Rapid evaluation method, system and equipment for power grid operation and storage medium

Abstract

The invention discloses a power grid operation rapid evaluation method, system and device and a storage medium, and the method comprises the steps: obtaining the probability distribution data of photovoltaic output and load demand in a power system, carrying out the preprocessing of the probability distribution data of photovoltaic output and load demand, and obtaining the dynamic characteristics of photovoltaic output and load demand; the optimal power flow is calculated by taking the minimum operation cost as the target and combining the power grid operation constraint, and the optimal power supply output condition and the load shedding amount under the optimal operation mode of the power grid are obtained; presetting a power grid operation state evaluation model, and training the preset power grid operation state evaluation model to obtain an optimal power grid operation state evaluation model; and evaluating the power grid operation state based on the optimal power grid operation state evaluation model. According to the method, the volatility and uncertainty caused by new energy high-proportion grid connection and flexible load can be effectively dealt with, the robustness of the model to a complex power grid environment is improved, and the operation state of the power grid is rapidly and accurately evaluated.

Description

Technical Field

[0001] This invention relates to the technical field of power system reliability assessment, and in particular to a method, system, device and storage medium for rapid assessment of power grid operation. Background Technology

[0002] The high proportion of renewable energy power generation connected to the grid and the addition of flexible loads have brought considerable volatility and uncertainty to grid operation. System operators must employ advanced uncertainty quantification to assess system risks, including generation planning (one day in advance, several hours in advance) and dispatching. Optimal operation decision-making algorithms, such as Optimal Power Flow (OPF) and Safe Unit Constraints (SCUC), play a crucial role in executing these tasks.

[0003] However, the methods described above require enormous computational resources and are not suitable for short-term real-time planning. They can only consider a limited number of scenarios, thus limiting the effectiveness of risk assessment in this manner. In contrast, high-resolution risk quantification can be achieved through sequential sampling-based methods. In this approach, many optimal solutions corresponding to various possible future scenarios are obtained, and the observations corresponding to these future scenarios and decisions are analyzed to quantify the risk. The need to solve a large number of OPF / SCUC problems presents significant computational challenges, as even solutions to these mixed-integer programming problems with a single future scenario are computationally challenging. Summary of the Invention

[0004] In view of the aforementioned existing problems, this invention is proposed. Therefore, this invention provides a method, system, device, and storage medium for rapid assessment of power grid operation to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] In a first aspect, embodiments of the present invention provide a method for rapid evaluation of power grid operation, comprising: acquiring probability distribution data of photovoltaic power output and load demand in a power system, and preprocessing the probability distribution data of photovoltaic power output and load demand to obtain dynamic characteristics of photovoltaic power output and load demand;

[0007] Optimal power flow is calculated by taking minimum operating cost as the objective and combining grid operation constraints, and the optimal power output and load reduction under the optimal grid operation mode are obtained.

[0008] A preset power grid operation status assessment model is trained based on the dynamic characteristics of photovoltaic power output and load demand, the optimal photovoltaic power output, and the load reduction to obtain the optimal power grid operation status assessment model.

[0009] The power grid operating status is evaluated based on the optimal power grid operating status evaluation model.

[0010] As a preferred embodiment of the rapid power grid operation assessment method of the present invention, the step of assessing the power grid operation status based on the optimal power grid operation status assessment model includes:

[0011] If the power grid operation status data value output by the optimal power grid operation status assessment model is greater than the preset threshold, then the power grid operation status is judged to be a fault state.

[0012] The probability of a fault state is quantified into the power grid operating state. For each power grid operating state that exceeds a preset threshold, its economic loss is calculated.

[0013] By generating spatiotemporally relevant scenarios, calculating the consequences of each scenario, and combining all risk indicators, the expected risk outcome corresponding to the power grid operation status data is calculated as a measure of system-level risk.

[0014] The beneficial effect of this preferred technical solution is that by introducing a spatiotemporal related scenario generation and risk assessment framework, it not only considers the intermittency and volatility of new energy output, but also enables a rapid and reliable assessment of the grid operation status.

[0015] As a preferred embodiment of the rapid power grid operation assessment method described in this invention, the optimal power grid operation assessment model is obtained by training a preset power grid operation status assessment model based on the dynamic characteristics of photovoltaic power output and load demand, the optimal power output, and the load shedding amount.

[0016] The dynamic characteristics of photovoltaic output and load demand are used as sample features of the preset power grid operation status assessment model, and the optimal photovoltaic output and load reduction are used as sample labels of the preset power grid operation status assessment model.

[0017] The sample features and sample labels are divided into a test set and a validation set according to a certain ratio. The test set is used to train the preset power grid operation status assessment model. In each round of training, the parameters are updated according to the loss function to obtain the optimal power grid operation status assessment model.

[0018] As a preferred embodiment of the rapid power grid operation assessment method described in this invention, the following steps are taken: In each training round, parameters are updated based on a loss function to obtain the optimal power grid operation state assessment model: The loss function is constructed from the mean squared error loss and a penalty term, and is expressed as:

[0019]

[0020] Among them, R b Represents a set of nodes. and xi ζ represents the basic true value and predicted value, respectively. i It is an auxiliary variable, x min and x max These represent the minimum and maximum allowed values ​​of the output variable, respectively.

[0021] As a preferred embodiment of the rapid power grid operation assessment method described in this invention, the preset power grid operation status assessment model includes: adding a self-loop to the symmetric normalized aggregation model in the graph convolutional neural network by changing the aggregation and update functions; and constructing the power grid operation status assessment model using five graph convolutional neural network layers and a fully connected readout layer.

[0022] Five graph convolutional neural network layers are used to update node representations, and the readout layer is used to output the output of traditional units, photovoltaic units, and node voltage amplitude and phase angle data of load reduction magnitude.

[0023] The aggregation model is represented as:

[0024]

[0025] A * =A+I N

[0026] Among them, A * It is the adjacency matrix of an undirected graph G plus self-connections, I N It is the identity matrix, D * It is A * The degree matrix, H (l) W is the activation unit matrix of the l-th layer. (l) It is the parameter matrix for each layer.

[0027] The beneficial effects of this preferred technical solution are as follows: by introducing a five-layer graph convolutional neural network model, it can efficiently process graph-structured data, extract key features, and output observations, significantly improving the timeliness and accuracy of the assessment; especially in complex and ever-changing large power grid environments, it can quickly respond to dynamic changes in network topology and achieve accurate assessment of power system reliability. Furthermore, by combining it with an optimal power flow algorithm for offline optimization, the accuracy and efficiency of the assessment are further improved.

[0028] As a preferred embodiment of the rapid power grid operation assessment method described in this invention, the optimal power flow is calculated by combining power grid operation constraints with the objective of minimizing operating cost, to obtain the optimal photovoltaic output and load shedding under the optimal power grid operation mode, including:

[0029] The photovoltaic power output calculation uses minimizing cost-effectiveness as the objective function:

[0030]

[0031] Where F is the objective function considering photovoltaic output, and P pv P Gi The outputs are photovoltaic power and traditional unit power, respectively. pv a is the output cost coefficient of the photovoltaic unit. i b i c i Let be the cost coefficient of the i-th traditional unit;

[0032] Power grid operation constraints include power balance constraints, maximum and minimum actual and reactive power output constraints of power sources, voltage phase angle constraints, and branch power flow constraints.

[0033] As a preferred embodiment of the rapid power grid operation assessment method described in this invention, the method includes: acquiring probability distribution data of photovoltaic power output and load demand in the power system, and preprocessing the probability distribution data of photovoltaic power output and load demand to obtain the dynamic characteristics of photovoltaic power output and load demand, including: modeling photovoltaic power output and load demand, and generating photovoltaic power output and load demand at different time steps based on the photovoltaic power output model and the load demand model.

[0034] A photovoltaic-load joint probability distribution model is constructed to obtain the joint probability density of photovoltaic and load. The photovoltaic-load joint probability distribution is sampled according to random sampling to obtain the initial photovoltaic output and load demand samples.

[0035] The initial photovoltaic power output and load demand samples are processed using the random walk method. The time series characteristics of photovoltaic power output and load demand are obtained by superimposing random disturbances from the state at the previous time step with the state at the current time step.

[0036] Secondly, the present invention provides a rapid power grid operation assessment system, comprising:

[0037] The data acquisition and preprocessing module is used to acquire probability distribution data of photovoltaic power output and load demand in the power system, and preprocess the probability distribution data of photovoltaic power output and load demand to obtain the dynamic characteristics of photovoltaic power output and load demand.

[0038] The optimal power flow calculation module is used to calculate the optimal power flow with the goal of minimizing operating costs and in combination with grid operation constraints, so as to obtain the optimal power output and load reduction under the optimal grid operation mode.

[0039] The model optimization module is used to preset the power grid operation status assessment model, and train the preset power grid operation status assessment model based on the dynamic characteristics of photovoltaic output and load demand, the optimal photovoltaic output situation and the load reduction amount, so as to obtain the optimal power grid operation status assessment model.

[0040] The evaluation module is used to evaluate the power grid operating status based on the optimal power grid operating status evaluation model.

[0041] Thirdly, the present invention provides an electronic device, comprising:

[0042] Memory and processor;

[0043] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the rapid power grid operation assessment method.

[0044] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the rapid power grid operation assessment method.

[0045] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention applies a combined GCN model to rapid power grid operation assessment, which can effectively cope with the volatility and uncertainty brought about by the high proportion of new energy grid connection and flexible loads, and improve the robustness of the model to complex power grid environments. By integrating the real-time advantages of the five-layer GCN model with the offline optimization characteristics of optimal power flow, a rapid and accurate assessment of the power grid operation status is achieved. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0047] Figure 1 This is a schematic flowchart illustrating a method, system, device, and storage medium for rapid assessment of power grid operation according to an embodiment of the present invention.

[0048] Figure 2 This is a schematic diagram of a standard IEEE 118-node system for a rapid power grid operation assessment method, system, device, and storage medium according to an embodiment of the present invention.

[0049] Figure 3 This is a flowchart illustrating a method, system, equipment, and storage medium for rapid power grid operation assessment according to an embodiment of the present invention.

[0050] Figure 4 This is a GCN model prediction result diagram of a power grid operation rapid assessment method, system, equipment and storage medium according to an embodiment of the present invention;

[0051] Figure 5 This is a comparison chart showing the convergence of a method, system, device, and storage medium for rapid evaluation of power grid operation according to an embodiment of the present invention. Detailed Implementation

[0052] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0053] Example 1, referring to Figure 1 As one embodiment of the present invention, this embodiment provides a method for rapid assessment of power grid operation, comprising:

[0054] S100: Obtain the probability distribution data of photovoltaic power output and load demand in the power system, and preprocess the probability distribution data of photovoltaic power output and load demand to obtain the dynamic characteristics of photovoltaic power output and load demand.

[0055] S200: Calculates the optimal power flow with the goal of minimizing operating costs and combining grid operation constraints, to obtain the optimal power output and load reduction under the optimal grid operation mode;

[0056] S300: Preset power grid operation status assessment model. The preset power grid operation status assessment model is trained based on the dynamic characteristics of photovoltaic output and load demand, the optimal photovoltaic output situation, and the load reduction amount to obtain the optimal power grid operation status assessment model.

[0057] S400: Based on the optimal power grid operation status assessment model, the power grid operation status is assessed.

[0058] It should be noted that optimal operating decision algorithms, such as Optimal Power Flow (OPF) and Safe Unit Constraints (SCUC), play a crucial role in performing these tasks. However, these methods require significant computational resources and are not suitable for short-term real-time planning; they can only consider a limited number of scenarios, resulting in limited risk assessment. In contrast, high-resolution risk quantification can be achieved through sequential sampling-based methods. In this approach, many optimal solutions corresponding to various possible future scenarios are obtained. Observations corresponding to these future scenarios and decisions are analyzed to quantify risk. The need to solve a large number of OPF / SCUC problems presents significant computational challenges, as even solutions to these mixed-integer programming problems with a single future scenario are computationally challenging. Advances in machine learning (ML) offer new opportunities to address this challenge. Among various machine learning model architectures, Graph Convolutional Neural Networks (GCNs) are particularly attractive for building proxy models for computational tasks in power grids, as they are capable of handling graph-structured data.

[0059] This application proposes a rapid power grid operation assessment method based on a combined GCN model, which integrates the real-time advantages of combined GCN with the offline optimization features of OPF. The five-layer GCN efficiently extracts features and outputs observations, significantly improving the timeliness and accuracy of the assessment. This method is particularly suitable for short-term planning in complex and ever-changing large power grid environments, enabling rapid response to dynamic changes in network topology and achieving accurate assessment of power system reliability. It provides a solid guarantee for the sustainable utilization of renewable energy in large power grids and the safe and stable operation of power systems.

[0060] In this embodiment of the application, step S100 involves obtaining probability distribution data of photovoltaic power output and load demand in the power system, and preprocessing the probability distribution data of photovoltaic power output and load demand to obtain the dynamic characteristics of photovoltaic power output and load demand, including: modeling photovoltaic power output and load demand, and generating photovoltaic power output and load demand at different time steps based on the photovoltaic power output model and the load demand model.

[0061] A photovoltaic-load joint probability distribution model is constructed to obtain the joint probability density of photovoltaic and load. The photovoltaic-load joint probability distribution is sampled according to random sampling to obtain the initial photovoltaic output and load demand samples.

[0062] The initial photovoltaic power output and load demand samples are processed using the random walk method. The time series characteristics of photovoltaic power output and load demand are obtained by superimposing random disturbances from the state at the previous time step with the state at the current time step.

[0063] Furthermore, based on the modeling of photovoltaic power generation and load demand, and combined with sequential Monte Carlo simulation, random initial photovoltaic output and load demand samples are generated. The main steps are as follows:

[0064] The photovoltaic power output model is expressed as follows:

[0065] P=rκη

[0066] Where P is the photovoltaic power generation, r is the irradiance, which follows a Beta distribution; к is the total area of ​​the photovoltaic module; and η is the photovoltaic power generation efficiency.

[0067] Photovoltaic power generation also follows a Beta distribution, with its probability distribution f p (P) is:

[0068]

[0069] Among them, P max Let Γ(·) be the maximum power of photovoltaic power generation, and let Γ(·) be the gama function; α and β are the shape parameters of the Beta distribution.

[0070] The calculation of the shape parameter of the Beta distribution is expressed as follows:

[0071]

[0072] Where, μ pv σ pv These represent the mean and standard deviation of the corresponding photovoltaic output sequence, respectively.

[0073] Photovoltaic reactive power output is represented as:

[0074] Q pv =P pv cosθ

[0075] Where θ is the power factor of the photovoltaic unit, which can be set to 0.85.

[0076] Furthermore, considering the photovoltaic-load joint probability distribution model of the load model, the load probability model following a normal distribution is expressed as:

[0077]

[0078] Where, μ D σ is the average value of the active power of the load. D Standard deviation;

[0079] Let x1,…,x be random variables representing all photovoltaic power output and load demand. n The marginal distribution functions are F1(x1), F2(x2), ..., F n (x n ):

[0080] F(x1,x2,…,x n )=C[F1(x1),…,Fn (x n )]

[0081] Where F(x1,x2,x3,…,x) n Let be the joint cumulative probability distribution function of the n-dimensional random variables, and let C[·] be the copula function.

[0082] In an optional embodiment, the copula function used is represented as follows:

[0083]

[0084] Among them, the values ​​of parameters a and b are in the range of (0,1); Let be the cumulative probability distributions of photovoltaic and load, respectively; the indicator function ∏[·] is 1 when it satisfies its own constraints, otherwise it is 0.

[0085] The joint probability density of photovoltaic and load is calculated by combining the copula function and the t-copula function; the joint probability distribution of photovoltaic and load is sampled according to random sampling to obtain the sample X0 of the initial random variable.

[0086] It should be noted that, considering the correlation and randomness of network random variables, the random walk method is used to process the network random variable X0 to obtain the processed sample X. The main steps are as follows:

[0087] The random walk approach introduces the time dependence of grid variables. The initial sample, X0, is generated through Monte Carlo simulation, reflecting only the independent temporal randomness of the variables. However, in actual grid operation, variables (such as load demand and photovoltaic output) have time inertia. Through the random walk model, the current state is generated by superimposing random disturbances from the previous time step. Specifically, a set of random initial variables x0 is randomly selected from X0, which is considered to be the initial operating state x0 of the system at time t. t Then x t+1 The state at time t can be calculated using the following formula:

[0088] x t+1 =x t +ε N ,ε N ∈N(0,1)

[0089] Where, ε N is a random number drawn from a standard normal distribution.

[0090] x t Converted to a standard normal distribution, it can be represented as:

[0091]

[0092] The random walk model described above simulates the dynamic process of variables fluctuating gradually over time, avoiding non-physical abrupt changes caused by independent sampling.

[0093] Then, Cholesky decomposition is used to transform uncorrelated variables into correlated variables. Spatiotemporal correlation is injected into the samples to ensure that the generated scene conforms to both the temporal evolution law and the spatial linkage characteristics between regions, represented as:

[0094] Where L is the lower triangular matrix in the Cholesky decomposition of the prediction covariance matrix C, and ′ denotes matrix transpose.

[0095] Given marginal probability W i The covariance matrix C, and the spatially correlated variables obtained from these marginal conditions are:

[0096]

[0097] in, For u t The i-th column is obtained from the cumulative density function of the standard normal distribution. The CDF is obtained, Φ Wi Let represent the CDF of the i-th edge.

[0098] Based on all generated initial photovoltaic power output and load demand samples X0(x t0 Generate all samples W after random walk processing. t =[w1 t w2 t …w Mt ], where M is the number of samples.

[0099] In this embodiment of the application, step S200 calculates the optimal power flow with the goal of minimizing operating cost and in conjunction with grid operation constraints, to obtain the optimal photovoltaic output and load shedding under the optimal grid operation mode, including:

[0100] The photovoltaic power output calculation uses minimizing cost-effectiveness as the objective function:

[0101]

[0102] Where F is the objective function considering photovoltaic output, and P pv P Gi The outputs are photovoltaic power and traditional unit power, respectively. pv a is the output cost coefficient of the photovoltaic unit. i b i c i Let be the cost coefficient of the i-th traditional unit;

[0103] Power grid operation constraints include power balance constraints, maximum and minimum actual and reactive power output constraints of power sources, voltage phase angle constraints, and branch power flow constraints.

[0104] Specifically, the power balance constraint is as follows:

[0105]

[0106] Among them, P D Q D These represent the total active and reactive power demand of the system load, respectively; P G Q G P represents the active and reactive power generation of a traditional generator unit. pv P represents the total active power output of photovoltaic power generation; L Q L These represent the active and reactive transmission losses of the line, respectively.

[0107] Furthermore, capacity constraints or inequality constraints, i.e., limits of all variables and functions, such as the upper and lower bounds of transmission lines, power generation output, stability, and safety limits, can be expressed as:

[0108] The maximum and minimum actual power output and reactive power output of the power source are expressed as follows:

[0109]

[0110] in, and These are the minimum and maximum power outputs of the power supply, respectively. and These represent the minimum and maximum power output of the photovoltaic power source, respectively; NG represents the number of conventional units in the system.

[0111] The voltage phase angle V and θ constraints are expressed as follows:

[0112]

[0113] Where, N b This represents the number of system nodes.

[0114] Branch flow constraints are represented as:

[0115]

[0116] Where, N L This represents the total number of branches in the network.

[0117] Furthermore, based on the constraints and objective function, the optimizer calculates the optimal power flow, obtaining the optimal power supply (including photovoltaic units and traditional existing units) output and load reduction under the optimal operating environment, which are then combined to form the model output sample labels. The specific implementation process is as follows: With the objective of minimizing the total economic cost including the output costs of photovoltaic and traditional units, the Gurobi mixed-integer linear programming solver is used to solve the constraint-laden optimization problem. The constraints are incorporated into the objective function using the Lagrange multiplier method, and the gradient descent method is used for iterative solving. The iteration terminates when the duality gap is less than a threshold, which is set to 10 in this application. -4 .

[0118] In this embodiment of the application, the preset power grid operation status assessment model in step S300 includes: adding a self-loop symmetric normalized aggregation model to the graph convolutional neural network by changing the aggregation and update functions; and constructing the power grid operation status assessment model using five graph convolutional neural network layers and a fully connected readout layer.

[0119] Five graph convolutional neural network layers are used to update node representations, and the readout layer is used to output the output of traditional units, photovoltaic units, and the voltage amplitude and phase angle data of nodes with load reduction magnitude; it can be used to evaluate the real-time operating status of the power grid (such as supply and demand balance, over-limit risk) and economics (such as power generation cost).

[0120] The aggregation model is represented as:

[0121]

[0122] A * =A+I N

[0123] Among them, A * It is the adjacency matrix of an undirected graph G plus self-connections, I N It is the identity matrix, D * It is A * The degree matrix, i.e. H (l) W is the activation unit matrix of the l-th layer. (l) It is the parameter matrix for each layer.

[0124] In this embodiment of the application, step S300, which trains a preset power grid operation status assessment model based on the dynamic characteristics of photovoltaic power output and load demand, the optimal power output, and the load shedding, to obtain the optimal power grid operation status assessment model, includes:

[0125] The dynamic characteristics of photovoltaic output and load demand are used as sample features of the preset power grid operation status assessment model, and the optimal photovoltaic output and load reduction are used as sample labels of the preset power grid operation status assessment model.

[0126] The sample features and sample labels are divided into a test set and a validation set according to a certain ratio. The test set is used to train the preset power grid operation status assessment model. In each round of training, the parameters are updated according to the loss function to obtain the optimal power grid operation status assessment model.

[0127] It should be noted that this application utilizes Monte Carlo simulation combined with photovoltaic and load modeling to generate random initial samples. Considering the correlation and randomness of network random variables, a random walk method is used for processing to obtain processed samples. Subsequently, the optimal power flow is calculated based on the lowest cost-effectiveness and overall network constraints, yielding output data under the optimal operating mode. This data serves as training data for training the combined five-layer GCN model. By combining the input and output training data with the error function, the optimal model is obtained. This process not only improves the model's generalization ability but also ensures the timeliness and accuracy of the evaluation results.

[0128] In an optional embodiment, the optimal power flow is calculated with the goal of minimizing operating cost and in conjunction with grid operation constraints. The output data under the optimal operating mode is used as sample features w, and the optimal power flow that satisfies all optimal constraints is used as sample labels y. These are divided into a test set and a validation set in a ratio of 7:3. The test set is fed into the combined GNN constructed in this invention, combined with a loss function, to train the regression model using multiple samples. The specific model settings are as follows: batch size 128; optimizer: Adam, initial learning rate set to 10. -3 A learning rate decay strategy is adopted (decaying by 50% every 10 rounds); regularization: L2 regularization, with the weight decay coefficient set to 10. -4 Early Stopping: Training terminates when the validation set loss does not decrease for 5 consecutive rounds, and training continues on the test set. The number of training rounds is determined by early stopping based on validation set performance, with a maximum of 200 rounds. The model performance is validated using the validation set. Model evaluation uses mean squared error and mean absolute percentage error to assess model performance. Based on the evaluation results, the optimal power grid operation state assessment model is obtained.

[0129] In this embodiment of the application, the parameter update process in each round of training in step S300, based on the loss function, to obtain the optimal power grid operation state evaluation model includes: the loss function is constructed from the mean squared error loss and the penalty term, and is expressed as:

[0130]

[0131] Among them, R b Represents a set of nodes. and x i ζ represents the basic true value and predicted value, respectively. i It is an auxiliary variable, x min and xmax These represent the minimum and maximum allowed values ​​of the output variable, respectively.

[0132] In this embodiment of the application, step S400, which evaluates the power grid operating state based on the optimal power grid operating state evaluation model, includes:

[0133] If the power grid operation status data value output by the optimal power grid operation status assessment model is greater than the preset threshold, then the power grid operation status is judged to be a fault state.

[0134] The probability of a fault state is quantified into the power grid operating state. For each power grid operating state that exceeds a preset threshold, its economic loss is calculated.

[0135] By generating spatiotemporally relevant scenarios, calculating the consequences of each scenario, and combining all risk indicators, the expected risk outcome corresponding to the power grid operation status data is calculated as a measure of system-level risk.

[0136] In an optional embodiment, a system operation evaluation scheme is introduced to evaluate the actual system operation status based on the trained optimal power grid operation status evaluation model. The main steps are as follows:

[0137] The stability and economic performance of a system are evaluated using the probability of adverse events and the costs incurred when they occur. Let A be the model output value, including key operating parameters such as system power generation output, load distribution, node voltage, and phase angle. These parameters directly reflect the grid's operating status and potential risks. Let E be the reliability threshold. Then, the probability of an adverse event (A>E) is expressed as:

[0138] P(A,E)=p(A>E)

[0139] The probability of an event, P(A,E), can be quantified by using Monte Carlo simulation to statistically determine the proportion of samples where A>E.

[0140]

[0141] Where N is the number of samples, if A i >E, then U i =1, otherwise U i =0.

[0142] The risk assessment logic of this application is based on both probability and consequences. Based on the above calculations, the probability of the system exceeding the reliability threshold and the quantified economic losses after exceeding the threshold are obtained. Finally, the overall risk of the system can be comprehensively assessed through the following: A certain failure consequence (economic cost) is defined as:

[0143]

[0144] Among them, A *=AE,C(A * ) indicates that it depends on A * The cost function is calculated by generating N spatiotemporally relevant scenarios, calculating the consequences of each scenario, and then considering all risk indicators to calculate the expected risk outcomes for all samples. This expected value is then taken as the system-level risk.

[0145]

[0146] It should be noted that the risk assessment framework in this application embodiment is applicable to various failure modes related to reserve adequacy, supply flexibility, etc.

[0147] It should be noted that this application applies the combined GCN model to rapid power grid operation assessment, effectively addressing the volatility and uncertainty brought about by the high proportion of renewable energy grid connection and flexible loads, and significantly improving the model's robustness to complex power grid environments. By integrating the real-time advantages of the five-layer GCN model with the offline optimization characteristics of optimal power flow, a rapid and accurate assessment of the power grid's operating status is achieved. Regarding the selection and optimization of the algorithm model, this application introduces a five-layer GCN model, which can efficiently process graph-structured data, extract key features, and output observations, significantly improving the timeliness and accuracy of the assessment. Especially in complex and ever-changing large power grid environments, it can quickly respond to dynamic changes in network topology, achieving accurate assessment of power system reliability. Furthermore, combined with the optimal power flow algorithm, this method performs optimization in the offline stage, further improving the accuracy and efficiency of the assessment.

[0148] Overall, this application provides a solid guarantee for the sustainable utilization of renewable energy in large power grids and the safe and stable operation of power systems. It not only improves computing efficiency, but also provides more refined considerations for factors such as the intermittency and volatility of new energy output in complex and ever-changing power grid environments, thereby enabling rapid and reliable assessment of the power grid's operating status.

[0149] Example 2: The above example is an illustrative scheme of a rapid power grid operation assessment method. It should be noted that the technical solution of this rapid power grid operation assessment system belongs to the same concept as the technical solution of the aforementioned rapid power grid operation assessment method. Details not described in detail in this example can be found in the description of the technical solution of the aforementioned rapid power grid operation assessment method.

[0150] This embodiment provides a rapid power grid operation assessment system, comprising:

[0151] The data acquisition and preprocessing module is used to acquire probability distribution data of photovoltaic output and load demand in the power system, and preprocess the probability distribution data of photovoltaic output and load demand to obtain the dynamic characteristics of photovoltaic output and load demand.

[0152] The optimal power flow calculation module is used to calculate the optimal power flow with the goal of minimizing operating costs and in combination with grid operation constraints, so as to obtain the optimal power output and load reduction under the optimal grid operation mode.

[0153] The model optimization module is used to preset the power grid operation status assessment model. Based on the dynamic characteristics of photovoltaic power output and load demand, as well as the optimal photovoltaic power output and load reduction, the preset power grid operation status assessment model is trained to obtain the optimal power grid operation status assessment model.

[0154] The evaluation module is used to evaluate the power grid operating status based on the optimal power grid operating status evaluation model.

[0155] This embodiment also provides an electronic device suitable for rapid power grid operation assessment methods, including:

[0156] The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement the method for rapid evaluation of power grid operation as proposed in the above embodiments.

[0157] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the method for rapid evaluation of power grid operation as proposed in the above embodiments.

[0158] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for rapid evaluation of power grid operation proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0159] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0160] Example 3, referring to Figures 2-4This is one embodiment of the present invention. This embodiment uses actual wind power data and system operation data measured in Southwest China to conduct test experiments to verify the beneficial effects of the present invention.

[0161] In this embodiment, the following is selected: Figure 2 The standard IEEE 118-node system shown serves as the basis for this study. It comprises 118 nodes (i.e., power transmission and distribution points), 186 lines, and various generator and load configurations, effectively simulating the complexity and diversity of real-world power grids. Secondly, based on the actual operation of the power grid, beta and normal distributions were used to simulate the grid's operating state on a one-hour timescale. Two initial samples covering various possible states of the power grid under different operating conditions were generated monthly (720 samples were generated for the one-hour timescale).

[0162] Processing flow as follows Figure 3 As shown, the initial samples were processed using a random walk method to simulate various random disturbances and changes that the power grid may encounter in actual operation. After processing, the number of samples on the 1-hour timescale increased to 720*1000 groups. Then, considering the system constraints of the power grid (such as line transmission capacity limitations, generator output limitations, etc.), the optimal power flow calculation was performed on the processed data to obtain all the generator outputs and load reductions that meet the conditions. These results, together with the corresponding power grid topology data (such as node connection relationships, line parameters, etc.), were used as input features (X), while generator outputs and load reductions were used as output labels (Y). Finally, the dataset was divided into training and validation sets in an 8:2 ratio to train the model. The model parameters were continuously adjusted to minimize the prediction error, thereby obtaining the optimal model g(w|X), where w is the model parameter, X is the input data, and g(·) is the explicit model.

[0163] Based on the calculated optimal model g(w|X), and using the actual wind power data measured in the southwest and system operation data, new actual input data X is constructed. _input The data is then fed into the optimal model g(w|X) to obtain the corresponding output data. Figure 4 The comparison between the actual and predicted values ​​shows that the method of this invention has a good effect on predicting the observed values ​​and can well present the reliability index under the actual system operation.

[0164] Example 4, refer to Figure 5 This is an embodiment of the present invention. This embodiment takes the generator sets (traditional units and new energy power supply units), load and their time-series operation status data of a certain region throughout 2022 as an example, and verifies the beneficial effects of the present invention through comparative experiments.

[0165] In this embodiment, the historical data for the southwestern region in 2022 includes traditional unit output curves (coal-fired and gas-fired), renewable energy output (wind power and solar power), and load time-series data (15-minute resolution). The evaluation results show that, in terms of prediction accuracy, the mean squared error of the GNN model is 2.1MW. 2 It significantly outperforms CNN's 15.3MW. 2 In terms of computational efficiency, GNN takes only 0.5ms per scene, a 240-fold improvement over MILP's 120ms; in terms of convergence speed, it is significantly faster than MILP. Figure 5 It can be seen that the method of this invention can basically converge in 25 iterations, which is much higher than the traditional method. In terms of risk coverage, it can be seen from the output of the two models that GNN can identify risk situations faster and more accurately, which is much higher than CNN. This is due to the error caused by the convergence problem. At the same time, it shows that the method of this invention can accurately evaluate the system. Combined with the dispatch center, the method of this invention fully demonstrates its significant advantages in improving the safety and economy of power grid operation at this stage.

[0166] This invention innovatively applies a combined GCN model to rapid power grid operation assessment. This strategy effectively addresses the volatility and uncertainty brought about by the high proportion of renewable energy grid connection and flexible loads, significantly improving the model's robustness to complex power grid environments. By integrating the real-time advantages of the five-layer GCN model with the offline optimization characteristics of Optimal Power Flow (OPF), a rapid and accurate assessment of the power grid's operating status is achieved. This provides a solid guarantee for the sustainable utilization of renewable energy in large power grids and the safe and stable operation of the power system. It not only improves computational efficiency but also provides more refined considerations for factors such as the intermittency and volatility of renewable energy output in complex and ever-changing power grid environments, enabling rapid and reliable assessment of the power grid's operating status.

[0167] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for rapid evaluation of power grid operation, characterized in that, include: The probability distribution data of photovoltaic power output and load demand in the power system are obtained, and the probability distribution data of photovoltaic power output and load demand are preprocessed to obtain the dynamic characteristics of photovoltaic power output and load demand. Optimal power flow is calculated by taking minimum operating cost as the objective and combining grid operation constraints, and the optimal power output and load reduction under the optimal grid operation mode are obtained. A preset power grid operation status assessment model is trained based on the dynamic characteristics of photovoltaic power output and load demand, the optimal photovoltaic power output, and the load reduction to obtain the optimal power grid operation status assessment model. The power grid operating status is evaluated based on the optimal power grid operating status evaluation model.

2. The rapid power grid operation assessment method as described in claim 1, characterized in that, The assessment of the power grid operating status based on the optimal power grid operating status assessment model includes: If the power grid operation status data value output by the optimal power grid operation status assessment model is greater than the preset threshold, then the power grid operation status is judged to be a fault state. The probability of a fault state is quantified into the power grid operating state. For each power grid operating state that exceeds a preset threshold, its economic loss is calculated. By generating spatiotemporally relevant scenarios, calculating the consequences of each scenario, and combining all risk indicators, the expected risk outcome corresponding to the power grid operation status data is calculated as a measure of system-level risk.

3. The rapid power grid operation assessment method as described in claim 2, characterized in that, The optimal power grid operation status assessment model is obtained by training a preset power grid operation status assessment model based on the dynamic characteristics of photovoltaic power output and load demand, the optimal power output, and the load reduction. The dynamic characteristics of photovoltaic output and load demand are used as sample features of the preset power grid operation status assessment model, and the optimal photovoltaic output and load reduction are used as sample labels of the preset power grid operation status assessment model. The sample features and sample labels are divided into a test set and a validation set according to a certain ratio. The test set is used to train the preset power grid operation status assessment model. In each round of training, the parameters are updated according to the loss function to obtain the optimal power grid operation status assessment model.

4. The rapid power grid operation assessment method as described in claim 3, characterized in that, In each round of training, parameters are updated based on the loss function to obtain the optimal power grid operation state assessment model. The loss function is constructed from the mean squared error loss and a penalty term, and is expressed as follows: Among them, R b Represents a set of nodes. and x i ζ represents the basic true value and predicted value, respectively. i It is an auxiliary variable, x min and x max These represent the minimum and maximum allowed values ​​of the output variable, respectively.

5. The rapid power grid operation assessment method as described in claim 3 or 4, characterized in that, The pre-defined power grid operation status assessment model includes: a symmetric normalized aggregation model with self-loops added to the graph convolutional neural network by changing the aggregation and update functions; and a power grid operation status assessment model constructed using five graph convolutional neural network layers and a fully connected readout layer. Five graph convolutional neural network layers are used to update node representations, and the readout layer is used to output the output of traditional units, photovoltaic units, and node voltage amplitude and phase angle data of load reduction magnitude. The aggregation model is represented as: A * =A+I N Among them, A * It is the adjacency matrix of an undirected graph G plus self-connections, I N It is the identity matrix, D * It is A * The degree matrix, H (l) W is the activation unit matrix of the l-th layer. (l) It is the parameter matrix for each layer.

6. The rapid power grid operation assessment method as described in claim 5, characterized in that, Optimal power flow is calculated by taking minimum operating cost as the objective and combining it with grid operation constraints, resulting in the optimal photovoltaic output and load shedding under the optimal grid operation mode, including: The photovoltaic power output calculation uses minimizing cost-effectiveness as the objective function: Where F is the objective function considering photovoltaic output, and P pv P Gi The outputs are photovoltaic power and traditional unit power, respectively. pv a is the output cost coefficient of the photovoltaic unit. i b i c i Let be the cost coefficient of the i-th traditional unit; Power grid operation constraints include power balance constraints, maximum and minimum actual and reactive power output constraints of power sources, voltage phase angle constraints, and branch power flow constraints.

7. The rapid power grid operation assessment method as described in claim 6, characterized in that, The probability distribution data of photovoltaic power output and load demand in the power system are obtained, and the probability distribution data of photovoltaic power output and load demand are preprocessed to obtain the dynamic characteristics of photovoltaic power output and load demand, including: modeling photovoltaic power output and load demand, and generating photovoltaic power output and load demand at different time steps based on the photovoltaic power output model and the load demand model. A photovoltaic-load joint probability distribution model is constructed to obtain the joint probability density of photovoltaic and load. The photovoltaic-load joint probability distribution is sampled according to random sampling to obtain the initial photovoltaic output and load demand samples. The initial photovoltaic power output and load demand samples are processed using the random walk method. The time series characteristics of photovoltaic power output and load demand are obtained by superimposing random disturbances from the state at the previous time step with the state at the current time step.

8. A rapid power grid operation assessment system, applied to the method described in any one of claims 1-7, characterized in that, include: The data acquisition and preprocessing module is used to acquire probability distribution data of photovoltaic power output and load demand in the power system, and preprocess the probability distribution data of photovoltaic power output and load demand to obtain the dynamic characteristics of photovoltaic power output and load demand. The optimal power flow calculation module is used to calculate the optimal power flow with the goal of minimizing operating costs and in combination with grid operation constraints, so as to obtain the optimal power output and load reduction under the optimal grid operation mode. The model optimization module is used to preset the power grid operation status assessment model, and train the preset power grid operation status assessment model based on the dynamic characteristics of photovoltaic output and load demand, the optimal photovoltaic output situation and the load reduction amount, so as to obtain the optimal power grid operation status assessment model. The evaluation module is used to evaluate the power grid operating status based on the optimal power grid operating status evaluation model.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the power grid operation rapid assessment method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the rapid power grid operation assessment method according to any one of claims 1 to 7.

Images