A spatiotemporal data-driven island microgrid stability evaluation method

Abstract

The present application relates to a kind of island micro-grid stability evaluation method driven by space-time data, belong to island micro-grid evaluation field.It includes the time sequence characteristic selection of implementation operating state variable collection and stability evaluation;Including improved island micro-grid space-time adjacency matrix, reaction system source and load associated state and contact tightness;It also includes island micro-grid node space-time relationship modeling;Graph neural network is used as basic algorithm, introduce attention mechanism to identify complex working condition interference caused by island micro-grid flexible operation characteristics by itself, give different node adaptive importance weight;The data-driven stability evaluation algorithm based on graph neural network, attention mechanism and LSTM network time series prediction layer construction, including offline training online application link, when judging instability, will implement alarm.This method has the advantages of fast evaluation speed, high accuracy, good reliability, etc., can adapt to noise and other factors interference existing in the complex environment of island micro-grid actual operation.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of island micro-grid evaluation, and particularly relates to a time-space data driven island micro-grid stability evaluation method. BACKGROUND

[0002] The island micro-grid usually relies on power electronic converters as the interface of the power supply to reliably access the distributed renewable energy such as wind energy and solar energy, to realize local power production and consumption, and to guarantee the reliable supply of energy on the island. However, the distributed power sources in the island micro-grid are dispersed in space, and the time scales and characteristics of the dynamic behaviors are different, which leads to strong nonlinear dynamic behaviors and low damping characteristics of the system during the fault, and there is frequent and deep coupling influence between the power sources, the power sources and the seawater desalination device, and the flexible loads such as air conditioners. The distributed power sources composed of power electronic devices have weak support capability for the system. Therefore, accurately and timely obtaining the operating state of the island micro-grid and reducing the operating risks faced by the island micro-grid become the key to ensuring the reliability and stability of the island micro-grid when facing major disturbances.

[0003] The transient stability evaluation of the micro-grid is mostly based on the traditional stability evaluation method of the large power grid dominated by synchronous generators, including the time domain simulation method and the direct method. These evaluation methods are widely used in traditional power systems, but have many limitations when applied to micro-grids. The time domain simulation method has high calculation accuracy, but the calculation burden is too heavy for the computing power of the micro-grid deployment, especially in the case of a large number of distributed power sources in the system, which makes it difficult to meet the real-time requirements. The direct method, i.e. the Lyapunov energy function method, can quickly evaluate the transient stability, but it relies on the accurate modeling of the system energy, and the nonlinear characteristics of the numerous power electronic equipment in the micro-grid make accurate modeling a challenge, and there is currently no systematic modeling method and theory that can accurately derive.

[0004] The transient stability of the micro-grid as a new type of power system is obviously different from that of the traditional power system. The transient stability of the traditional power system refers to the ability of the system to recover to the original operating state or transition to a new stable state after being disturbed, while the distributed micro-sources of the micro-grid with power electronic inverter interface make the transient stability mechanism more complex, bringing further challenges to the stability evaluation of the micro-grid. The island micro-grid is a special micro-grid, not only in terms of spatial location, but also in terms of time. The island micro-grid covers a smaller range, has more types of power sources and loads, and the multi-source and load brings more complex dynamic coupling in the time scale, further restricting the application of traditional stability evaluation methods.

[0005] Aiming at the limitations of traditional methods, data-driven stability evaluation can well deal with nonlinear problems, and because a large number of offline training-online operation modes are adopted, the training with high computing power requirement can be carried out on the remote or cloud, and only a lower computing power calculation unit is needed for real-time operation. However, the current data-driven evaluation does not consider the space-time characteristics of the island micro-grid, only analyzes and mines the operation data in the time dimension, and independently considers the nodes, sources and loads in the island micro-grid, which is easy to cause the problems of poor evaluation accuracy and large feature quantity requirement. SUMMARY

[0006] The purpose of the present application is to solve the problems of large voltage and frequency fluctuation amplitude, high instability risk and difficult stability control of the island micro-grid with high proportion of new energy access after large disturbance, and to provide a space-time data-driven island micro-grid stability evaluation method, which realizes accurate and rapid determination of the operation state of the island micro-grid after faults such as typhoon, rainstorm and submarine cable removal.

[0007] To achieve the above purpose, the technical scheme of the present application is: a space-time data-driven island micro-grid stability evaluation method, which comprises: stability evaluation feature selection, island micro-grid node space-time relationship modeling, data-driven evaluation algorithm construction and offline training online application. The method has the advantages of fast evaluation speed, high accuracy and good reliability, and can adapt to the noise and other factor interference in the complex environment of the actual operation of the island micro-grid.

[0008] In an embodiment of the present application, the stability evaluation feature selection is specifically implemented as follows:

[0009] When the island micro-grid is in a stable state, the state variables between each node or each power source remain relatively constant, at this time the voltage, frequency and power in the system form a multi-dimensional space, which collectively reflects the current operating point, operating state and stability of the system. The sampling period is set to T, and the voltage sampling value Vs is

[0010] Vs=[V1,V2,...V k·T ]

[0011] Wherein, k is the sampling number, the single sampling window length Wd satisfies Wd=kT, and the frequency sampling value Fs, the active power sampling value Ps and the reactive power sampling value Qs are

[0012] Fs=[F1,F2,...F k·T ]

[0013] Ps=[P1,P2,...P k·T ]

[0014] Qs = [Q1, Q2,... Qn] k·T ]

[0015] In actual operation, the four selected characteristic quantities differ in sampling frequency due to different time scales of the respective dynamic states; therefore, for the frequency and voltage, a sampling time Tt is selected, and T = n*Tt, and the characteristic quantity X is represented as

[0016]

[0017] In an embodiment of the present application, the stability evaluation feature selection further comprises pre-processing the collected characteristic quantities, that is, first filtering noise and clutter through state estimation, and eliminating outliers caused by including measurement errors or transmission packet loss.

[0018] In an embodiment of the present application, the island micro-grid node space-time relationship modeling is specifically implemented as follows:

[0019] The topology structure of the island micro-grid is a two-dimensional graph Graph, but due to the great difference in the state of each node of the two-dimensional graph Graph, the space-time relationship modeling of the island micro-grid is divided into two links; wherein,

[0020] Link 1 implements graph modeling of the spatial topology:

[0021] Let the adjacency matrix A of the island micro-grid be

[0022]

[0023] Wherein, i represents node i, m represents the total number of nodes in the island micro-grid, a ij represents the connection relationship between node i and node j, 1 if there is a connection, otherwise 0;

[0024] Since the distance of spatial distribution will determine the influence of disturbance under the same conditions, the adjacency matrix A is improved, and the improved adjacency matrix A e is

[0025]

[0026] Wherein, a eij represents the improved connection relationship between node i and node j, and the value range is 0-1, and the calculation method is

[0027]

[0028] Wherein, Z ij represents the impedance between node i and j, represents any one node in the set of all nodes connected to node i, and μ represents all nodes connected to node i;

[0029] Step 2 realizes the integration of time information on the basis of the above graph modeling:

[0030] The feature quantity X is adjusted to obtain a node feature matrix X, that is,

[0031]

[0032] wherein the subscript base represents the reference value of each node selected in advance, that is, the node feature matrix X is changed into a three-dimensional matrix, and is normalized to ensure that each node does not interfere with the accuracy of the stable evaluation algorithm in subsequent calculation due to too large or too small value; for F base and V base , the rated frequency value and the rated voltage value consistent with the whole system are selected, for P base and Q base , the rated power value of each node is selected due to the difference between nodes;

[0033] Through steps 1 and 2, the time information and the space information of the island micro-grid are integrated.

[0034] In an embodiment of the present application, the data-driven evaluation algorithm is constructed and implemented as follows:

[0035] Due to the graph structure of the island micro-grid, the graph neural network GNN is adopted as the basis of the stable evaluation algorithm; the random latent variable z i is introduced to form a matrix Z to enhance the representation ability and the implicit relationship extraction ability of the stable evaluation algorithm, and the spatio-temporal feature encoding layer is

[0036]

[0037] wherein g is a mapping function; for the island micro-grid, the roles between the power supply and the load may be switched, therefore, the attention mechanism is introduced to learn the importance of each node and produce individualized evaluation conclusions for different situations;

[0038] The attention mechanism layer of the stable evaluation inputs the output value of the spatio-temporal feature encoding layer as input, and has

[0039]

[0040] wherein S atten is the output of the attention mechanism layer, v, W1, W2, W3 are weight parameter matrices to be learned and optimized for the neural network, b is a bias matrix, and sigma is a sigmoid function;

[0041] The decoding layer of the spatio-temporal feature is the reverse calculation of the encoding layer, and the output value calculation process is

[0042]

[0043] Wherein, p represents probability, the decoding layer output of the space-time feature of the current moment and the decoding layer output of the space-time feature of the previous moment are input into the long short-term memory network together, the evaluation conclusion of the time sequence is established, that is, the value input into the long short-term memory network is

[0044]

[0045] Wherein, T w Indicates the length of the input data, generally measured by time;

[0046] The output is

[0047]

[0048] Wherein, h t-1 Indicates the hidden value of the LSTM network at t-1 moment, when y(t)>y0, y(t) is set as 1 to indicate that the system is unstable, when y(t)≤y0, y(t) is set as 0 to indicate that the system is stable, and y0 is a set threshold value.

[0049] In an embodiment of the present application, the data-driven evaluation algorithm constructed is trained before being applied online, that is, the data-driven evaluation algorithm is trained, and is applied online.

[0050] In an embodiment of the present application, the data-driven evaluation algorithm constructed is trained, that is, the data-driven evaluation algorithm is trained on the cloud server by using the gradient descent method based on historical operation data and simulation data until the expected accuracy is achieved on the test set.

[0051] In an embodiment of the present application, the online application, that is, the trained data-driven evaluation algorithm is deployed locally in the island micro-grid and is run online until the termination instruction from the control center is received.

[0052] The present application also provides a space-time data-driven island micro-grid stability evaluation system, comprising a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor, when the processor executes the computer program instructions, the method steps as described above can be realized.

[0053] The present application also provides a computer readable storage medium, which stores computer program instructions capable of being executed by a processor, when the processor executes the computer program instructions, the method steps as described above can be realized.

[0054] Compared to existing technologies, this invention offers the following advantages: It considers the strong spatiotemporal coupling characteristics of island microgrids during transient stability due to their close source-load distance and strong dynamic coupling, whereas traditional methods merely extract stability features and trends from time-series data. In contrast, this invention employs a multi-dimensional, multi-spatiotemporal approach, integrating implicit mathematical relationships related to stability assessment in both temporal and spatial correlations. Leveraging the intelligence and nonlinear representation capabilities of data-driven methods, it accurately and promptly senses the system's operating status in the rapidly changing and harsh environment of island microgrids, facilitating timely emergency control intervention and maintaining the long-term reliable and stable operation of the island microgrid. Attached Figure Description

[0055] Figure 1 This is a structural diagram of each component of the present invention;

[0056] Figure 2 This paper compares the accuracy of the proposed method with that of the LSTM algorithm in stability assessment in 100 simulation cases of a microgrid on an island after a three-phase ground fault. Detailed Implementation

[0057] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0058] This invention provides a spatiotemporal data-driven method for stability assessment of island microgrids, including: selection of stability assessment features, modeling of spatiotemporal relationships between island microgrid nodes, construction of a data-driven assessment algorithm, and offline training and online application.

[0059] The following is a detailed implementation process of the present invention.

[0060] like Figure 1 As shown, a spatiotemporal data-driven method for stability assessment of island microgrids has the following operational steps:

[0061] Step 1: Selection of Stability Assessment Features. When the island microgrid is in a steady state, the state variables of each node or power source remain relatively constant. At this time, the voltage, frequency, and power within the system constitute a multi-dimensional space, jointly reflecting the current operating point, operating state, and stability of the system. Setting the sampling period to T, the voltage sample value Vs is...

[0062] Vs = [V1, V2, ... V k·T ]

[0063] Where k is the number of samples. The single sampling window length Wd holds true if Wd = kT. Similarly, the frequency sample value Fs, active power sample value Ps, and reactive power sample value Qs are...

[0064] Fs = [F1, F2,... F k·T ]

[0065] Ps = [P1, P2,... P k·T ]

[0066] Qs = [Q1, Q2,... Q k·T ]

[0067] In actual operation, the four selected characteristic quantities differ in sampling frequency due to different time scales of each dynamic state. Therefore, for frequency and voltage, a smaller sampling time Tt is selected, and T = n*Tt, then the characteristic quantity X can be expressed as

[0068]

[0069] The collected characteristic quantities are first filtered by state estimation to remove noise and clutter, and outliers caused by measurement errors or transmission packet loss are removed to ensure the authenticity of the original data input into the evaluation method.

[0070] Step 2: Modeling of the spatial and temporal relationship of the island microgrid node. The topology structure of the island microgrid is a two-dimensional graph, but due to the great difference in the state of each node of this graph, the spatial and temporal relationship modeling of the island microgrid is divided into two links.

[0071] Link 1 realizes the graph modeling of the spatial topology. Let the adjacency matrix A of the island microgrid be

[0072]

[0073] Where i represents node i, m represents the total number of nodes in the island microgrid. a ij represents the connection relationship between node i and node j, 1 if there is a connection, otherwise 0.

[0074] Since the distance of spatial distribution will determine the influence of disturbance under the same conditions, the above adjacency matrix A is improved in the present application, and the improved adjacency matrix A e is

[0075]

[0076] Where a eij represents the improved connection relationship between node i and node j, and the value range is 0-1, and the calculation method is

[0077]

[0078] Where, Z ij represents the impedance between nodes i and j, represents any one of the set of nodes connected to node i, and μ represents all nodes connected to node i.

[0079] Step 2 realizes the integration of time information on the basis of the above graph modeling. The characteristic quantity X in step 1 is adjusted to obtain the node feature matrix X, that is,

[0080]

[0081] wherein the subscript base represents the reference value of each node selected in advance, that is, the original two-dimensional feature matrix X is changed to a three-dimensional matrix, and is normalized to ensure that each node does not interfere with the accuracy of the stable evaluation algorithm in subsequent calculations due to too large or too small values. For F base and V base , the rated frequency value and the rated voltage value consistent with the whole system are generally selected, and for P base and Q base , the rated power value of each node is generally selected.

[0082] Through the above steps 1 and 2, step 2 integrates the time information and the spatial information of the island microgrid, not only expressing the objective position, but also considering the closeness of electrical connection, and further considering the time sequence characteristics in the data input into the stable evaluation method.

[0083] Step 3: Data-driven evaluation algorithm construction. Due to the graph structure of the island microgrid, the graph neural network (GNN) is adopted as the basis of the stable evaluation algorithm. The random latent variable z i is introduced to form the matrix Z to enhance the representation ability and the implicit relationship extraction ability of the stable evaluation algorithm, and the space-time feature encoding layer is

[0084]

[0085] wherein g is a mapping function;

[0086] Here, a two-layer graph convolutional neural network (GCN) is taken as an example, that is,

[0087] γ1=GCN(X,A)

[0088]

[0089] wherein γ1 is the output of the first layer GCN, ReLu is an activation function, W 0 and W are weight matrices. The calculation is

[0090]

[0091] D is a degree matrix.

[0092] For island microgrid, the role between power supply and load may switch, for example, the energy storage converter is a load when wind power and photovoltaic are abundant, and is a power supply when night load increases, photovoltaic is not available, or power gap is caused by disturbance. Therefore, the attention mechanism is introduced to learn the importance of each node and produce personalized evaluation conclusions for different situations.

[0093] The input of the attention mechanism layer of the stability evaluation is the output value of the space-time feature encoding layer, and

[0094]

[0095] wherein S atten is the output of the attention mechanism layer, v, W1, W2, W3 are weight parameter matrices to be learned and optimized by the neural network, b is a bias matrix, and sigma is a sigmoid function.

[0096] The decoding layer of the space-time feature is the reverse calculation of the encoding layer, but slightly different, and the output value calculation process is

[0097]

[0098] wherein p represents probability. The decoding layer output and the decoding layer output at the previous time are jointly input into the long short-term memory network to establish the time sequence evaluation conclusion, that is, the value input into the long short-term memory network is

[0099]

[0100] The output is

[0101]

[0102] wherein h t-1 represents the hidden value of the LSTM network at the previous time (i.e. at time t-1). When y(t)>y0, y(t) is set to 1 to represent system instability, and when y(t)≤y0, y(t) is set to 0 to represent system stability. y0 is a threshold value set by man.

[0103] Step 4: training the above stability evaluation method. In the cloud server, the gradient descent method is used to train the stability evaluation method of the present application based on historical operation data and simulation data until the desired accuracy is achieved on the test set.

[0104] Step 5: deploying the trained stability evaluation method of the present application in the island microgrid and running online until receiving the termination instruction from the control center.

[0105] As Figure 2The comparison shows that in 100 simulation cases of the island micro-grid of the application, the accuracy of the stability evaluation of the application is significantly higher than that of the LSTM algorithm after a three-phase ground fault.

[0106] The application further provides an island micro-grid stability evaluation system driven by space-time data, comprising a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor, and when the processor executes the computer program instructions, the method steps described above can be realized.

[0107] The application further provides a computer readable storage medium, which stores computer program instructions capable of being executed by a processor, and when the processor executes the computer program instructions, the method steps described above can be realized.

[0108] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "exemplary embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the exemplary description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0109] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0110] The present application is described with reference to flowcharts and / or block diagrams according to the method, device (system) and computer program product of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks.

[0111] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 The flow or flows and / or blocks Figure 1 The functions specified in the flow or flows and / or blocks

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that are executed on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 The flow or flows and / or blocks Figure 1 The functions specified in the flow or flows and / or blocks

[0113] The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application to other forms. Any person skilled in the art can make modifications or alterations to the above-described technical contents, or make equivalent embodiments with equivalent changes. However, any simple modifications, equivalent changes and alterations made to the above embodiments without departing from the technical solutions of the present application, and according to the technical essence of the present application, still belong to the protection scope of the present application.

Claims

1. A spatiotemporal data-driven method for stability assessment of island microgrids, characterized in that, include: The process includes selecting stability assessment features, modeling the spatiotemporal relationships of nodes in an island microgrid, constructing a data-driven assessment algorithm, and offline training followed by online application. The specific implementation of selecting stability assessment features is as follows: When the island microgrid is in a steady state, the state variables of each node or power source remain relatively constant. At this time, the voltage, frequency, and power within the system constitute a multi-dimensional space, jointly reflecting the current operating point, operating state, and stability of the system. Setting the sampling period to T, the voltage sample value Vs is... Vs=[V1,V2,...V k·T ] Where k is the number of samples, and the single sampling window length Wd holds true as Wd = kT. Similarly, the frequency sampled value Fs, active power sampled value Ps, and reactive power sampled value Qs are... Fs=[F1,F2,...F k·T ] Ps=[P1,P2,...P k·T ] Qs=[Q1,Q2,...Q k·T ] In actual operation, the four selected characteristic quantities have different sampling frequencies due to their different dynamic time scales; therefore, for frequency and voltage, a sampling time Tt is chosen, let T = n*Tt, then the characteristic quantity X is expressed as... The spatiotemporal relationship modeling of the island microgrid nodes is specifically implemented as follows: The topology of an island microgrid is a two-dimensional graph. However, due to the significant differences in the states of the nodes in the two-dimensional graph, the spatiotemporal modeling of the island microgrid is divided into two stages; among them, Step 1 involves graph modeling of the spatial topology: Let the adjacency matrix A of the island microgrid be Where i represents node i, m represents the total number of nodes in the island microgrid, and a ij This represents the connection relationship between node i and node j; a value of 1 indicates a connection, and a value of 0 indicates otherwise. Since the distance of the spatial distribution will determine the magnitude of the disturbance under the same conditions, the adjacency matrix A is improved. The improved adjacency matrix A e for Among them, a eij This represents the improved connection relationship between node i and node j, with values ​​ranging from 0 to 1. Its calculation method is as follows: Among them, Z ij This represents the impedance between nodes i and j. Let represent any node in the set of all nodes connected to node i, and let μ represent all nodes connected to node i. Step 2 incorporates time information based on the graph modeling described above: Adjusting the feature quantity X yields the node feature matrix X, i.e. Wherein, the subscript base represents the pre-selected baseline value for each node, that is, converting the node feature matrix X into a three-dimensional matrix and normalizing it to ensure that the value of each node is not too large or too small in subsequent calculations, thus preventing interference with the accuracy of the stable evaluation algorithm; for F base and V base In this regard, selecting a consistent rated frequency and rated voltage value for the entire system is beneficial for P. base and Q base The rated power value for each node varies depending on the node. Through steps 1 and 2, the temporal and spatial information of the island microgrid is integrated; The specific implementation of the data-driven evaluation algorithm is as follows: Due to the graph structure of the island microgrid, a graph neural network (GNN) is used as the basis for the stability evaluation algorithm; a random latent variable z is introduced. i To construct matrix Z and enhance the representational and implicit relation extraction capabilities of the stable evaluation algorithm, the spatiotemporal feature coding layer is then: Where g is the mapping function; for island microgrids, there is a possibility of role switching between power source and load. Therefore, an attention mechanism is introduced to learn the importance of each node and generate personalized evaluation conclusions for different situations. The attention mechanism layer for stable evaluation takes the output value of the spatiotemporal feature encoding layer as input, and has Among them, S atten ν is the output of the attention mechanism layer, W1, W2, and W3 are the weight parameter matrices to be learned and optimized in the neural network, b is the bias matrix, and σ is the sigmoid function. The decoding layer for spatiotemporal features is the reverse computation of the encoding layer, and its output value is calculated as follows: Where p represents probability, the decoding layer output of the spatiotemporal features at the current time step and the decoding layer output of the spatiotemporal features at previous time steps are jointly input into the Long Short-Term Memory (LSTM) network to establish the temporal evaluation conclusion, that is, the value input into the LTM network is... Among them, T w Indicates the length of the input data; The output is Among them, h t-1 Let y(t) represent the hidden value of the LSTM network at time t-1. When y(t)>y0, let y(t)=1 to indicate that the system is unstable. When y(t)≤y0, let y(t)=0 to indicate that the system is stable. y0 is a set threshold.

2. The spatiotemporal data-driven method for evaluating the stability of island microgrids according to claim 1, characterized in that, The selection of stability assessment features also includes preprocessing the collected feature quantities, namely, firstly filtering out noise and clutter through state estimation, and removing outliers caused by measurement errors or packet loss.

3. The spatiotemporal data-driven method for evaluating the stability of island microgrids according to claim 1, characterized in that, The data-driven evaluation algorithm is trained offline and then applied online before the application is trained online.

4. The spatiotemporal data-driven method for evaluating the stability of island microgrids according to claim 3, characterized in that, The data-driven evaluation algorithm is trained on a cloud server using gradient descent based on historical running data and simulation data until the desired accuracy is achieved on the test set.

5. The spatiotemporal data-driven method for evaluating the stability of island microgrids according to claim 3, characterized in that, The online application refers to the deployment and online operation of a pre-trained data-driven evaluation algorithm on the island microgrid until a termination command is received from the control center.

6. A spatiotemporal data-driven stability assessment system for island microgrids, characterized in that, It includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it can implement the steps of the spatiotemporal data-driven island microgrid stability assessment method as described in any one of claims 1-5.

7. A computer-readable storage medium having stored thereon computer program instructions executable by a processor, wherein when the processor executes the computer program instructions, it is able to implement the steps of the spatiotemporal data-driven island microgrid stability assessment method as described in any one of claims 1-5.

Images