Method, device and terminal for predicting operation state of micro-grid digital twinborn body
By using the tandem echo state network in the operation state prediction model of the microgrid digital twin for lightweight abstraction and cropping, and parameter optimization is performed on the edge computing terminal, the problem of low operation efficiency of the microgrid digital twin on the edge computing terminal is solved, achieving efficient operation and reduced complexity.
Patent Information
- Application Number
- CN202510260550.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-06
- Publication Date
- 2025-06-24
AI Technical Summary
Existing microgrid digital twins are difficult to operate efficiently on edge computing terminals, mainly due to the complexity and computing power burden caused by the isomerization of equipment types and information differentiation.
By building an operating state prediction model based on multiple echo state networks connected in series, lightweight abstraction and cropping of microgrid digital twins is performed, model complexity is reduced, and parameter optimization is performed on edge computing terminals.
It realizes the efficient operation of the microgrid digital twin on the edge computing terminal, reduces the computing power and communication burden, and supports the refined simulation and high-complexity computing of the microgrid.
Smart Images
Figure CN120196889A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of microgrid digital twins, and specifically relates to a method for predicting the operating state of a microgrid digital twin, a device for predicting the operating state of a microgrid digital twin, an edge computing terminal, a machine-readable storage medium, and a computer program product. Background Art
[0002] The microgrid digital twin (MG-DT: Micro-Grid Digital Twin) corresponds to the microgrid in the real world and is a complete mapping of the microgrid system in the virtual digital space. Through high-density dynamic data, it fully reflects the dynamic changes of entities and their relationships within the time scale of the entire life cycle. It can achieve the interaction of information and actions with the physical microgrid system, and at the same time can support the application system to realize intelligent analysis, dynamic decision-making, and mutual sensing and cooperation based on the full model data of the digital twin. It is a collection of digital life forms or a digital twin on a larger scale and scope.
[0003] The microgrid has two operating states: islanded operation and grid-connected operation, with grid-connected operation being the main one. Only by coordinating with the distribution network can its efficiency be maximized. Existing microgrids have characteristics such as heterogeneous equipment types and information differences, which make it difficult to establish a refined unified information model for the microgrid digital twin, increasing the computational complexity and amount of digital twins. If the microgrid digital twin adopts the cloud computing method, it will cause a huge computing power burden and communication burden on the cloud platform master station; if the edge computing method is adopted, the limited computing power is difficult to support the refined simulation and high-complexity operation of the microgrid. Therefore, it is urgent to perform lightweight abstraction and trimming on the microgrid digital twin model to adapt to the efficient operation of the microgrid digital twin on the edge computing terminal. Summary of the Invention
[0004] The purpose of the embodiments of the present invention is to provide a method, device, and terminal for predicting the operating state of a microgrid digital twin, so as to solve the problem of how to perform lightweight abstraction and trimming on the microgrid digital twin model to adapt to the efficient operation of the microgrid digital twin on the edge computing terminal.
[0005] To achieve the above purpose, the embodiments of the present invention provide a method for predicting the operating state of a microgrid digital twin, which is applied to an edge computing terminal. The method includes:
[0006] Obtain the performance data of the microgrid;
[0007] Input the performance data into the operating state prediction model of the microgrid digital twin to obtain the operating state prediction result of the microgrid digital twin output by the operating state prediction model;
[0008] Among them, the operating state prediction model includes a feature extraction layer and an output layer. The feature extraction layer is constructed based on a plurality of echo state networks connected in series, and is used to extract features from the performance data to obtain operating state features; the output layer is used to perform feature transformation on the operating state features to obtain the operating state prediction result.
[0009] The operating state prediction model is trained based on the sample performance data of the microgrid and the operating state labels of the digital twin of the microgrid corresponding to the sample performance data; when training the operating state prediction model, the model parameters of the output layer are optimized.
[0010] Optionally, the operating state prediction model further includes a Bayesian optimization layer, which is used to optimize the model parameters of the output layer by the Bayesian optimization method when training the operating state prediction model.
[0011] Optionally, the model parameters of all the echo state networks are the same.
[0012] Optionally, the operating state prediction model is trained through the following steps:
[0013] Obtain a sample performance data set of the microgrid; the sample performance data set includes a plurality of sample performance data.
[0014] Determine a plurality of initial model parameters of the output layer, and determine the objective function value corresponding to each initial model parameter based on the plurality of first sample performance data.
[0015] Construct a data set based on a plurality of sample points; each sample point includes the initial model parameter and the objective function value corresponding to the initial model parameter.
[0016] Repeat the following steps until the set stop condition is reached:
[0017] Construct a Gaussian process surrogate model based on the plurality of sample points in the data set to fit the posterior distribution of the objective function value.
[0018] Solve the Gaussian process surrogate model based on the acquisition function to obtain the next optimized model parameter of the output layer.
[0019] Determine the objective function value of the next optimized model parameter based on the plurality of first sample performance data, and add the next optimized model parameter and the objective function value of the next optimized model parameter as updated sample points to the data set.
[0020] Determine the model parameter that makes the objective function value optimal in the data set as the model parameter of the output layer of the operating state prediction model.
[0021] Optionally, the performance data includes at least one of voltage data and current data of the microgrid.
[0022] On the other hand, an embodiment of the present invention further provides an operating state prediction device for a microgrid digital twin, including:
[0023] An acquisition module, configured to acquire performance data of the microgrid;
[0024] A prediction module, configured to input the performance data into an operating state prediction model of the microgrid digital twin to obtain an operating state prediction result of the microgrid digital twin output by the operating state prediction model;
[0025] Wherein, the operating state prediction model includes a feature extraction layer and an output layer. The feature extraction layer is constructed based on a plurality of echo state networks connected in series, and is configured to extract features from the performance data to obtain operating state features; the output layer is configured to perform feature transformation on the operating state features to obtain the operating state prediction result;
[0026] The operating state prediction model is trained based on sample performance data of the microgrid and operating state labels of the microgrid digital twin corresponding to the sample performance data; when training the operating state prediction model, the model parameters of the output layer are optimized.
[0027] Optionally, the operating state prediction model further includes a Bayesian optimization layer, configured to optimize the model parameters of the output layer by a Bayesian optimization method when training the operating state prediction model.
[0028] Optionally, the model parameters of all the echo state networks are the same.
[0029] Optionally, the operating state prediction model is trained through the following steps:
[0030] Acquire a sample performance data set of the microgrid; the sample performance data set includes a plurality of sample performance data;
[0031] Determine a plurality of initial model parameters of the output layer, and determine the objective function value corresponding to each initial model parameter based on the plurality of first sample performance data;
[0032] Construct a data set based on a plurality of sample points; each sample point includes the initial model parameter and the objective function value corresponding to the initial model parameter;
[0033] Repeat the following steps until a set stop condition is reached:
[0034] Construct a Gaussian process surrogate model based on the plurality of sample points of the data set to fit the posterior distribution of the objective function value;
[0035] Solving the Gaussian process proxy model based on the acquisition function to obtain the next optimized model parameter of the output layer;
[0036] Determine the objective function value of the next optimization model parameter based on the plurality of first sample performance data, and add the next optimization model parameter and the objective function value of the next optimization model parameter as update sample points to the data set;
[0037] The model parameters that optimize the objective function value in the data set are determined as the model parameters of the output layer of the operating status prediction model.
[0038] Optionally, the performance data includes at least one of voltage data and current data of the microgrid.
[0039] On the other hand, the present invention also provides an edge computing terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for predicting the operating status of the microgrid digital twin when executing the program.
[0040] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for predicting the operating status of the microgrid digital twin.
[0041] On the other hand, the present invention also provides a computer program product, including a computer program, which implements the above-mentioned method for predicting the operating status of the microgrid digital twin when executed by a processor.
[0042] Through the above-mentioned technical scheme, the feature extraction layer of the operation status prediction model of the embodiment of the present invention is constructed based on multiple echo state networks connected in series, so as to realize lightweight abstraction and tailoring of the operation status model of the microgrid digital twin, reduce the complexity to support the deployment of its edge computing terminal; and when training the operation status prediction model of the embodiment of the present invention, only the model parameters of the output layer are adjusted, which helps to reduce the complexity and computational complexity of the model, and realize efficient operation of the microgrid digital twin in the edge computing terminal.
[0043] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:
[0045] Figure 1 It is a schematic flow chart of the operation state prediction method of the microgrid digital twin provided by the present invention;
[0046] Figure 2 It is a schematic structural diagram of the operation state prediction model of the microgrid digital twin provided by the present invention;
[0047] Figure 3 It is a schematic structural diagram of the operation state prediction device of the microgrid digital twin provided by the present invention;
[0048] Figure 4 It is a schematic structural diagram of the edge computing terminal provided by the present invention. Detailed implementation manners
[0049] The following will describe in detail the specific implementation manners of the embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific implementation manners described herein are only used to illustrate and explain the embodiments of the present invention, and are not used to limit the embodiments of the present invention.
[0050] Method embodiment
[0051] Please refer to Figure 1 , an embodiment of the present invention provides an operation state prediction method for a microgrid digital twin, which is applied to an edge computing terminal. The method includes:
[0052] Step 100, obtain the performance data of the microgrid.
[0053] The edge computing terminal obtains the performance data of the microgrid. The performance data is data representing the main performance characteristics of the operation state of the microgrid. In some embodiments, the performance data may include at least one of the voltage data and current data of the microgrid. That is, the performance data may include the voltage data of the microgrid, or the current data of the microgrid, or the voltage data and current data of the microgrid. These data are collected by the monitoring and acquisition devices or sensors of the microgrid. In other embodiments, the performance data of the microgrid may be at least one of the power consumption load, harmonic distortion rate, voltage fluctuation and voltage flicker. That is, the performance data of the microgrid may be one or more of the power consumption load, harmonic distortion rate, voltage fluctuation and voltage flicker.
[0054] Step 200, input the performance data into the operation state prediction model of the microgrid digital twin to obtain the operation state prediction result of the microgrid digital twin output by the operation state prediction model.
[0055] The edge computing terminal inputs the performance data into the operation state prediction model of the microgrid digital twin, and obtains the operation state prediction result of the microgrid digital twin output by the operation state prediction model. In some embodiments, the operation state prediction result can be characterized by active power and / or reactive power. That is, the operation state prediction result can be characterized by active power, or reactive power, or both active power and reactive power. For example, in the embodiments of the present invention, the current data and voltage data detected at the microgrid connection point are used as the input to the operation state prediction model of the microgrid digital twin, and the active power and reactive power of the microgrid digital twin are obtained as the output results for the operation state inference of the microgrid digital twin.
[0056] Please refer to Figure 2 , the operation state prediction model of the microgrid digital twin in the embodiments of the present invention includes an input layer, a feature extraction layer (or called a cascaded ESN layer), and an output layer.
[0057] The input layer is used to receive the performance data characterizing the operation state of the microgrid, and these data come from the monitoring and acquisition devices or sensors of the microgrid. The operation state data collected by the monitoring and acquisition devices or sensors of the microgrid are stored in the historical data sample library, providing the sample data source for the operation state prediction model of the microgrid digital twin.
[0058] The feature extraction layer is constructed based on a plurality of cascaded echo state networks, and is used to extract features from the performance data to obtain operation state features. Specifically, please refer to Figure 2 , the feature extraction layer is composed of k echo state networks (Echo State Network, ESN) connected in series in sequence. That is, except for the first echo state network, the input value of each echo state network is the output value of the previous-level echo state network.
[0059] Specifically, the structural formula of a plurality of cascaded echo state networks (Cas-ESN) is as follows:
[0060]
[0061] Among them, u(t) in the input layer is the performance data of the microgrid input at time t, and the number of neurons is K. u(t + 1) is the performance data of the microgrid input at time t + 1, and W in is the input weight matrix from the input layer to the reservoir. x1(t) is the output value of the first reservoir at time t. The reservoir neurons store historical sequence information, and the connection state is random and fixed. W is the internal weight matrix of each reservoir. f is the activation function of the reservoir neurons, and usually the Sigmoid function is selected. o k (t + 1) is the value of the connection state between the kth reservoir and the (k - 1)th reservoir at time t + 1. fout is the activation function of the output layer neurons. out is the output weight matrix of the output layer. k-1 (t) is the value of the connection state between the k-1th storage pool and the k-2th storage pool at time t. k (t) is the output value of the kth storage pool at time t. k (t+1) is the output value of the kth reservoir at time t+1. x1(t+1) is the output value of the first reservoir at time t+1. The output layer y(t) is the final output state at time t, and the number of neurons is L. The output layer y(t+1) is the final output state at time t+1. k (t) is the value of the connection state between the kth reserve pool and the k-1th reserve pool at time t.
[0062] The output layer is used to perform feature conversion on the running status features to obtain the running status prediction result.
[0063] The feature extraction layer of the operation state prediction model of the embodiment of the present invention is constructed based on multiple echo state networks connected in series. In the echo state network, the neuron connections and weights of the reservoir are randomly initialized and kept fixed. This design avoids the complex weight training process in the traditional neural network, thereby greatly simplifying the network structure. The embodiment of the present invention realizes lightweight abstraction and tailoring of the operation state model of the microgrid digital twin, reducing the complexity to support the deployment of its edge computing terminal. The echo state network maps the input signal to a high-dimensional "echo state" space and completes the prediction task by performing linear regression on the space. This high-dimensional mapping helps to capture subtle features and patterns in the data and improve the prediction accuracy. And the series-connected echo state network can be regarded as a series of multiple echo state network modules. Each module is responsible for processing a part of the features or patterns of the input microgrid sample performance data, and passing the information to the next module in series. This modular design makes the model more flexible and scalable.
[0064] The operating status prediction model is trained based on the sample performance data of the microgrid and the operating status labels of the microgrid digital twin corresponding to the sample performance data. The sample performance data can be historical performance data collected by the monitoring and collection device or sensor of the microgrid. These historical performance data are stored in the historical data sample library, providing a sample data source for the operating status prediction model of the microgrid digital twin. When training the operating status prediction model, the model parameters (or hyperparameters) of the output layer are optimized. Specifically, the input weight matrix W from the input layer to the reserve pool in the operating status prediction model of the embodiment of the present invention inThe weight matrix W inside the reserve pool is generated before training and will not change during training. During training, only the weight matrix W of the output layer is updated. out Therefore, the running state prediction model of the embodiment of the present invention only adjusts the weight matrix W of the output layer out , which helps to reduce the complexity and computational effort of the model.
[0065] The feature extraction layer of the operating status prediction model of the embodiment of the present invention is constructed based on multiple echo state networks connected in series, which realizes lightweight abstraction and tailoring of the operating status model of the microgrid digital twin, reduces the complexity to support the deployment of its edge computing terminals; and when training the operating status prediction model of the embodiment of the present invention, only the model parameters of the output layer are adjusted, which helps to reduce the complexity and computational complexity of the model, and realize efficient operation of the microgrid digital twin in the edge computing terminal.
[0066] In other aspects of the embodiments of the present invention, in order to further reduce the number of model parameters. In the series structure, the number of model parameters can be further reduced by parameter sharing. For example, the model parameters of all the echo state networks are the same. Specifically, multiple echo state networks can use the same reservoir structure and weight matrix, and only adjust the weight matrix of the output layer, so that the embodiments of the present invention help to further reduce the complexity and computational complexity of the model.
[0067] For other aspects of the embodiments of the present invention, please refer to Figure 2 The running state prediction model also includes a Bayesian optimization layer (Bayes optimization layer), which is used to optimize the model parameters of the output layer by Bayesian optimization method when the running state prediction model is trained. The Bayesian optimization method is an algorithm that effectively finds model parameters with good generalization ability. It samples the acquisition function based on prior information, establishes an evaluation index for the performance of the cascade echo state network model, and adopts an optimization algorithm with Gaussian process as the probability proxy model.
[0068] Specifically, the running state prediction model of the embodiment of the present invention is trained through the following steps:
[0069] Step 10: Obtain a sample performance data set of the microgrid; the sample performance data set includes a plurality of sample performance data.
[0070] The edge computing terminal obtains a sample performance dataset. The sample performance data in the sample performance dataset can be historical performance data collected by the monitoring and acquisition devices or sensors of the microgrid. These historical performance data are stored in the historical data sample library, providing a sample data source for the operation status prediction model of the microgrid digital twin. In addition, the edge computing terminal also needs to determine the model parameter search space (or hyperparameter search space) A and the objective function f. The model parameter search space represents the value range of the model parameters of the output layer. The objective function f can be the minimum error between the output prediction value and the true label value of the operation status prediction model.
[0071] Step 20: Determine multiple initial model parameters of the output layer, and determine the objective function value corresponding to each of the initial model parameters based on the multiple first sample performance data.
[0072] The edge computing terminal determines multiple initial model parameters of the output layer. For example, multiple initial model parameters {z1, z2, …, z t} are generated by random sampling or Latin hypercube sampling. The edge computing terminal inputs the multiple first sample performance data into the operation status prediction model to determine the objective function value f(z i ), where i = 1, 2, …, t.
[0073] Step 30: Construct a dataset based on multiple sample points; each of the sample points includes the initial model parameter and the objective function value corresponding to the initial model parameter.
[0074] The edge computing terminal takes each initial model parameter and the objective function value corresponding to the initial model parameter as a sample point (z i , f(z i )) to construct a dataset D = {(z i , f(z i ))} composed of multiple sample points, where i = 1, 2, …, t.
[0075] Repeat the following steps until the set stop condition is reached:
[0076] Step 40: Construct a Gaussian process surrogate model based on the multiple sample points of the dataset to fit the posterior distribution of the objective function value.
[0077] The edge computing terminal trains a Gaussian process surrogate model L based on the dataset to fit the posterior distribution of the objective function value. y = f(z) + ε; where y is the observed value of the Gaussian process surrogate model, z is the model parameter of the output layer, ε is the noise with an independent and identically distributed mean of 0 and a variance of a 2 , and f is a function containing model parameters.
[0078] Step 50: Solve the Gaussian process surrogate model based on the acquisition function to obtain the next optimized model parameter of the output layer.
[0079] The edge computing terminal solves the Gaussian process surrogate model based on any one of the acquisition functions S of expected improvement, upper confidence bound, and probability improvement to obtain the next optimized model parameter z of the output layer t+1 .
[0080] Step 60: Determine the objective function value of the next optimized model parameter based on the multiple first sample performance data, and use the next optimized model parameter and the objective function value of the next optimized model parameter as updated sample points to be added to the data set.
[0081] The edge computing terminal inputs the multiple first sample performance data into the operating state prediction model to determine the next optimized model parameter z t+1 corresponding objective function value f * , and add this set of updated sample points (z t+1 , f * ) to the data set D to repeatedly execute the model parameter iteration of the output layer of the operating state prediction model in steps 40 to 60 until the loop optimization times T are reached, and then exit the loop.
[0082] The distribution of the Gaussian process can be represented by the mean function m() and the covariance function k():
[0083] f2(z) ∼ GP(m(z), k(z, z')); (Formula - 2)
[0084]
[0085] Among them, f2(z) is the Gaussian process function. K is the covariance matrix, which is calculated from the hyperparameter group data z 1:t = {z1,..., z t} obtained from the previous t optimizations. According to the probability surrogate model distributions obtained from the previous t and t + 1 optimizations. z and z' represent model parameters. GP() is the Gaussian distribution function.
[0086]
[0087] Among them, k t+1 = [k(z t+1 , z1),..., k(z t+1 , z t )] T ; (Formula - 5)
[0088] Obtain the prediction distribution:
[0089]
[0090]
[0091] where z t+1 is the next optimized model parameter. f * is the next optimized model parameter z t+1 corresponding objective function value. μ(z t+1 ) is the mean function corresponding to the next optimized model parameter z t+1 . is the next optimized model parameter z t+1 corresponding covariance function. E is the mathematical expectation matrix. y 1:t is the Gaussian surrogate model observation value of the first t optimizations. N() is the Poisson distribution with mean u.
[0092] Step 70: Determine the model parameter that optimizes the objective function value in the dataset as the model parameter of the output layer of the operating state prediction model.
[0093] The edge computing terminal finally uses the model parameter that optimizes the objective function value in the dataset as the model parameter of the output layer of the operating state prediction model. Thus, the embodiment of the present invention designs an operating state prediction model with a Bayes-Cas-ESN structure, and optimizes the model parameters of the output layer of the cascaded echo state network (Cas-ESN) through the Bayesian optimization method to improve its operation accuracy.
[0094] In one embodiment, steps 10 to 70 are simplified. The embodiment of the present invention uses the gradient descent algorithm to train the operating state prediction model of the microgrid digital twin as follows:
[0095] Step 1: Input the dataset D, the objective function f, the hyperparameter search space A, the acquisition function S, the Gaussian process surrogate model L, and the number of loop optimizations T.
[0096] Step 2: Calculate the possible distribution p(f * |Z 1:t ,z t+1 ) of the objective function value of this iteration according to the dataset D and the Gaussian process surrogate model L, and solve the next optimized model parameter z t+1 based on the distribution using the acquisition function S;
[0097] Step 3: Use z t+1 to calculate the objective function value f * , and add this set of data (z t+1 ,f * ) to D, t = t + 1;
[0098] Step 4: When the number of iterations exceeds the loop optimization times T, output the model parameters of the output layer; otherwise, repeat Steps 2-3;
[0099] Step 5: Select the training set data and update the state of the echo state network; train the weight matrix W of the output layer out ;
[0100] Step 6: Test the model and verify the output.
[0101] In summary, the embodiment of the present invention constructs a lightweight digital twin model architecture of a microgrid based on Bayes-Cas-ESN, trims the operation state prediction model of the microgrid digital twin body, reduces the complexity to support the deployment of its edge computing terminal; in addition, the embodiment of the present invention designs a digital twin model based on Bayes-Cas-ESN to infer the change trend of its operation state to support the preventive operation and maintenance decision of the microgrid, and simplifies the operation amount of the digital twin body. Finally, the embodiment of the present invention designs a digital twin model of Bayes-Cas-ESN, and optimizes the parameters of the cascaded echo state network (Cas-ESN) by the Bayesian optimization method to improve its operation accuracy.
[0102] The embodiment of the present invention proposes a lightweight operation method for the microgrid digital twin body, and obtains the overall operation eigenvalue of the microgrid through the above model with the performance data (i.e., the current eigenvalue) of the microgrid. The microgrid has a complex structure, and the voltage and current data can better reflect the overall characteristics of the microgrid during operation. Therefore, the embodiment of the present invention can use the current data and voltage data of the microgrid connection point as the input data set, and the active power and reactive power as the output results to infer the operation state of the microgrid digital twin body. The embodiment of the present invention realizes the lightweight abstraction and trimming of the microgrid digital twin model to adapt to the problem of the efficient operation of the microgrid digital twin body at the edge computing terminal.
[0103] Device embodiment
[0104] Please refer to Figure 3 On the other hand, the embodiment of the present invention also provides an operation state prediction device for a microgrid digital twin body, including:
[0105] An acquisition module 301, configured to acquire the performance data of the microgrid;
[0106] A prediction module 302, configured to input the performance data into the operation state prediction model of the microgrid digital twin body to obtain the operation state prediction result of the microgrid digital twin body output by the operation state prediction model;
[0107] Among them, the operating state prediction model includes a feature extraction layer and an output layer. The feature extraction layer is constructed based on a plurality of echo state networks connected in series and is used to extract features from the performance data to obtain operating state features. The output layer is used to perform feature transformation on the operating state features to obtain the operating state prediction result.
[0108] The operating state prediction model is trained based on the sample performance data of the microgrid and the operating state labels of the digital twin corresponding to the sample performance data. When training the operating state prediction model, the model parameters of the output layer are optimized.
[0109] Optionally, the operating state prediction model further includes a Bayesian optimization layer, which is used to optimize the model parameters of the output layer by the Bayesian optimization method when training the operating state prediction model.
[0110] Optionally, the model parameters of all the echo state networks are the same.
[0111] Optionally, the operating state prediction model is trained through the following steps:
[0112] Obtain the sample performance data set of the microgrid. The sample performance data set includes a plurality of sample performance data.
[0113] Determine a plurality of initial model parameters of the output layer, and determine the objective function value corresponding to each initial model parameter based on the plurality of first sample performance data.
[0114] Construct a data set based on a plurality of sample points. Each sample point includes the initial model parameter and the objective function value corresponding to the initial model parameter.
[0115] Repeat the following steps until the set stop condition is reached:
[0116] Construct a Gaussian process surrogate model based on the plurality of sample points in the data set to fit the posterior distribution of the objective function value.
[0117] Solve the Gaussian process surrogate model based on the acquisition function to obtain the next optimized model parameter of the output layer.
[0118] Determine the objective function value of the next optimized model parameter based on the plurality of first sample performance data, and add the next optimized model parameter and the objective function value of the next optimized model parameter as updated sample points to the data set.
[0119] Determine the model parameter that makes the objective function value optimal in the data set as the model parameter of the output layer of the operating state prediction model.
[0120] Optionally, the performance data includes at least one of voltage data and current data of the microgrid.
[0121] The operating state prediction device of the microgrid digital twin includes a processor and a memory. The above-mentioned acquisition module 301, prediction module 302, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to implement corresponding functions.
[0122] The processor contains a kernel, and the kernel retrieves the corresponding program units from the memory. One or more kernels can be set.
[0123] The memory may include non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory includes at least one memory chip.
[0124] Figure 4 An example of the physical structure diagram of an edge computing terminal is shown as Figure 4 As shown, the edge computing terminal may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440. Among them, the processor 410, the communication interface 420, and the memory 430 complete communication with each other through the communication bus 440. The processor 410 can call the logical instructions in the memory 430 to execute the operating state prediction method of the microgrid digital twin, and the method includes: acquiring the performance data of the microgrid; inputting the performance data into the operating state prediction model of the microgrid digital twin to obtain the operating state prediction result of the microgrid digital twin output by the operating state prediction model; wherein, the operating state prediction model includes a feature extraction layer and an output layer, the feature extraction layer is constructed based on a plurality of echo state networks connected in series, and is used for extracting features from the performance data to obtain operating state features; the output layer is used for performing feature transformation on the operating state features to obtain the operating state prediction result; the operating state prediction model is trained based on the sample performance data of the microgrid and the operating state labels of the microgrid digital twin corresponding to the sample performance data; when training the operating state prediction model, the model parameters of the output layer are optimized.
[0125] In addition, when the logical instructions in the above-mentioned memory 430 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.
[0126] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute a method for predicting the operating state of a microgrid digital twin. The method includes: obtaining performance data of the microgrid; inputting the performance data into an operating state prediction model of the microgrid digital twin to obtain an operating state prediction result output by the operating state prediction model. Wherein, the operating state prediction model includes a feature extraction layer and an output layer. The feature extraction layer is constructed based on a plurality of echo state networks connected in series and is used to extract features from the performance data to obtain operating state features. The output layer is used to perform feature transformation on the operating state features to obtain the operating state prediction result. The operating state prediction model is trained based on sample performance data of the microgrid and operating state labels of the microgrid digital twin corresponding to the sample performance data. When training the operating state prediction model, parameter optimization is performed on the model parameters of the output layer.
[0127] In another aspect, the present invention also provides a machine-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements a method for predicting the operating state of a microgrid digital twin. The method includes: obtaining performance data of the microgrid; inputting the performance data into an operating state prediction model of the microgrid digital twin to obtain an operating state prediction result of the microgrid digital twin output by the operating state prediction model; wherein, the operating state prediction model includes a feature extraction layer and an output layer. The feature extraction layer is constructed based on a plurality of echo state networks connected in series and is used to extract features from the performance data to obtain operating state features; the output layer is used to perform feature transformation on the operating state features to obtain the operating state prediction result; the operating state prediction model is trained based on sample performance data of the microgrid and operating state labels of the microgrid digital twin corresponding to the sample performance data; when training the operating state prediction model, the model parameters of the output layer are optimized.
[0128] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.
[0129] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting the operating status of a microgrid digital twin, characterized in that: Applied to an edge computing terminal, the method includes: Obtain microgrid performance data; Inputting the performance data into an operation status prediction model of a microgrid digital twin to obtain an operation status prediction result of the microgrid digital twin output by the operation status prediction model; The running state prediction model includes a feature extraction layer and an output layer. The feature extraction layer is constructed based on a plurality of echo state networks connected in series and is used to extract features from the performance data to obtain running state features. The output layer is used to perform feature conversion on the running state features to obtain the running state prediction result. The operating status prediction model is trained based on sample performance data of the microgrid and the operating status labels of the microgrid digital twin corresponding to the sample performance data; when training the operating status prediction model, the model parameters of the output layer are optimized.
2. The method for predicting the operating status of a microgrid digital twin according to claim 1, characterized in that: The running state prediction model also includes a Bayesian optimization layer, which is used to optimize the model parameters of the output layer by using a Bayesian optimization method when the running state prediction model is trained.
3. The method for predicting the operating status of a microgrid digital twin according to claim 1, characterized in that: The model parameters of all the echo state networks are the same.
4. The method for predicting the operating status of a microgrid digital twin according to claim 1, characterized in that: The running status prediction model is trained by the following steps: Acquire a sample performance data set of a microgrid; the sample performance data set includes a plurality of sample performance data; Determine a plurality of initial model parameters of the output layer, and determine an objective function value corresponding to each of the initial model parameters based on the plurality of first sample performance data; Constructing a data set based on a plurality of sample points; each of the sample points includes the initial model parameters and the objective function value corresponding to the initial model parameters; Repeat the following steps until the set stop condition is reached: Constructing a Gaussian process proxy model based on multiple sample points of the data set to fit the posterior distribution of the objective function value; Solving the Gaussian process proxy model based on the acquisition function to obtain the next optimized model parameter of the output layer; Determine the objective function value of the next optimization model parameter based on the plurality of first sample performance data, and add the next optimization model parameter and the objective function value of the next optimization model parameter as update sample points to the data set; The model parameters that optimize the objective function value in the data set are determined as the model parameters of the output layer of the operating status prediction model.
5. The method for predicting the operating status of a microgrid digital twin according to claim 1, characterized in that: The performance data includes at least one of voltage data and current data of the microgrid.
6. A device for predicting the operating status of a microgrid digital twin, characterized in that: include: An acquisition module, used to acquire performance data of the microgrid; A prediction module, used to input the performance data into an operation status prediction model of a microgrid digital twin, and obtain an operation status prediction result of the microgrid digital twin output by the operation status prediction model; The running state prediction model includes a feature extraction layer and an output layer. The feature extraction layer is constructed based on a plurality of echo state networks connected in series and is used to extract features from the performance data to obtain running state features. The output layer is used to perform feature conversion on the running state features to obtain the running state prediction result. The operating status prediction model is trained based on sample performance data of the microgrid and the operating status labels of the microgrid digital twin corresponding to the sample performance data; when training the operating status prediction model, the model parameters of the output layer are optimized.
7. The operating state prediction device of the microgrid digital twin according to claim 6, characterized in that: The running state prediction model also includes a Bayesian optimization layer, which is used to optimize the model parameters of the output layer by using a Bayesian optimization method when the running state prediction model is trained.
8. The operating state prediction device of the microgrid digital twin according to claim 6, characterized in that: The model parameters of all the echo state networks are the same.
9. The operating state prediction device of the microgrid digital twin according to claim 6, characterized in that: The running status prediction model is trained by the following steps: Acquire a sample performance data set of a microgrid; the sample performance data set includes a plurality of sample performance data; Determine a plurality of initial model parameters of the output layer, and determine an objective function value corresponding to each of the initial model parameters based on the plurality of first sample performance data; Constructing a data set based on a plurality of sample points; each of the sample points includes the initial model parameters and the objective function value corresponding to the initial model parameters; Repeat the following steps until the set stop condition is reached: Constructing a Gaussian process proxy model based on multiple sample points of the data set to fit the posterior distribution of the objective function value; Solving the Gaussian process proxy model based on the acquisition function to obtain the next optimized model parameter of the output layer; Determine the objective function value of the next optimization model parameter based on the plurality of first sample performance data, and add the next optimization model parameter and the objective function value of the next optimization model parameter as update sample points to the data set; The model parameters that optimize the objective function value in the data set are determined as the model parameters of the output layer of the operating status prediction model.
10. The operating state prediction device of the microgrid digital twin according to claim 6, characterized in that: The performance data includes at least one of voltage data and current data of the microgrid.
11. An edge computing terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the operating status prediction method of the microgrid digital twin according to any one of claims 1 to 5 is implemented.
12. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the operating status prediction method of the microgrid digital twin according to any one of claims 1 to 5 is implemented.
13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the operating status prediction method of the microgrid digital twin according to any one of claims 1 to 5 is implemented.