Evaluation method based on operation reliability intelligent evaluation model and related device

Abstract

The invention provides an evaluation method based on an operation reliability intelligent evaluation model and a related device, and the method comprises the steps: obtaining system data of an electric power system, building a minimum load shedding model according to the system data of the electric power system, determining an inherent mode of the minimum load shedding model, and carrying out the evaluation of the operation reliability intelligent evaluation model based on the inherent mode of the minimum load shedding model. Constructing a robust learning loss function; and finally, performing strong generalization learning according to the minimum load shedding neural network, the robust learning loss function and load disturbance to obtain an operation reliability intelligent evaluation model, and inputting system data of the power system into the operation reliability intelligent evaluation model to obtain a result capable of being used for evaluating the operation reliability of the system. And the efficiency and the accuracy of the evaluation result are effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of power systems and automation thereof, and in particular to an evaluation method and related devices based on an intelligent evaluation model of operation reliability. Background Art

[0002] Operation reliability assessment is an important tool for power system operation risk assessment and early warning. In the context of countries around the world vigorously developing renewable energy, due to the intermittent and random nature of wind and solar energy, the power system faces strong uncertainty, and it is necessary to conduct real-time operation reliability assessment.

[0003] However, the existing numerical calculation operation reliability assessment method needs to iteratively solve the minimum load shedding optimization problem under a large number of system states, and the calculation bottleneck is prominent, which is difficult to meet the current evaluation calculation time requirements. Traditional numerical operation reliability assessment methods can be divided into two categories: analytical method and simulation method. The analytical method directly derives the analytical formula for operation reliability calculation based on the probability model of uncertainty and failure of important equipment. The analytical method works well in small power systems with fewer system states and smaller calculation scales. However, when large-scale power systems involve complex operating conditions and a large number of serious events, the analytical method may be very complicated and even unable to consider certain operating conditions. The simulation method simulates different events and then statistically calculates the operation reliability index, so there is no deficiency that the analytical method is not suitable for complex conditions and large-scale power systems. However, the real-time application of the simulation method also has a large number of cumulative calculation bottlenecks for solving system states, and the evaluation cannot be completed within the 5-minute operation time window, making it difficult to apply to the online evaluation of power systems. To this end, scholars have proposed a data-driven evaluation method based on neural networks.

[0004] The data-driven evaluation method is an effective way to break the calculation bottleneck of online evaluation of the simulation method, relying on the advantages of neural networks that do not require iteration and have fast calculation speed. However, the data-driven evaluation method requires the use of a large number of load shedding samples to train the neural network, and then use the trained neural network to replace the time-consuming minimum load shedding calculation of the simulation method. However, the actual power system has strong risk resistance, and the probability of load shedding scenarios is usually low, resulting in a strong imbalance in the load shedding samples. The neural network based on this training is biased, resulting in the inaccuracy of the data-driven evaluation method. The load shedding samples meet the minimum load shedding model. How to efficiently expand the load shedding samples through the minimum load shedding model to avoid large-scale random sampling is an urgent problem to be solved. Summary of the invention

[0005] In view of this, the present application provides an evaluation method and related devices based on an operation reliability intelligent evaluation model to effectively improve the performance of an operation reliability evaluation neural network.

[0006] The first aspect of the present application provides an evaluation method based on an intelligent evaluation model for operating reliability, including:

[0007] Obtain the system data of the power system;

[0008] Establish a minimum load shedding model according to the system data of the power system;

[0009] Determine the inherent mode of the minimum load shedding model;

[0010] Construct a robust learning loss function based on the inherent mode of the minimum load shedding model;

[0011] Perform strong generalization learning according to the minimum load shedding model, the robust learning loss function, and load disturbances to obtain an intelligent evaluation model for operating reliability;

[0012] Input the system data of the power system into the intelligent evaluation model for operating reliability to obtain a result that can be used to evaluate the operating reliability of the system.

[0013] Optionally, the establishing a minimum load shedding model according to the system data of the power system includes:

[0014] Construct a minimum load shedding model according to the bus node load reduction amount information, cost coefficient, load information, voltage amplitude and phase angle information, output of the generator connected to the bus node, phase angle difference between bus nodes, node conductance matrix, node susceptance matrix, active power between nodes, generator status, active power output of the generator at the previous moment, maximum ramp rate of the generator, generator set, bus node set, and branch set.

[0015] Optionally, the constructing a robust learning loss function based on the inherent mode of the minimum load shedding model includes:

[0016] Construct an initial loss function according to the output features of the minimum load shedding neural network and the corresponding labels, the sample size of the training set, and the output dimension;

[0017] Construct a robust learning loss function according to the initial loss function, disturbance rate, load input amount, load reduction amount of the output feature, and weight coefficient of disturbance sample learning.

[0018] Optionally, the performing strong generalization learning according to the minimum load shedding model, the robust learning loss function, and load disturbances to obtain an intelligent evaluation model for operating reliability includes:

[0019] Receive model training configuration information; wherein, the model training configuration information at least includes a training set and a target number of iterations;

[0020] Initialize the parameters of the minimum load shedding neural network to obtain the neural network for initially calculating the minimum load shedding;

[0021] Randomly shuffle the training set;

[0022] Obtain the original sample input features from the shuffled training set;

[0023] Input the original sample input features into the initial minimum load shedding neural network, and output the initially predicted minimum load shedding amount of the sample;

[0024] Randomly generate a load perturbation within the target interval;

[0025] Add the load perturbation to the training set to obtain a perturbed training set;

[0026] Obtain the perturbed sample input features from the perturbed training set;

[0027] Input the perturbed sample input features into the initial minimum load shedding neural network, and output the predicted minimum load shedding amount of the perturbed sample;

[0028] Use the robust learning loss function to calculate the loss values of the initially predicted minimum load shedding amount of the sample and the predicted minimum load shedding amount of the perturbed sample;

[0029] Use the loss values to perform backpropagation to calculate the gradients of the neural network parameters;

[0030] Use an optimizer to update the neural network parameters to obtain the trained minimum load shedding neural network;

[0031] If the current iteration number reaches the target iteration number, use the trained minimum load shedding neural network as the operation reliability intelligent evaluation model;

[0032] If the current iteration number does not reach the target iteration number, use the trained minimum load shedding neural network as the new initial minimum load shedding neural network, and return to execute the step of randomly shuffling the training set.

[0033] The second aspect of this application provides an evaluation device based on the operation reliability intelligent evaluation model, including:

[0034] A system data acquisition unit for acquiring the system data of the power system;

[0035] A model establishment unit for establishing a minimum load shedding model according to the system data of the power system;

[0036] An inherent mode determination unit for determining the inherent mode of the minimum load shedding model;

[0037] A loss function construction unit, configured to construct a robust learning loss function based on the inherent mode of the minimum load shedding model;

[0038] A strong generalization learning unit, configured to perform strong generalization learning according to the minimum load shedding model, the robust learning loss function, and load disturbances, to obtain an intelligent operation reliability evaluation model;

[0039] An evaluation unit, configured to input the system data of the power system into the intelligent operation reliability evaluation model, to obtain a result that can be used to evaluate the system operation reliability.

[0040] Optionally, the model establishment unit includes:

[0041] A model establishment subunit, configured to construct a minimum load shedding model according to the bus node load shedding amount information, cost coefficient, load information, voltage amplitude and phase angle information, output of the generator connected to the bus node, phase angle difference between bus nodes, node conductance matrix, node susceptance matrix, active power between nodes, generator status, active power output of the generator at the previous moment, maximum ramp rate of the generator, generator set, bus node set, and branch set.

[0042] Optionally, the loss function construction unit includes:

[0043] An initial sample loss function construction subunit, configured to construct an initial loss function according to the output features of the minimum load shedding neural network and the corresponding labels, the sample size of the training set, and the output dimension;

[0044] A perturbed sample loss function construction subunit, configured to construct a robust learning loss function according to the initial loss function, perturbation rate, load input amount, load shedding amount of the output features, and weight coefficient of perturbed sample learning.

[0045] Optionally, the strong generalization learning unit includes:

[0046] A receiving unit, configured to receive model training configuration information; wherein, the model training configuration information at least includes a training set and a target number of iterations;

[0047] An initialization unit, configured to initialize the parameters of the minimum load shedding neural network, to obtain an initial neural network for calculating the minimum load shedding;

[0048] A shuffling unit, configured to randomly shuffle the training set;

[0049] A first obtaining unit, configured to obtain original sample input features in the shuffled training set;

[0050] A first input unit, configured to input the original sample input features into the initial minimum cut load neural network, and output an initial sample predicted minimum cut load;

[0051] A load disturbance generation unit, configured to randomly generate a load disturbance within a target interval;

[0052] An addition unit, configured to add the load disturbance to a training set to obtain a disturbed training set;

[0053] A second acquisition unit, configured to acquire disturbed sample input features from the disturbed training set;

[0054] A second input unit, configured to input the disturbed sample input features into the initial minimum cut load neural network, and output a disturbed sample predicted minimum cut load;

[0055] A loss unit, configured to calculate a loss value between the initial sample minimum cut load and the disturbed sample minimum cut load using a robust learning loss function;

[0056] A backpropagation unit, configured to perform backpropagation calculation on the gradient of the neural network parameters using the loss value;

[0057] An update unit, configured to update the neural network parameters using an optimizer to obtain a trained minimum cut load neural network;

[0058] A determination unit, configured to, if the current iteration number reaches a target iteration number, use the trained minimum cut load neural network as a running reliability intelligent evaluation model;

[0059] An activation unit, configured to, if the current iteration number does not reach the target iteration number, use the trained minimum cut load neural network as a new initial minimum cut load neural network, and activate the shuffling unit to perform the random shuffling of the training set.

[0060] The third aspect of the present application provides an electronic device, including:

[0061] One or more processors;

[0062] A storage device, on which one or more programs are stored;

[0063] When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the evaluation method of the running reliability intelligent evaluation model according to any item of the first aspect.

[0064] The fourth aspect of the present application provides a computer storage medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the evaluation method of the running reliability intelligent evaluation model according to any item of the first aspect is implemented.

[0065] As can be seen from the above solution, the present application provides an evaluation method and related device based on an intelligent evaluation model for operation reliability. System data of a power system is obtained. After establishing a minimum load shedding model based on the system data of the power system, the inherent mode of the minimum load shedding model is determined. Based on the inherent mode of the minimum load shedding model, a robust learning loss function is constructed. Finally, strong generalization learning is performed according to the minimum load shedding model, the robust learning loss function, and load disturbances to obtain an intelligent evaluation model for operation reliability. Inputting the system data of the power system into the intelligent evaluation model for operation reliability can obtain results that can be used to evaluate the system operation reliability, effectively improving the efficiency and accuracy of the evaluation results. Description of the Drawings

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0067] Figure 1 Specific flowchart of an evaluation method based on an intelligent evaluation model for operation reliability provided by an embodiment of the present application;

[0068] Figure 2 Flowchart of a method for constructing a robust learning loss function provided by another embodiment of the present application;

[0069] Figure 3 Flowchart of a strong generalization learning method provided by another embodiment of the present application;

[0070] Figure 4 Schematic diagram of the calculation error of a neural network under different training set sample sizes provided by another embodiment of the present application;

[0071] Figure 5 Schematic diagram of an evaluation device based on an intelligent evaluation model for operation reliability provided by another embodiment of the present application;

[0072] Figure 6 Schematic diagram of an electronic device for implementing an evaluation method based on an intelligent evaluation model for operation reliability provided by another embodiment of the present application. Detailed Description of the Embodiments

[0073] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the scope of protection of the present application.

[0074] As used herein, the term "including" and its variants are open-ended, that is, "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0075] It should be noted that the information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0076] It should be noted that the concepts such as "first" and "second" mentioned in the present application are only used to distinguish different devices, modules, or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules, or units.

[0077] It should be noted that the modifications of "one" and "multiple" mentioned in the present application are illustrative rather than restrictive. Those skilled in the art should understand that unless clearly indicated otherwise in the context, it should be understood as "one or more".

[0078] The embodiments of the present application provide an evaluation method based on an intelligent evaluation model for operating reliability, as Figure 1 shown, which specifically includes the following steps:

[0079] S101. Obtain the system data of the power system.

[0080] Among them, the system data of the power system includes but is not limited to information on the load reduction amount of bus nodes, cost coefficients, load information, voltage amplitude and phase angle information, node conductance matrix, node susceptance matrix, generator status, active power output of the generator at the previous moment, etc., which are not limited here.

[0081] S102. Establish a minimum load shedding model according to the system data of the power system.

[0082] In the specific implementation process of this application, a minimum load shedding model can be constructed based on the bus node load shedding amount information, cost coefficient, load information, voltage amplitude and phase angle information, output power information of the generators connected to the bus node, phase angle difference between bus nodes, nodal conductance matrix, nodal susceptance matrix, active power between nodes, generator status, active power output of the generator at the previous moment, maximum ramp rate of the generator, generator set, bus node set, and branch set, without limitation here.

[0083] The minimum load shedding model can be as follows:

[0084] (1)

[0085] (2)

[0086] (3)

[0087] (4)

[0088] (5)

[0089] (6)

[0090] (7)

[0091] (8)

[0092] (9)

[0093] (10)

[0094] Among them, and are the active and reactive load shedding amounts of bus node i respectively; is the cost coefficient of * (for example, in formula (1), which represents the active cost coefficient of bus node i); and are the active and reactive loads of bus node i respectively; and are the magnitudes of the active and reactive power outputs of the generators connected to bus node i; and are the voltage amplitude and phase angle of bus node i respectively; is the phase angle difference between the i-th node and the j-th bus node; and represent the elements in the i-th row and j-th column of the nodal conductance matrix and nodal susceptance matrix respectively; Indicates the active power between the i-th node and the j-th node; Indicates the generator status; Is the active power output of generator i at the previous moment; Is the maximum ramp rate of generator i; , , Are the sets of generators, bus nodes, and branches respectively; And Respectively represent the upper and lower limits of * (e.g., in Equation (10) Represents the upper limit of the active power between the i-th node and the j-th node, The lower limit of the active power between the i-th node and the j-th node).

[0095] S103. Determine the natural mode of the minimum load shedding model.

[0096] It can be understood that the minimum load shedding problem is to calculate the amount of load to be shed without violating the power grid operation constraints, where whether the load shedding amount is zero is related to the load size. Then, if it is known that a certain node needs to shed load, it means that the load at that node is overloaded. If the load is continuously increased at this node, the additional load calculated by the optimal load shedding model also needs to be shed. Conversely, if the load on the node is not overloaded, there is no need to shed load. Writing this rule in mathematical form can be expressed as:

[0097] (11)

[0098] Where, Is the load change at node i; Represents the function of the minimum load shedding model from the input features to the load shedding amount at node i; Represents the remaining input features. The remaining input features include but are not limited to active load, reactive load, equipment status, power grid topology, new energy output, etc., which are not limited here.

[0099] The present invention proves Equation (11) by demonstrating that the new samples after adding load perturbations (perturbed samples obtained by perturbing the initial samples according to Equation (11)) do not affect the optimality and feasibility of the initial samples (obtaining input features by randomly sampling the power system state and using the interior point method to solve the power system minimum load shedding model to obtain output features). For the initial samples ( , , where the load shedding amount of node i is not zero), when given (including but not limited to input features for minimum load shedding calculation such as active load, reactive load, equipment status, power grid topology, new energy output, etc., which are not limited here), the output features It can be calculated by solving the minimum load shedding model of equations (1)-(10). According to equation (11), the perturbed samples based on the initial samples can be directly calculated, that is , , where is the load perturbation amount, . When substituting the perturbed samples into the minimum load shedding model, the constraint conditions (3)-(6) and (8)-(10) have nothing to do with the load and load shedding and will not change. In equation (2), since the load and the load shedding amount increase by the same amount, the injection power of node i remains unchanged, so the power flow equation remains unchanged. For (7), adding the same perturbation to both sides of the inequality, the inequality remains unchanged. Therefore, all the constraint conditions of the perturbed samples in the minimum load shedding model are exactly the same as the initial samples, and the feasibility remains unchanged. The objective function then becomes . And when solving the minimum load shedding model, the load perturbation amount is a constant and does not affect the optimality of the load shedding model. Therefore, when adding the perturbation to the minimum load shedding model of the initial samples, neither the optimality nor the feasibility will change, and except for the perturbed node i, other input-output characteristics are the same. The proof is completed, and the proof result that adding perturbation to the minimum load shedding model does not change the optimality and feasibility is obtained.

[0100] The process of obtaining the perturbed samples by perturbing the initial samples according to equation (11) includes but is not limited to: selecting the samples in the initial sample set with non-zero minimum load shedding amount, and then adding the perturbation amount to the output characteristics of the selected samples. The perturbation amount satisfies the conditions of equation (11). After that, according to equation (11), directly add the corresponding perturbation amount to the input characteristics of the perturbed nodes. Finally, save the input-output characteristics after perturbation to obtain the perturbed samples.

[0101] S104. Construct a robust learning loss function based on the inherent mode of the minimum load shedding model.

[0102] It can be understood that in the data-driven minimum load shedding calculation, due to the complex non-linearity and topological combination explosion characteristics, hundreds of thousands or even more samples are required for the training of the minimum load shedding neural network. The present invention constructs a loss function for the robust learning of the neural network based on the inherent mode of the minimum load shedding, that is, automatically adding perturbed samples, so as to improve the mapping learning of the neural network for solving the minimum load shedding.

[0103] Optionally, in another embodiment of the present application, an implementation manner of step S104, as Figure 2 shown, includes:

[0104] S201. Construct an initial loss function according to the output features of the minimum load shedding neural network, the corresponding labels, the sample size of the training set, and the output dimension.

[0105] The learning of the neural network is to optimize the trainable parameters. Usually, the neural network is supervised and trained using sample labels, that is, minimizing the loss function L, which can be shown as follows:

[0106] (12)

[0107] Where, and respectively represent the i-th output feature of the k-th sample neural network and the corresponding label; M represents the sample size of the training set; N represents the output dimension.

[0108] S202. Construct a robust learning loss function according to the initial loss function, the perturbation rate, the load input amount, the load shedding amount of the output feature, and the weight coefficient of the perturbed sample learning.

[0109] Continuing with the above example, the inherent mode of the minimum load shedding model can be known. If the initial load shedding amount is not zero, a perturbation amount can be added to the load and its shedding amount of the original sample, and the optimality and feasibility will not change. Therefore, the present invention adds perturbations to the non-zero load shedding amounts of all samples to generate more training data. The loss function can be rewritten as Equation (13):

[0110] (13)

[0111] Where, is the perturbation rate of node i, and its value range is ; is the load shedding amount of the i-th output feature of the neural network when the load is used as the input; is the weight coefficient of the perturbed sample learning, which can be set to 0.1.

[0112] S105. Perform strong generalization learning according to the minimum load shedding model, the robust learning loss function, and the load perturbation to obtain an intelligent operation reliability evaluation model.

[0113] Optionally, in a specific implementation process of the present application, an implementation manner of step S105, as shown in Figure 3 includes:

[0114] S301. Receive model training configuration information.

[0115] Among them, receiving the model training configuration information includes at least the training set and the target number of iterations.

[0116] It should be noted that the content in the training set includes, but is not limited to, the active load, reactive load, active power output of the unit, reactive power output of the unit, nodal conductance matrix, nodal susceptance matrix, minimum load shedding amount, etc. of the i-th sample, which is not limited here.

[0117] S302. Initialize the parameters of the minimum load shedding neural network to obtain the neural network for initial calculation of the minimum load shedding.

[0118] S303. Randomly shuffle the training set and obtain the original sample input features from the shuffled training set.

[0119] S304. Input the original sample input features into the initial minimum load shedding neural network, and output the initial sample predicted minimum load shedding amount.

[0120] S305. Randomly generate a load perturbation within the target interval, add the load perturbation to the training set to obtain a perturbed training set, and obtain the perturbed sample input features from the perturbed training set.

[0121] Among them, the target interval is preset by technicians, experts, etc., for example: , which is not limited here.

[0122] S306. Input the perturbed sample input features into the initial minimum load shedding neural network, and output the perturbed sample predicted minimum load shedding amount.

[0123] S307. Use the robust learning loss function to calculate the loss values of the initial sample minimum load shedding amount and the perturbed sample minimum load shedding amount.

[0124] S308. Use the loss value to perform backpropagation to calculate the gradient of the neural network parameters.

[0125] S309. Use the optimizer to update the neural network parameters to obtain the trained minimum load shedding neural network.

[0126] S310. Determine whether the current iteration number has reached the target iteration number.

[0127] Specifically, if the current iteration number has reached the target iteration number, then execute step S311; if the current iteration number has not reached the target iteration number, then execute step S312.

[0128] S311. Use the trained minimum load shedding neural network as the operation reliability intelligent evaluation model.

[0129] S312. Use the trained minimum load shedding neural network as the new initial minimum load shedding neural network.

[0130] The following is the strong generalization learning algorithm (Algorithm 1) provided by the embodiments of the present application:

[0131] Algorithm 1

[0132] Input The learning rate is r, the number of iterations is E, the training dataset is #timg#, initialize the neural network parameters H, and e = 0. Output The trained neural network parameters H. 1 While e < E: 2 Randomly shuffle the training sample set. 3 For all #timg#: 4 When the input is #timg#, calculate the output #timg# of the neural network with parameters H. 5 Generate a load disturbance #timg# with values in the #timg# interval; 6 Determine to explore a new sample #timg# and calculate the neural network output #timg#; 7 Use Equation (13) to calculate the loss value and perform backpropagation to calculate the gradient of the neural network parameters; 8 Use the optimizer to update the neural network parameters H; 9 End for 10 e = e + 1 11 End while 12 Output H

[0133] Among them, the data in represent the active load, reactive load, active power output of the unit, reactive power output of the unit, nodal conductance matrix, nodal susceptance matrix, and minimum load shedding amount of the i-th sample respectively.

[0134] It should be noted that is the minimum load shedding amount predicted by the neural network of the initial sample, corresponding to in formula (13); is the minimum load shedding amount predicted by the neural network of the perturbed sample, corresponding to in formula (13).

[0135] S106. Input the system data of the power system into the intelligent evaluation model of operation reliability to obtain the results that can be used to evaluate the operation reliability of the system.

[0136] Continuing the above embodiment, after performing strong generalization learning based on the minimum load shedding model, robust learning loss function, and load perturbation to obtain the intelligent evaluation model of operation reliability, the system data of the power system can be input into the intelligent evaluation model of operation reliability to obtain the results that can be used to evaluate the operation reliability of the system, thereby effectively improving the efficiency and accuracy of the evaluation results.

[0137] The present invention will be further described below in conjunction with specific implementation schemes.

[0138] Embodiment 1:

[0139] The evaluation of operation reliability in a power system usually involves load fluctuations, random renewable energy generation, and the outage of important power equipment. For load fluctuations, in this embodiment, it is assumed to follow a normal distribution, with the default value as the mean and 0.3 as the standard deviation. In this embodiment, two types of renewable energy, wind power and photovoltaic power, are considered. It is assumed that the wind speed and solar irradiance follow the Weibull distribution and the Beta distribution respectively, where the wind speed follows the Weibull distribution, where , and the solar irradiance follows distribution, where In this embodiment, multiple wind farms and photovoltaic power stations are connected to different nodes, enabling the penetration rate of renewable energy to exceed 20%. In this embodiment, in the IEEE 39-node system, four wind farms with a capacity of 200 MW and four photovoltaic power stations with a capacity of 200 MW are randomly connected to different nodes of the power system. The penetration rate of renewable energy is 20.26%. To reduce the imbalance of training data, this embodiment samples N-1 and N-2 scenario samples to cover all training situations. This embodiment generates 20K and 10K samples for training and testing.

[0140] To prove the effectiveness of the proposed method, this embodiment compares the traditional operation reliability assessment method with the method without using the proposed learning method and the method using the proposed learning method, denoted as M0, M1, and M2 respectively. All neural networks are constructed and trained using the PyTorch framework on a desktop computer with an Intel(R) Core(TM) i7-10700K CPU@ 3.80GHz 3.80 GHz, 16 GB RAM, and an NVIDIA GeForce RTX 2080Ti. The neural network is trained using the Adam optimizer with a learning rate of 0.001. The number of training iterations is 2000 times.

[0141] M0: The traditional operation reliability assessment method that uses the interior point method to solve the minimum load shedding amount. This method is the benchmark method for data.

[0142] M1: This method utilizes a model-embedded graph convolutional neural network architecture and trains the minimum load shedding calculation neural network using the traditional supervised learning method.

[0143] M2: This method utilizes a model-embedded graph convolutional neural network architecture and trains the minimum load shedding calculation neural network using the strong generalization learning method proposed in the present invention.

[0144] In this embodiment, the neural network is trained using the initial training sample set with different data volumes, and then the prediction error of the load shedding amount of the trained neural network in the test set is statistically analyzed. The results are as Figure 4 shown.

[0145] From Figure 4It can be seen that under different training set sample sizes, the calculation error of M2 is always smaller than that of M1. In addition, as the number of training samples decreases, the calculation error of M1 increases sharply, especially when the number of training data is less than 6000. On the contrary, the calculation error of M2 increases slowly and only starts to rise when the data volume is less than 1000. The main reason for such a phenomenon is that the inherent pattern-guided strong generalization learning method proposed in the present invention is not limited to the given training data and can generate more load reduction samples by using known samples. It can not only overcome the problem of unbalanced load shedding samples, but also reduce the dependence of the neural network on a large number of training samples, verifying the effectiveness of the minimum load shedding inherent pattern-guided strong generalization learning algorithm proposed in the present invention.

[0146] Embodiment 2:

[0147] The operation reliability assessment in a power system usually involves load fluctuations, random renewable energy generation, and the outage of important power equipment. For load fluctuations, in this embodiment, it is assumed to follow a normal distribution, with the default value as the mean and 0.3 as the standard deviation. In this embodiment, two types of renewable energy, wind power and photovoltaic power, are considered. It is assumed that the wind speed and solar irradiance follow the Weibull distribution and the Beta distribution respectively, where the wind speed follows the Weibull distribution, where , and the solar irradiance follows distribution, where . In this embodiment, multiple wind farms and photovoltaic power plants are connected to different nodes, so that the penetration rate of renewable energy exceeds 20%. In this embodiment, in the IEEE 39-bus system, four wind farms with a capacity of 200 MW and four photovoltaic power plants with a capacity of 200 MW are randomly connected to different nodes of the power system. The penetration rate of renewable energy is 20.26%. In order to reduce the imbalance of training data, this embodiment samples N-1 and N-2 scenario samples to cover all training situations. In the operation reliability assessment stage, this embodiment simulates a 1% failure probability for reliability assessment. This embodiment generates 20K and 10K samples for training and testing.

[0148] To prove the effectiveness of the proposed method, this embodiment compares the traditional operation reliability assessment method with the method without using the proposed learning method and the method using the proposed learning method, denoted as M0, M1, and M2 respectively. All neural networks are built and trained using the PyTorch framework on a desktop computer with an Intel(R) Core(TM) i7-10700K CPU@ 3.80GHz 3.80 GHz, 16 GB of RAM, and an NVIDIA GeForce RTX 2080Ti. The neural network is trained using the Adam optimizer with a learning rate of 0.001. The number of training iterations is 2000 times.

[0149] M0: The traditional operation reliability evaluation method uses the interior point method to solve the minimum load shedding amount. This method is the benchmark method for data.

[0150] M1: This method utilizes a graph convolutional neural network architecture embedded in the model and trains the minimum load shedding calculation neural network using traditional supervised learning methods.

[0151] M2: This method utilizes a graph convolutional neural network architecture embedded in the model and trains the minimum load shedding calculation neural network using the strong generalization learning method proposed in the present invention.

[0152] This embodiment uses the adequacy indexes of operation reliability for analysis, including Loss of Load Probability (LOLP) and Expected Demand Not Supplied (EDNS), which are expressed as:

[0153] ;

[0154] ;

[0155] Among them, is the probability of the i-th system state; S is the set of system states where load shedding occurs; represents the load shedding amount of node j under system state i.

[0156] After completing the training with 20K samples in the IEEE 39-bus system, the probability of generator and line outages is set to 1% for operation reliability evaluation, and the reliability indexes are statistically analyzed. The results are shown in Table 1. It can be observed that when using the neural network trained with traditional learning methods for operation reliability evaluation, although the calculation error is small, the relative error still reaches more than 10%. However, the strong generalization learning algorithm guided by the inherent mode of minimum load shedding proposed in this paper expands the load shedding samples through random perturbations, promotes the neural network's learning of load shedding samples, and can further reduce the calculation error of the loss of load probability to below 10%, reaching 4.82%, verifying the effectiveness of the proposed learning algorithm.

[0157] Table 1

[0158] Method LOLP #timg# EDNS #timg# M0 0.04479 / 9.90 / M1 0.04932 10.11% 10.158 2.61% M2 0.04263 4.82% 10.035 1.36%

[0159] Among them, represents the relative error of LOLP, represents the relative error of EDNS.

[0160] As can be seen from the above solution, the present application provides an evaluation method based on an intelligent evaluation model for operating reliability. System data of the power system is obtained, and after establishing a minimum load shedding model based on the system data of the power system, the inherent mode of the minimum load shedding model is determined. Based on the inherent mode of the minimum load shedding model, a robust learning loss function is constructed. Finally, strong generalization learning is performed according to the minimum load shedding model, the robust learning loss function, and load disturbances to obtain an intelligent evaluation model for operating reliability. By inputting the system data of the power system into the intelligent evaluation model for operating reliability, a result that can be used to evaluate the operating reliability of the system can be obtained, effectively improving the efficiency and accuracy of the evaluation result.

[0161] Another embodiment of the present application provides an evaluation device based on an intelligent evaluation model for operating reliability, as Figure 5 shown, specifically including:

[0162] A system data acquisition unit 501, configured to acquire system data of the power system.

[0163] A model establishment unit 502, configured to establish a minimum load shedding model.

[0164] Optionally, in another embodiment of the present application, an implementation manner of the model establishment unit 502 includes:

[0165] A model establishment subunit, configured to construct a minimum load shedding model according to the bus node load curtailment amount information, cost coefficient, load information, voltage amplitude and phase angle information, output of the generator connected to the bus node, phase angle difference between bus nodes, node conductance matrix, node susceptance matrix, active power between nodes, generator status, active power output of the generator at the previous moment, maximum ramp rate of the generator, generator set, bus node set, and branch set.

[0166] For the specific working process of the unit disclosed in the above embodiment of the present application, reference may be made to the content of the corresponding method embodiment, which will not be elaborated here.

[0167] An inherent mode determination unit 503, configured to determine the inherent mode of the minimum load shedding model.

[0168] A loss function construction unit 504, configured to construct a robust learning loss function based on the inherent mode of the minimum load shedding model.

[0169] Optionally, in another embodiment of the present application, an implementation manner of the loss function construction unit 504 includes:

[0170] An initial sample loss function construction subunit, configured to construct an initial loss function according to the output features of the minimum load shedding neural network and the corresponding labels, the sample size of the training set, and the output dimension.

[0171] A disturbance sample loss function construction subunit, configured to construct a robust learning loss function according to an initial loss function, a disturbance rate, a load input amount, a load shedding amount of an output feature, and a weight coefficient learned by disturbance samples.

[0172] For the specific working process of the unit disclosed in the foregoing embodiments of the present application, reference may be made to the corresponding method embodiment content, such as Figure 2 shown, which will not be elaborated herein.

[0173] A strong generalization learning unit 505, configured to perform strong generalization learning according to a minimum load shedding model, a robust learning loss function, and a load disturbance to obtain an operation reliability intelligent evaluation model.

[0174] An evaluation unit 506, configured to input system data of a power system into the operation reliability intelligent evaluation model to obtain a result capable of evaluating the system operation reliability.

[0175] For the specific working process of the unit disclosed in the foregoing embodiments of the present application, reference may be made to the corresponding method embodiment content, such as Figure 1 shown, which will not be elaborated herein.

[0176] Optionally, in another embodiment of the present application, an implementation manner of the strong generalization learning unit 505 includes:

[0177] A receiving unit, configured to receive model training configuration information.

[0178] Wherein, the model training configuration information at least includes a training set and a target number of iterations.

[0179] An initialization unit, configured to initialize parameters of a minimum load shedding neural network to obtain an initial neural network for calculating the minimum load shedding.

[0180] A shuffling unit, configured to randomly shuffle the training set.

[0181] A first obtaining unit, configured to obtain original sample input features in the shuffled training set.

[0182] A first input unit, configured to input the original sample input features into the initial minimum load shedding neural network and output an initial sample predicted minimum load shedding amount.

[0183] A load disturbance generation unit, configured to randomly generate a load disturbance within a target interval.

[0184] An adding unit, configured to add the load disturbance to the training set to obtain a disturbed training set.

[0185] A second obtaining unit, configured to obtain disturbed sample input features in the disturbed training set.

[0186] A second input unit for inputting a disturbance sample feature into an initial minimum load shedding neural network and outputting a predicted minimum load shedding amount of the disturbance sample.

[0187] A loss unit for calculating a loss value between the initial sample minimum load shedding amount and the disturbance sample minimum load shedding amount using a robust learning loss function.

[0188] A backpropagation unit for performing backpropagation calculation of the gradient of the neural network parameters using the loss value.

[0189] An update unit for updating the neural network parameters using an optimizer to obtain a trained minimum load shedding neural network.

[0190] A determination unit for, if the current iteration number reaches the target iteration number, using the trained minimum load shedding neural network as an operation reliability intelligent evaluation model.

[0191] An activation unit for, if the current iteration number does not reach the target iteration number, using the trained minimum load shedding neural network as a new initial minimum load shedding neural network and activating the shuffling unit to perform random shuffling of the training set.

[0192] For the specific working process of the units disclosed in the above embodiments of the present application, reference may be made to the corresponding method embodiment content, as Figure 3 shown, and details are not described herein again.

[0193] As can be seen from the above solution, the present application provides an evaluation device based on an operation reliability intelligent evaluation model, which obtains system data of a power system, determines an inherent mode of a minimum load shedding model after establishing the minimum load shedding model according to the system data of the power system, constructs a robust learning loss function based on the inherent mode of the minimum load shedding model; finally, performs strong generalization learning according to the minimum load shedding model, the robust learning loss function, and load disturbances to obtain an operation reliability intelligent evaluation model. Inputting the system data of the power system into the operation reliability intelligent evaluation model can obtain a result that can be used to evaluate the operation reliability of the system, effectively improving the efficiency and accuracy of the evaluation result.

[0194] The functions described above in this article can be at least partially performed by one or more hardware logic components. For example, without limitation, the exemplary types of hardware logic components that can be used include: Field Programmable Gate Array (FPGA), Application Specific Integrated Circuit (ASIC), Application Specific Standard Product (ASSP), System on Chip (SOC), Complex Programmable Logic Device (CPLD), and so on.

[0195] Another embodiment of the present application provides an electronic device, as Figure 6 shown, including:

[0196] One or more processors 601.

[0197] A storage device 602, on which one or more programs are stored.

[0198] When the one or more programs are executed by the one or more processors 601, the one or more processors 601 are caused to implement the evaluation method based on the intelligent evaluation model of running reliability as described in the above embodiments.

[0199] Another embodiment of the present application provides a computer storage medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the evaluation method based on the intelligent evaluation model of running reliability as described in the above embodiments is implemented.

[0200] In the context of the present application, a machine-readable medium may be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0201] It should be noted that the above computer-readable medium in the present application may be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer-readable storage medium may be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device. And in the present application, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium, and this computer-readable signal medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted by any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0202] The above computer-readable medium may be included in the above electronic device; or it may exist separately and not be assembled into the electronic device.

[0203] Another embodiment of the present application provides a computer program product, which is used to execute the above evaluation method based on the operation reliability intelligent evaluation model when the computer program product is executed.

[0204] Specifically, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments of the present application include a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and this computer program contains program code for executing the method shown in the flowchart. In such an embodiment, this computer program can be downloaded and installed from a network through a communication device, or installed from a storage device, or installed from a ROM. When this computer program is executed by a processing device, it executes the above functions defined in the method of the embodiments of the present application.

[0205] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in this application is not necessarily limited to the specific features or acts described above. On the contrary, the specific features and acts described above are merely example forms for implementing this application.

[0206] Although a number of specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this application. Certain features described in the context of separate embodiments can also be implemented in combination in a single embodiment. On the contrary, the various features described in the context of a single embodiment can also be implemented separately or in any suitable sub-combination in multiple embodiments.

[0207] The above description is only a preferred embodiment of this application and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the application involved in this application is not limited to the technical solution formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above application concept. For example, a technical solution formed by mutually replacing the above features with technical features having similar functions (but not limited to) applied in this application.

Claims

1. An evaluation method based on an intelligent evaluation model of operational reliability, characterized in that: include: Obtain system data of the power system; establishing a minimum load shedding model according to system data of the power system; determining an inherent mode of the minimum load shedding model; Constructing a robust learning loss function based on the inherent mode of the minimum load shedding model; Performing strong generalization learning according to the minimum load shedding model, the robust learning loss function and the load disturbance to obtain an intelligent evaluation model for operation reliability; The system data of the power system is input into the intelligent evaluation model of operation reliability to obtain the results that can be used to evaluate the system operation reliability.

2. The evaluation method based on the operation reliability intelligent evaluation model according to claim 1 is characterized in that: The step of establishing a minimum load shedding model according to the system data of the power system comprises: A minimum load shedding model is constructed based on the bus node load reduction information, cost coefficient, load information, voltage amplitude and phase angle information, output information of generators connected to the bus node, phase angle difference between bus nodes, node conductance matrix, node susceptance matrix, active power between nodes, generator status, active output of generator at the previous moment, maximum ramp rate of generator, generator set, bus node set and branch set.

3. The evaluation method based on the operation reliability intelligent evaluation model according to claim 1 is characterized in that: The method of constructing a robust learning loss function based on the inherent mode of the minimum load shedding model includes: Construct an initial loss function based on the output features of the minimum load shedding neural network and the corresponding labels, sample size of the training set, and output dimensions; A robust learning loss function is constructed based on the initial loss function, the disturbance rate, the load input amount, the load reduction amount of the output feature and the weight coefficient of the disturbance sample learning.

4. The evaluation method based on the operation reliability intelligent evaluation model according to claim 1 is characterized in that: The strong generalization learning is performed according to the minimum load shedding model, the robust learning loss function and the load disturbance to obtain an operation reliability intelligent evaluation model, including: Receive model training configuration information; wherein the model training configuration information at least includes a training set and a target number of iterations; Initializing the parameters of the minimum load shedding neural network to obtain the neural network for initial calculation of the minimum load shedding; Randomly shuffle the training set; Get the original sample input features in the shuffled training set; Inputting the original sample input features into the initial minimum load shedding neural network, and outputting the initial sample predicted minimum load shedding amount; A load disturbance is randomly generated within the target interval; Adding the load disturbance to a training set to obtain a disturbance training set; Obtain perturbation sample input features in the perturbation training set; Inputting the disturbance sample input feature into the initial minimum load shedding neural network, and outputting the disturbance sample predicted minimum load shedding amount; Use the robust learning loss function to calculate the loss value of the minimum load shedding of the initial sample and the minimum load shedding of the perturbation sample; Using the loss value to perform back propagation to calculate the gradient of the neural network parameters; Use the optimizer to update the neural network parameters and obtain the trained minimum load shedding neural network; If the current number of iterations reaches the target number of iterations, the trained minimum load shedding neural network is used as an intelligent evaluation model for operation reliability; If the current number of iterations does not reach the target number of iterations, the trained minimum load shedding neural network is used as a new initial minimum load shedding neural network, and the process returns to execute the step of randomly shuffling the training set.

5. An evaluation device based on an intelligent evaluation model of operation reliability, characterized in that: include: A system data acquisition unit, used for acquiring system data of the power system; A model building unit, used for building a minimum load shedding model according to system data of the power system; an inherent mode determination unit, used to determine the inherent mode of the minimum load shedding model; A loss function construction unit, used for constructing a robust learning loss function based on an inherent mode of the minimum load shedding model; A strong generalization learning unit, used for performing strong generalization learning according to the minimum load shedding model, the robust learning loss function and the load disturbance to obtain an operation reliability intelligent evaluation model; The evaluation unit is used to input the system data of the power system into the operation reliability intelligent evaluation model to obtain the results that can be used to evaluate the system operation reliability.

6. The evaluation device based on the operation reliability intelligent evaluation model according to claim 5 is characterized in that: The model building unit comprises: The model building subunit is used to build a minimum load shedding model based on the bus node load reduction information, cost coefficient, load information, voltage amplitude and phase angle information, output information of generators connected to the bus node, phase angle difference between bus nodes, node conductance matrix, node susceptance matrix, active power between nodes, generator status, active output of generator at the previous moment, maximum ramp rate of generator, generator set, bus node set and branch set.

7. The evaluation device based on the operation reliability intelligent evaluation model according to claim 5 is characterized in that: The loss function construction unit comprises: An initial sample loss function construction subunit is used to construct an initial loss function according to the output characteristics of the minimum load shedding neural network and the corresponding labels, the sample size of the training set and the output dimension; The disturbance sample loss function construction subunit is used to construct a robust learning loss function according to the initial loss function, the disturbance rate, the load input amount, the load reduction amount of the output feature and the weight coefficient of the disturbance sample learning.

8. The evaluation device based on the operation reliability intelligent evaluation model according to claim 5, characterized in that: The strong generalization learning unit includes: A receiving unit, configured to receive model training configuration information; wherein the model training configuration information at least includes a training set and a target number of iterations; An initialization unit, used to initialize the parameters of the minimum load shedding neural network, and obtain the neural network for initially calculating the minimum load shedding; Shuffle unit, used to randomly shuffle the training set; A first acquisition unit is used to acquire original sample input features from the shuffled training set; A first input unit is used to input the original sample input feature into the initial minimum load shedding neural network, and output the initial sample predicted minimum load shedding amount; A load disturbance generating unit, used for randomly generating a load disturbance within a target interval; An adding unit, used for adding the load disturbance to a training set to obtain a disturbance training set; A second acquisition unit is used to acquire the perturbation sample input features in the perturbation training set; A second input unit is used to input the disturbance sample input feature into the initial minimum load shedding neural network, and output the disturbance sample predicted minimum load shedding amount; A loss unit, used to calculate the loss value of the minimum load shedding of the initial sample and the minimum load shedding of the perturbation sample using a robust learning loss function; A back propagation unit, used to calculate the gradient of the neural network parameters by back propagation using the loss value; An updating unit, used for updating the neural network parameters using an optimizer to obtain a trained minimum load shedding neural network; a determination unit, configured to use the trained minimum load shedding neural network as an intelligent evaluation model for operation reliability if the current number of iterations reaches a target number of iterations; The activation unit is used to use the trained minimum load shedding neural network as a new initial minimum load shedding neural network if the current number of iterations does not reach the target number of iterations, and activate the shuffling unit to execute the randomly shuffled training set.

9. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the evaluation method based on the operation reliability intelligent evaluation model as described in any one of claims 1 to 4.

10. A computer storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is executed by a processor, the evaluation method based on the operation reliability intelligent evaluation model as described in any one of claims 1 to 4 is implemented.

Images