome the defects in the above-mentioned prior art and provide a method, device, equipment and storage medium for configuring flexibility resources in the electricity market.

[0004] This application provides a method for allocating flexibility resources in an electricity market, including:

[0005] Obtain real-time measurement data, weather forecast data, and equipment status data of power grid nodes;

[0006] Extracting spatiotemporal features associated with power grid topology based on the real-time measurement data;

[0007] Calculating new energy output and equipment availability based on the weather forecast data and equipment status data;

[0008] generating a dynamic adjustment margin matrix according to the spatiotemporal characteristics, the output of new energy sources, and the availability of equipment;

[0009] Based on the dynamic adjustment margin matrix, predicting a quantile map of short-term flexibility demand;

[0010] Decomposing the global optimization problem into regional sub-problems according to the quantile map;

[0011] Solve the regional sub-problems and generate regulating power instructions on the power generation side and the load side;

[0012] According to the power regulation instruction, flexible resource allocation in the electricity market is achieved.

[0013] Optionally, extracting spatiotemporal features associated with the power grid topology based on the real-time measurement data includes:

[0014] Capturing the topological relationships between power grid nodes through graph convolution operations;

[0015] The receptive field of the time dimension is expanded through the dilated convolution operation.

[0016] Optionally, predicting a quantile map of short-term flexibility demand based on the dynamic adjustment margin matrix includes:

[0017] Use the temporal attention mechanism to identify key historical fluctuation periods;

[0018] Adopting spatial attention mechanism to capture the coupling relationship between supply and demand among regions;

[0019] According to the historical key fluctuation period and the inter-regional supply and demand coupling relationship, a demand range containing at least 0.1, 0.5, and 0.9 percentiles is output.

[0020] Optionally, based on the quantile map, decomposing the global optimization problem into regional sub-problems includes:

[0021] Decompose the global problem into regional autonomous sub-problems, and each region independently optimizes local resources;

[0022] The cross-regional supply and demand differences are balanced through iterative updates of global coordination variables and Lagrange multipliers.

[0023] Optionally, a dynamic adjustment margin matrix is ​​generated according to the spatiotemporal characteristics, the output of new energy, and the availability of equipment, including:

[0024] The Sigmoid function is used to constrain the output range of the adjustable capacity of the grid node; the Softplus function is used to ensure the non-negativity of the grid node climbing rate; and the exponential rounding function is used to discretize the sustainable adjustment time of the grid node.

[0025] Optionally, a dynamic adjustment margin matrix is ​​generated according to the spatiotemporal characteristics, the output of new energy, and the availability of equipment, including:

[0026] Based on the spatiotemporal characteristics, the spatial coupling constraints of the adjustable capacity of the grid nodes are dynamically modified.

[0027] Optionally, global coordination variables and Lagrange multipliers are iteratively updated to balance cross-regional supply and demand differences, including:

[0028] The update rate of coordination variables is dynamically optimized based on the real-time changes in supply and demand differences between regions.

[0029] The present application also provides a power market flexibility resource configuration device, comprising:

[0030] Acquisition module, which obtains real-time measurement data, weather forecast data and equipment status data of power grid nodes;

[0031] An extraction module, which extracts spatiotemporal features associated with the power grid topology based on the real-time measurement data;

[0032] a calculation module for calculating the output of new energy and the availability of equipment based on the weather forecast data and the equipment status data;

[0033] A generation module generates a dynamic adjustment margin matrix according to the spatiotemporal characteristics, the output of new energy and the availability of equipment;

[0034] A prediction module, which predicts a quantile map of short-term flexibility demand based on the dynamic adjustment margin matrix;

[0035] a decomposition module, which decomposes the global optimization problem into regional sub-problems according to the quantile map;

[0036] An instruction module solves the regional sub-problem and generates power regulation instructions for the power generation side and the load side;

[0037] The configuration module realizes the flexibility resource configuration of the power market according to the power adjustment instruction.

[0038] The present application also provides a power market flexibility resource configuration device, comprising:

[0039] Memory;

[0040] A processor for storing a computer executable program for the above-mentioned method for configuring flexibility resources in an electricity market, the computer executable program being used to obtain real-time measurement data, weather forecast data and equipment status data of power grid nodes; extracting spatiotemporal characteristics associated with power grid topology based on the real-time measurement data; calculating new energy output and equipment availability based on the weather forecast data and equipment status data; generating a dynamic adjustment margin matrix based on the spatiotemporal characteristics, new energy output and equipment availability; predicting a quantile map of short-term flexibility demand based on the dynamic adjustment margin matrix; decomposing a global optimization problem into regional sub-problems based on the quantile map; solving the regional sub-problems to generate adjustment power instructions for the power generation side and the load side; and realizing flexibility resource configuration in the electricity market based on the adjustment power instructions.

[0041] The present application also provides a storage medium, comprising: a computer executable program stored therein, wherein the computer executable program is used to be called by a processor to execute the steps of the above-mentioned method for configuring flexibility resources in an electricity market.

[0042] The beneficial effects of this application are:

[0043] The present application provides a method for configuring flexibility resources in the electricity market, including: obtaining real-time measurement data, weather forecast data and equipment status data of power grid nodes; extracting spatiotemporal characteristics associated with the power grid topology based on the real-time measurement data; calculating the output of new energy and equipment availability based on the weather forecast data and equipment status data; generating a dynamic adjustment margin matrix based on the spatiotemporal characteristics, the output of new energy and the availability of equipment; predicting the quantile map of short-term flexibility demand based on the dynamic adjustment margin matrix; decomposing the global optimization problem into regional sub-problems based on the quantile map; solving the regional sub-problems to generate adjustment power instructions for the power generation side and the load side; and realizing flexibility resource configuration in the electricity market based on the adjustment power instructions. Based on real-time measurement, meteorological and equipment data, the present application extracts spatiotemporal characteristics and calculates adjustment margins, predicts flexibility demand, decomposes and optimizes the solution of adjustment instructions, realizes dynamic resource configuration, improves evaluation accuracy and response speed, and ensures grid stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the power market flexibility resource allocation process in this application;

[0045] Figure 2 It is a schematic diagram of the power market flexibility resource configuration device in this application. DETAILED DESCRIPTION

[0046] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it is understood that various forms of implementing the present disclosure are not limited by the embodiments set forth herein. Rather, the embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0047] Please refer to Figure 1 As shown, the present application provides a method for allocating flexibility resources in the power market, including:

[0048] S101. Acquire real-time measurement data, weather forecast data, and equipment status data of power grid nodes;

[0049] Collect real-time power system measurement data (the voltage of the grid node i at time t) The power flow of the line between the grid nodes i and j at time t ), weather forecast (wind speed forecast value Light intensity prediction ) and equipment status (the available capacity of generator g at time t The state of charge of the energy storage device e at time t ).

[0050] Then, the above data is standardized:

[0051]

[0052] Among them, μ X is the data mean, σ X is the standard deviation of the data.

[0053] S102, extracting spatiotemporal features associated with power grid topology based on the real-time measurement data;

[0054] The three-dimensional tensor Input the spatio-temporal convolutional network (STCN) to extract the spatio-temporal features of the three-dimensional tensor D.

[0055] Wherein, N is the number of the grid nodes in the power system, T_in is the input time window length, and F is the characteristic dimension of the measurement data of the grid node.

[0056] A. Spatial Convolution, specifically a graph convolution network (GCN) operation based on power grid topology; used to capture power grid topology correlations, improve cross-region coordination capabilities, and reduce command conflicts:

[0057]

[0058] Among them, H (l-1) Represents the output feature matrix of the l-1 layer, H 0 is a three-dimensional tensor D; is the adjacency matrix with self-loops added (A is the adjacency matrix based on the power grid topology, and I_N is the identity matrix); is the degree matrix, the diagonal elements

[0059] is the trainable weight matrix, b spa is the bias term, σ represents the nonlinear activation function, is the output spatial feature.

[0060] By calculation The adjacency matrix can be symmetrically normalized to avoid gradient explosion. The normalized adjacency matrix is ​​compared with the input feature matrix H (l-1) Multiplying together can aggregate the grid node information of neighbors.

[0061] B. Temporal Convolution: This is used to expand the temporal receptive field, enhance adaptability to new energy fluctuations, and reduce evaluation lag to minutes.

[0062]

[0063] Among them, d is the dilation rate of the void convolution, which is used to control the size of the receptive field; W temp is the temporal convolution kernel weight, DilatedConv() is the hole convolution function, is the output spatiotemporal joint feature.

[0064] The total number of layers in the spatiotemporal convolutional network is L, where l is the current layer index. Each layer performs spatial convolution → temporal convolution → residual connection in sequence. The residual connection is calculated as follows:

[0065] H (l) =f(H (l-1) , W (l) )+H (l-1)

[0066] Among them, f() represents the above-mentioned spatial convolution and temporal convolution operations.

[0067] Grid measurement data (such as voltage and power flow) contains both low-frequency global features (such as regional load levels) and high-frequency local features (such as renewable energy fluctuations). The residual mechanism allows the network to retain key information in the original input (such as the grid node voltage baseline), avoiding information loss caused by multi-layer nonlinear transformations.

[0068] S103. Calculating new energy output and equipment availability based on the weather forecast data and equipment status data;

[0069] After temporally aligning the meteorological data with the equipment status, the features of the two types of data are first extracted.

[0070] A. Calculate equivalent new energy output based on meteorological data:

[0071]

[0072] Where f is a function based on the wind turbine / photovoltaic characteristic curve, such as:

[0073]

[0074] Among them, v_cut-in, v_cut-out, and v_rated represent the cut-in wind speed, cut-out wind speed, and rated wind speed, respectively; ρ, C_p, and A represent the air density, wind energy utilization coefficient, and blade swept area, respectively.

[0075]

[0076] Among them, η, A pv 、T amb represent the photoelectric conversion efficiency, photovoltaic panel area and ambient temperature respectively.

[0077] B. Calculate equipment availability based on equipment status:

[0078]

[0079] Among them, C g,max is the maximum capacity of generator g.

[0080] Then map it through the fully connected layer to get the feature F meta , feature F meta Used to introduce external influencing factors (weather / equipment) to reduce the error in regulating capacity prediction:

[0081]

[0082] Among them, ReLU is the activation function, W meta is the weight matrix of the fully connected layer, b meta is the bias term of the fully connected layer.

[0083] The feature F meta and spatiotemporal characteristics Splicing to obtain fusion features:

[0084]

[0085] Among them, d embed Represents the embedding dimension. The original feature dimension 3 is transformed through the fully connected layer Upgrade to d embed Dimension (such as 64 dimensions).

[0086] Then calculate through Multilayer Perceptron (MLP):

[0087]

[0088] Then, the MLP output is converted into physical dimensions, and the output range is constrained by the activation function to ensure the feasibility of the optimization model and the instruction feasibility rate:

[0089]

[0090] Among them, C i,nax is the maximum adjustable capacity of the grid node i, R ibase is the benchmark climbing rate, and round() is the rounding function.

[0091] S104: Generate a dynamic adjustment margin matrix based on the spatiotemporal characteristics, new energy output, and equipment availability;

[0092] According to the equipment utilization, the scheduling margin matrix is ​​obtained:

[0093]

[0094] in, represents the adjustable capacity of the grid node i at time t (MW), represents the ramp power of the grid node i (MW / min), Represents the sustainable adjustment time (min). The sigmoid function is used to limit the output to [0,1] and then multiply it by the maximum capacity of the device C. i,max ; The softplus function is used to ensure that the climbing rate is non-negative, R i,base is the reference value (such as 10MW / min); round(exp()) is the exponential + rounding function, which is used to discretize the duration into integer minutes.

[0095] The physical topology and dynamic characteristics of the power grid are captured through spatiotemporal convolutional networks to improve the assessment accuracy of the regulation margin; the "assessment lag" problem is solved through minute-level updates (rather than the traditional hourly level); the meteorological fusion mechanism enhances the adaptability to new energy fluctuations and improves the new energy absorption capacity.

[0096] S105. Predicting a quantile map of short-term flexibility demand based on the dynamic adjustment margin matrix;

[0097] Based on historical transaction data, new energy output forecasts and load forecast data, the demand forecast model is used to output a 15-minute spatiotemporal distribution map of flexibility demand.

[0098] Construct an Attention-LSTM model as a demand forecasting model. The structure of the demand forecasting model is as follows:

[0099] A. LSTM layer: h t =LSTM(x t ,h t-1 )

[0100] Among them, input Contains historical transaction data New Energy Forecast and load forecasting h t is the hidden state of LSTM at time step t.

[0101] B. Dual Attention Mechanism: 1. Temporal Attention:

[0102]

[0103]

[0104] in, is the temporal attention weight at time step t, W Q 、W K is the learnable query and key matrix, d k is the temporal attention dimension (usually 64), and H = [h1,h2,...,hT] is the concatenation of all time-step hidden states. T is the temporal context vector.

[0105] The temporal attention mechanism calculates the importance weights of different time steps to identify key fluctuation periods (such as the 15-minute window when photovoltaic output drops sharply) and assigns higher weights to these periods. Specifically, the model considers the hidden state at the current moment (Query = W Q ·h t ) and the hidden state of all historical moments (Key=W K H) performs similarity calculations (i.e., the above formula). If the pattern of a certain historical period is highly correlated with the current forecast demand (e.g., a similar PV drop occurred in the past), its similarity score will be higher. The similarity is then converted into probability weights using the softmax function.

[0106] In scenarios with high penetration of new energy, the fluctuation of photovoltaic output caused by cloud movement will form obvious spikes in the time dimension. Temporal attention automatically captures this mutation pattern and calculates the context vector c. T Higher weights are given to these periods so that the model pays more attention to these key fluctuations when predicting.

[0107] Spatial Attention:

[0108]

[0109] in, is the spatial attention weight, h i is the eigenvector of the grid node i, H S is the characteristic matrix of all the grid nodes, W Q '、W K ' is the spatial attention-specific weight. c S is the spatial context vector, d' k is the spatial attention dimension

[0110] Spatial attention is used to capture inter-regional demand correlations (such as the supply and demand coupling between load centers and new energy bases).

[0111] Generate context vector c:

[0112] c=W c [c T ||c S ]+b c

[0113] Where || is the vector concatenation operation, W c is the fusion weight matrix, b c is the fusion bias term.

[0114] The context vector c is the result of comprehensive spatiotemporal attention and represents the global key information.

[0115] C. Output layer:

[0116]

[0117] Used to output the spatiotemporal distribution map of flexibility demand in the next 15 minutes, N is the number of nodes in the power grid, T is the length of the time window, and Q is the number of quantiles. In statistics, the number of quantiles refers to the number of values ​​at a specific quantile position after a set of data is arranged in order of size.

[0118] For example, the 0.1 / 0.5 (i.e., median) / 0.9 quantile values ​​of the predicted demand of the grid node 1 in the three time periods within the time window can be expressed as follows:

[0119] The grid node 1:

[0120] [Time period 9:00-9:05; [2.1MW, 3.3MW, 4.8MW], / / The demand is less than 2.1MW 10% of the time, less than 3.3MW 50% of the time, and less than 4.8MW 90% of the time.

[0121] Time period 9:05-9:10; [2.4MW, 3.6MW, 5.1MW];

[0122] Time period 9:10-9:15; [2.7MW, 4.0MW, 5.5MW]].

[0123] The spatiotemporal distribution map clarifies the differences in demand between regions and provides input for cross-regional optimization.

[0124] Training strategy, using quantile loss function to enhance prediction robustness:

[0125]

[0126] Among them, τ is the target quantile value (such as 0.1, 0.5, 0.9), y i,t is the actual demand of the grid node i at time t, is the quantile demand value predicted by the model, and Λ is the quantile set (usually 5-9 quantiles).

[0127] 0.1 percentile: predicted demand lower limit (conservative estimate, used to ensure minimum adjustment capacity);

[0128] 0.5 quantile: median forecast (benchmark scheduling scheme);

[0129] 0.9 percentile: predicted upper limit of demand (to guard against unexpected fluctuations in new energy).

[0130] The quantile loss function can improve the prediction reliability in extreme scenarios and improve resource utilization efficiency.

[0131] Uncertainty in power demand forecasting primarily stems from renewable energy fluctuations and load randomness. Temporal attention can identify similar historical fluctuation patterns (such as the periodic drop in PV output in the afternoon) and enhance feature learning during these periods. Spatial attention can capture inter-regional chain reactions (for example, when wind power in Area A plummets, Area B needs to increase reserve capacity), improving cross-regional coordination accuracy. By combining forecasts at different quantiles, a demand fluctuation range is formed (for example, [50MW, 80MW, 110MW] corresponding to quantiles 0.1, 0.5, and 0.9), providing a safety margin for dispatchers. Different penalties are then applied for overestimation and underestimation. For example, if the quantile τ = 0.9 (the upper limit prediction), the model prioritizes underestimation and insufficient reserve capacity.

[0132] By combining spatiotemporal attention with quantiles, the uncertainty problem brought by new energy to the power system can be solved.

[0133] S106. Decomposing the global optimization problem into regional sub-problems according to the quantile map;

[0134] Distributed optimization model construction.

[0135] Objective function:

[0136]

[0137] in, is the regulation cost coefficient of the grid node i, is the regulation power of the grid node i, λ is the regulation smoothness weight, To adjust the smoothing term, p t The power vector represents the power of all grid nodes. The smoothing term is the squared L2 norm of the power change vectors of all grid nodes at adjacent moments. Specifically, each component of the power change vector (the power change of each grid node) is squared and summed. The smoothing term is used to limit the power change amplitude between adjacent moments (e.g., a 5-minute interval) to avoid frequent device adjustments.

[0138] Constraints:

[0139] 1. Power balance:

[0140] 2. Adjustment capacity limitations:

[0141] 3. Climbing rate limit:

[0142] 4. Duration constraints:

[0143] Among them, T min is the minimum continuous action time (to prevent frequent start and stop); is the indicator function, when The function value is 1; otherwise it is 0; the constraint summation range is from the current time t to is the sustainable regulation time of the grid node i.

[0144] This constraint ensures that once the grid node i is activated In the subsequent At least T in minutes min This constraint avoids frequent startup and shutdown of equipment (such as gas turbines), thus extending the life of the equipment.

[0145] The distributed optimization model uses the alternating direction method of multipliers (ADMM) for parallel computing. The optimization objectives include:

[0146] Minimize the regulation cost function; meet the grid security constraints; balance the differences in resource supply and demand across regions.

[0147] S107, solving the regional sub-problem and generating regulating power instructions for the power generation side and the load side;

[0148] ADMM solves the optimization model.

[0149] Decompose the global problem into regional subproblems:

[0150] Lower level (regional autonomy):

[0151] Each region independently optimizes local resources (taking into account local costs and constraints).

[0152] Regional local issues:

[0153]

[0154] Among them, f k (p k ) is the objective function of region k (adjustment cost + local constraint), pk is the optimization variable of region k, z k is the global regional coordination variable, u k is the Lagrange multiplier, and ρ is the penalty factor.

[0155] Upper layer (global coordination layer):

[0156] Globally coordinated updates:

[0157]

[0158] Where m is the number of iterations and K is the number of regions.

[0159] Multiplier update:

[0160]

[0161] For example:

[0162] 1. Solving regional subproblems (taking region k as an example)

[0163] Objective function:

[0164]

[0165] Among them, N_k is the number of the grid nodes in area k, and is the regulated power obtained by the last iterative calculation of the grid node i.

[0166] Constraints:

[0167] 1. Power balance:

[0168]

[0169] is the total demand in area k;

[0170] 2. Adjustment capacity limitations:

[0171]

[0172] For the assessment of regulatory margin;

[0173] 3. Climbing rate limit:

[0174]

[0175] The solver can choose:

[0176] Quadratic programming (QP) or interior point method relies on the convexity of the problem; commercial solvers (such as Gurobi and CPLEX) are often used in actual engineering.

[0177] 2. Globally coordinated updates

[0178] Each region uploads p to the coordination center k +u k The coordination center calculates the average value z and broadcasts it to all areas. Multiplier u k It masks the actual power deviation and avoids directly exposing regional operation data.

[0179] 3. Convergence conditions

[0180] a. Residual conditions: is the preset tolerance (such as 1e-4).

[0181] b. The maximum number of iterations is 100 to prevent infinite loops.

[0182] S108. Implementing flexible resource allocation in the electricity market according to the power adjustment instruction.

[0183] The adjustment margin matrix and the demand spatiotemporal distribution map are input into a distributed optimization model, dynamic resource allocation instructions and an adjustment cost curve are solved in a rolling manner, and resource configuration instructions are issued to the power trading system.

[0184] Please refer to Figure 2 As shown, the present application also provides a power market flexibility resource configuration device, comprising:

[0185] Acquisition module 201, acquires real-time measurement data, weather forecast data and equipment status data of power grid nodes;

[0186] An extraction module 202 extracts spatiotemporal features associated with the power grid topology based on the real-time measurement data;

[0187] A calculation module 203 calculates the new energy output and equipment availability based on the weather forecast data and equipment status data;

[0188] A generation module 204 generates a dynamic adjustment margin matrix based on the spatiotemporal characteristics, the output of new energy sources, and the availability of equipment;

[0189] A prediction module 205 predicts a quantile map of short-term flexibility demand based on the dynamic adjustment margin matrix;

[0190] A decomposition module 206 decomposes the global optimization problem into regional sub-problems according to the quantile map;

[0191] The instruction module 207 solves the regional sub-problem and generates power regulation instructions for the power generation side and the load side;

[0192] The configuration module 208 implements the flexibility resource configuration of the electricity market according to the power adjustment instruction.

[0193] The present application also provides a power market flexibility resource configuration device, comprising:

[0194] Memory;

[0195] A processor for storing a computer executable program for the above-mentioned method for configuring flexibility resources in an electricity market, the computer executable program being used to obtain real-time measurement data, weather forecast data and equipment status data of power grid nodes; extracting spatiotemporal characteristics associated with power grid topology based on the real-time measurement data; calculating new energy output and equipment availability based on the weather forecast data and equipment status data; generating a dynamic adjustment margin matrix based on the spatiotemporal characteristics, new energy output and equipment availability; predicting a quantile map of short-term flexibility demand based on the dynamic adjustment margin matrix; decomposing a global optimization problem into regional sub-problems based on the quantile map; solving the regional sub-problems to generate adjustment power instructions for the power generation side and the load side; and realizing flexibility resource configuration in the electricity market based on the adjustment power instructions.

[0196] The present application also provides a storage medium, comprising: a computer executable program stored therein, wherein the computer executable program is used to be called by a processor to execute the steps of the above-mentioned method for configuring flexibility resources in an electricity market.

[0197] The above description of the embodiments is intended to facilitate understanding and application of this application by those skilled in the art. It will be readily apparent to those skilled in the art that various modifications to the above embodiments can be made, and the general principles described herein can be applied to other embodiments without requiring creative effort. Therefore, this application is not limited to the above embodiments. Any improvements or modifications made to this application by those skilled in the art based on the disclosure of this application should fall within the scope of protection of this application.

Claims

1. A method for allocating flexibility resources in the power market, characterized in that: include: Obtain real-time measurement data, weather forecast data, and equipment status data of power grid nodes; Extracting spatiotemporal features associated with power grid topology based on the real-time measurement data; Calculating new energy output and equipment availability based on the weather forecast data and equipment status data; generating a dynamic adjustment margin matrix according to the spatiotemporal characteristics, the output of new energy sources, and the availability of equipment; Based on the dynamic adjustment margin matrix, predicting a quantile map of short-term flexibility demand; Decomposing the global optimization problem into regional sub-problems according to the quantile map; Solve the regional sub-problems and generate regulating power instructions on the power generation side and the load side; According to the power regulation instruction, flexible resource allocation in the electricity market is achieved.

2. A method for allocating flexibility resources in the power market according to claim 1, characterized in that: Extracting spatiotemporal features associated with the power grid topology based on the real-time measurement data includes: Capturing the topological relationships between power grid nodes through graph convolution operations; The receptive field of the time dimension is expanded through the dilated convolution operation.

3. The method for configuring flexibility resources in the power market according to claim 1, characterized in that: Based on the dynamic adjustment margin matrix, a quantile map of short-term flexibility demand is predicted, including: Use the temporal attention mechanism to identify key historical fluctuation periods; Adopting spatial attention mechanism to capture the coupling relationship between supply and demand among regions; According to the historical key fluctuation period and the inter-regional supply and demand coupling relationship, a demand range containing at least 0.1, 0.5, and 0.9 percentiles is output.

4. A method for allocating flexibility resources in the power market according to claim 1, characterized in that: Based on the quantile map, the global optimization problem is decomposed into regional sub-problems, including: Decompose the global problem into regional autonomous sub-problems, and each region independently optimizes local resources; The cross-regional supply and demand differences are balanced through iterative updates of global coordination variables and Lagrange multipliers.

5. The method for allocating flexibility resources in the power market according to claim 1, characterized in that: According to the spatiotemporal characteristics, new energy output and equipment availability, a dynamic adjustment margin matrix is ​​generated, including: The Sigmoid function is used to constrain the output range of the adjustable capacity of the grid node; the Softplus function is used to ensure the non-negativity of the grid node climbing rate; and the exponential rounding function is used to discretize the sustainable adjustment time of the grid node.

6. A method for allocating flexibility resources in the power market according to claim 1, characterized in that: According to the spatiotemporal characteristics, new energy output and equipment availability, a dynamic adjustment margin matrix is ​​generated, including: Based on the spatiotemporal characteristics, the spatial coupling constraints of the adjustable capacity of the grid nodes are dynamically modified.

7. A method for allocating flexibility resources in the power market according to claim 4, characterized in that: Balance cross-regional supply and demand differences through iterative updates of global coordination variables and Lagrange multipliers, including: The update rate of coordination variables is dynamically optimized based on the real-time changes in supply and demand differences between regions.

8. A power market flexibility resource allocation device, characterized in that: include: Acquisition module, which obtains real-time measurement data, weather forecast data and equipment status data of power grid nodes; An extraction module, which extracts spatiotemporal features associated with the power grid topology based on the real-time measurement data; a calculation module for calculating the output of new energy and the availability of equipment based on the weather forecast data and the equipment status data; A generation module generates a dynamic adjustment margin matrix according to the spatiotemporal characteristics, the output of new energy and the availability of equipment; A prediction module, which predicts a quantile map of short-term flexibility demand based on the dynamic adjustment margin matrix; a decomposition module, which decomposes the global optimization problem into regional sub-problems according to the quantile map; An instruction module solves the regional sub-problem and generates power regulation instructions for the power generation side and the load side; The configuration module realizes the flexibility resource configuration of the power market according to the power adjustment instruction.

9. A power market flexibility resource configuration device, characterized in that: include: Memory; A processor for storing a computer executable program for the method for configuring flexibility resources in a power market according to any one of claims 1 to 7, the computer executable program for acquiring real-time measurement data, weather forecast data, and device status data of power grid nodes; and extracting spatiotemporal characteristics associated with power grid topology based on the real-time measurement data; Based on the weather forecast data and equipment status data, the new energy output and equipment availability are calculated; based on the spatiotemporal characteristics, the new energy output and the equipment availability, a dynamic adjustment margin matrix is ​​generated; based on the dynamic adjustment margin matrix, a quantile map of short-term flexibility demand is predicted; based on the quantile map, the global optimization problem is decomposed into regional sub-problems; the regional sub-problems are solved to generate adjustment power instructions for the power generation side and the load side; based on the adjustment power instructions, the flexibility resource allocation of the power market is realized.

10. A storage medium, characterized in that: include: A computer executable program is stored, which is used to be called by a processor to execute the steps of a method for configuring flexibility resources in an electricity market as described in any one of claims 1 to 7.

Images