Modeling method for evaluating regulation potential of distributed resources in micro-grid
By using a long and short-term memory network model in the microgrid combined with an automatic encoder and a convolutional neural network for data processing, and combining a cloud model with a Bayesian network to build an evaluation model, the problems of insufficient accuracy and poor real-time performance of traditional methods when processing microgrid data are solved, and a more efficient assessment of distributed resource regulation potential is achieved.
Patent Information
- Application Number
- CN202411851470.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-16
- Publication Date
- 2025-04-08
AI Technical Summary
When traditional data acquisition methods and modeling technologies deal with large amounts of distributed resource data of multi-source and high-dimensionality in microgrids, they have problems such as insufficient accuracy, poor real-time performance, and difficulty in capturing complex dynamic characteristics, especially when dealing with uncertainty and ambiguity.
The long and short-term memory network model is used to combine automatic encoder and convolutional neural network to collect, feature extraction and dimensionality reduction of the operating data of distributed resources in the microgrid. Then, through the combination of cloud models and Bayesian networks, a moderating potential evaluation model is constructed to deal with uncertainty and ambiguity.
It improves the efficiency of the distributed resource regulation potential evaluation within the microgrid, can more accurately capture the dynamic changes in resource operation data, handle complex uncertainties and ambiguity, and enhances the robustness and adaptability of the evaluation.
Smart Images
Figure CN119782738A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of microgrids, and in particular to a modeling method for evaluating the regulation potential of distributed resources within a microgrid. Background Art
[0002] With the development of the world economy and the continuous growth of energy demand, global energy shortage and environmental problems have become major issues facing all countries. Energy conservation and rational utilization have become a hot topic of common concern to the world. Vigorously developing clean energy has become a consensus of the international community and an important part of my country's energy strategy. The distributed resources within the microgrid are diverse and complex in nature, which brings great challenges to the accurate modeling of resource regulation potential.
[0003] Traditional data acquisition methods and modeling techniques often have problems such as insufficient accuracy, poor real-time performance, and difficulty in capturing complex dynamic characteristics when processing large amounts of multi-source, high-dimensional distributed resource data. The development of deep learning technology has provided new ideas for processing complex data, but how to effectively apply it to the field of microgrids still requires in-depth research. For example, the use of a microgrid distributed power optimization configuration method based on the particle swarm optimization algorithm (PSO) has a large amount of calculation and a long solution time for large-scale microgrid systems. The parameter selection of the algorithm has a great impact on the optimization results, and multiple experiments are required to determine the appropriate parameters. At the same time, uncertainty and fuzziness are prevalent in microgrids, such as uncertainty in user behavior and changes in the external environment. Traditional methods have limitations in dealing with these problems.
[0004] Therefore, an evaluation modeling method for the regulation potential of distributed resources in microgrids is urgently needed to improve the evaluation efficiency. Summary of the invention
[0005] In view of at least one of the above technical problems, the present invention provides a modeling method for evaluating the regulation potential of distributed resources within a microgrid, and improves the evaluation efficiency of the regulation potential of distributed resources within a microgrid by using an improved method.
[0006] According to a first aspect of the present invention, a modeling method for evaluating the regulation potential of distributed resources within a microgrid is provided, comprising:
[0007] Step S1, real-time collection of operating data of distributed resources in the microgrid and construction of a long short-term memory network model;
[0008] Step S2, using an autoencoder and a convolutional neural network to perform feature extraction and dimensionality reduction processing on the collected operation data, and to mine patterns and features of the operation data;
[0009] Step S3, based on the patterns and characteristics of the operating data, a regulation potential assessment model is constructed using a combination of the cloud model and the Bayesian network.
[0010] In some embodiments of the present invention, constructing the long short-term memory network model includes processing and predicting the operation data sorted in a time series mode, and the prediction formula of the long short-term memory network model is as follows:
[0011] y t =σ(W f *[h t-1 , x t ]+b f )*h t-1 +σ(W i *[h t-1 , x t ]+b i )*tanh(W c *[h t-1 , x t ]+b c )
[0012] Among them, y t represents the output of the long short-term memory network model at time t, h t-1 represents the hidden state at time t-1, x t represents the input of the long short-term memory network model at time t, W f , W i , W c is the weight matrix, b f , b i , b c is the bias vector, σ represents the sigmoid function, and tanh represents the hyperbolic tangent function.
[0013] In some embodiments of the present invention, step S1 specifically includes the following steps:
[0014] Step S11, deploying photoelectric sensors on various distributed resources in the microgrid, the formula for measuring light intensity by the photoelectric sensors is:
[0015] I=k*E
[0016] Among them, I is the light intensity, k is the sensitivity coefficient of the photoelectric sensor, and E is the received light energy;
[0017] The calculation is performed by taking the average value of multiple measurements. The calculation formula is as follows:
[0018]
[0019] Among them, I avgis the average light intensity, n is the number of measurements, I i is the light intensity measured at the i-th time;
[0020] Step S12, establish a data transmission network to transmit the collected operation data to the processor in a timely manner. The data transmission rate is calculated as follows:
[0021] R=B*llg2(1+S / N)
[0022] Where R is the data transmission rate, B is the channel bandwidth, S is the signal power, and N is the noise power;
[0023] Step S13, preprocessing the collected operating data, wherein the preprocessing includes removing abnormal values and filling missing values.
[0024] In some embodiments of the present invention, in step S13, the 3σ principle is used to identify the outliers, and the outlier judgment formula is:
[0025] |x-μ|>3σ
[0026] Where x is the data point, μ is the mean, and σ is the standard deviation.
[0027] In some embodiments of the present invention, in step S13, linear interpolation or a method based on similar days is used to fill in the missing values, and the linear interpolation formula is:
[0028]
[0029] Among them, y is the interpolation result, y1 and y2 are adjacent known data points, t1 and t2 are corresponding time points, and t is the time point to be interpolated.
[0030] In some embodiments of the present invention, the use of the autoencoder and the convolutional neural network includes fusing the convolutional neural network with the encoding part of the autoencoder to jointly extract features from the operating data, and then reconstructing it through the decoding part, and the convolution operation formula of the convolutional neural network is:
[0031]
[0032] Among them, s(i, j) represents the value of the output feature map, x(i+m, j+n) represents the value of the input image, and k(m, n) represents the value of the convolution kernel.
[0033] In some embodiments of the present invention, step S2 specifically includes the following steps:
[0034] Step S21, designing the structure of the autoencoder, including the number of layers and neurons of the encoder and decoder. The encoder gradually compresses the data dimension, and the decoder attempts to reconstruct the original data. The output formula of the encoder is:
[0035] h=σ(W enc *x+b enc )
[0036] Among them, h is the encoded output, W enc is the weight matrix of the encoder, x is the original data, b enc is the bias vector;
[0037] The output formula of the decoder is:
[0038] x′=σ(W dec *x+b dec )
[0039] Among them, x′ is the reconstructed data after decoding, σ is the standard deviation, and W dec is the weight matrix of the decoder, b dec is the bias vector;
[0040] Step S22, introducing a convolution layer into the encoding layer of the autoencoder, using convolution kernels of different sizes and numbers to extract multi-scale features, the output formula of the convolution layer is:
[0041]
[0042] Among them, s(i, j) represents the value of the output feature map, x(i+m, j+n) represents the value of the input image, and k(m, n) represents the value of the convolution kernel;
[0043] Step S23, using L1 and L2 regularization to prevent the model from overfitting, the formula of the L1 regularization term is:
[0044]
[0045] The formula for the L2 regularization term is:
[0046]
[0047] Among them, λ is the regularization parameter, w i are the parameters of the model.
[0048] In some embodiments of the present invention, the cloud model is combined with the Bayesian network, including determining the nodes and edges of the Bayesian network, estimating the probability distribution of the nodes based on prior knowledge and data, and using the output of the cloud model as the input of the Bayesian network for reasoning. The probability reasoning formula of the Bayesian network is:
[0049]
[0050] Among them, P(A|B) represents the probability of event A occurring under the condition that event B occurs.
[0051] In some embodiments of the present invention, step S3 specifically includes the following steps:
[0052] Step S31, determining the parameters of the cloud model to describe the characteristics of uncertainty, the parameters of the cloud model include expected E x and entropy E n , the generation formula of the cloud model is:
[0053] x=E x +E n *randn
[0054] Among them, x is the generated cloud drop value, and randn is a random number of normal distribution;
[0055] Step S32, constructing the Bayesian network structure, and determining the connection relationship between nodes according to the domain knowledge and the relevance of the data;
[0056] Step S33, using a gradient-based optimization algorithm to learn the parameters of the Bayesian network.
[0057] In some embodiments of the present invention, the gradient-based optimization algorithm is a stochastic gradient descent algorithm, and the parameter update formula of the Bayesian network is:
[0058]
[0059] Among them, θ t is the current parameter, α is the learning rate, is the gradient of the loss function.
[0060] The beneficial effects of the present invention are as follows: the present invention constructs a long short-term memory network model, integrates an autoencoder and a convolutional neural network in the data processing stage to perform feature extraction and dimensionality reduction processing on the collected operating data, and combines a cloud model and a Bayesian network; compared with the prior art, the present invention provides a novel method for dealing with uncertainty and ambiguity, and improves the evaluation efficiency of the distributed resource regulation potential within the microgrid. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0062] Figure 1 A schematic diagram of a process for performing potential assessment modeling in an embodiment of the present invention;
[0063] Figure 2 Schematic diagram of the specific steps in step S1 in the embodiment of the present invention;
[0064] Figure 3 Schematic diagram of the specific steps in step S2 in the embodiment of the present invention;
[0065] Figure 4 Schematic diagram of the specific steps in step S3 in the embodiment of the present invention. DETAILED DESCRIPTION
[0066] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.
[0067] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only implementation method.
[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0069] like Figures 1 to 4 The modeling method for evaluating the regulation potential of distributed resources within the microgrid shown includes the following steps:
[0070] Step S1, collect the operating data of distributed resources in the microgrid in real time and build a long short-term memory network model. When collecting operating data, it is preferred to use IoT sensors and intelligent monitoring equipment. The operating data of various distributed resources include the light intensity I of distributed photovoltaics, the charge and discharge state S of the energy storage system, and the power E of the energy storage system.
[0071] In an embodiment of the present invention, constructing a long short-term memory network model includes processing and predicting the operation data sorted in a time series mode. The prediction formula of the long short-term memory network model is as follows:
[0072] y t =σ(W f *[h t-1 , x t ]+b f )*h t-1 +σ(W i *[h t-1 , x t ]+b i )*tanh(W c *[h t-1 , x t ]+b c )
[0073] Among them, y t represents the output of the long short-term memory network model at time t, h t-1 represents the hidden state at time t-1, x t represents the input of the long short-term memory network model at time t, W f , W i , W c is the weight matrix, b f , b i , b c is the bias vector, σ represents the sigmoid function, and tanh represents the hyperbolic tangent function.
[0074] The long short-term memory network model can handle dependencies in long time series data, accurately capture the dynamic changes of resource operation data, provide rich and accurate raw data for subsequent modeling, and ensure that the model can reflect the actual operation of resources. The construction method of the long short-term memory network model includes determining the number of layers, number of neurons and other hyperparameters of the long short-term memory network model, using the collected data for training, and adjusting the model parameters through the back propagation algorithm.
[0075] Step S2, use autoencoders and convolutional neural networks to perform feature extraction and dimensionality reduction processing on the collected operation data, and mine the patterns and features of the operation data. It should be noted that autoencoders are a technology in deep learning. They are a type of artificial neural network used in semi-supervised learning and unsupervised learning. Their function is to characterize and learn the input information by taking the input information as the learning target. Autoencoders can learn the compressed representation of data, and convolutional neural networks (CNNs) can extract local features. The combination of the two improves data processing efficiency and the learning ability of the model, and more accurately extracts the multi-dimensional regulatory attributes of resources.
[0076] In an embodiment of the present invention, the use of the autoencoder and the convolutional neural network includes fusing the convolutional neural network with the encoding part of the autoencoder, jointly extracting features from the running data, and then reconstructing it through the decoding part. The convolution operation formula of the convolutional neural network is:
[0077]
[0078] Among them, s(i, j) represents the value of the output feature map, x(i+m, j+n) represents the value of the input image, and k(m, n) represents the value of the convolution kernel.
[0079] Step S3, based on the patterns and characteristics of the operating data, the cloud model is combined with the Bayesian network to construct a regulation potential assessment model. The cloud model is used to deal with the uncertainty and ambiguity of user behavior, external environmental factors, etc. in the patterns and characteristics of the operating data, and to improve the robustness and adaptability of the regulation potential assessment modeling. The cloud model can describe uncertainty, and the Bayesian network can perform probabilistic reasoning. The combination of the two can better cope with complex and changeable actual situations, making the regulation potential assessment modeling more reliable.
[0080] In an embodiment of the present invention, the cloud model is combined with the Bayesian network, including determining the nodes and edges of the Bayesian network, estimating the probability distribution of the nodes based on prior knowledge and data, and using the output of the cloud model as the input of the Bayesian network for reasoning. The probability reasoning formula of the Bayesian network is:
[0081]
[0082] Among them, P(A|B) represents the probability of event A occurring under the condition that event B occurs.
[0083] In an embodiment of the present invention, the above step S1 specifically includes the following steps:
[0084] Step S11, deploy photoelectric sensors on various distributed resources in the microgrid. Specifically, photoelectric sensors that can accurately measure light intensity can be installed on distributed photovoltaic equipment, and their measurement error is controlled within ±5%, which can ensure the accuracy and real-time nature of data collection. The formula for photoelectric sensors to measure light intensity is:
[0085] I=k*E
[0086] Among them, I is the light intensity, k is the sensitivity coefficient of the photoelectric sensor, and E is the received light energy.
[0087] At the same time, in order to improve the measurement stability, the method of taking the average value of multiple measurements is used for calculation. The calculation formula is as follows:
[0088]
[0089] Among them, I avg is the average light intensity, n is the number of measurements, I i is the light intensity measured at the i-th time.
[0090] Step S12, establish a data transmission network to transmit the collected operation data to the processor in a timely manner. The data transmission network preferably uses 5G communication technology or a dedicated wireless communication protocol to ensure that the data transmission delay is less than 100 milliseconds and the data packet loss rate is less than 1%. The data transmission rate is calculated as follows:
[0091] R=B*log2(1+S / N)
[0092] Among them, R is the data transmission rate, B is the channel bandwidth, S is the signal power, and N is the noise power.
[0093] Step S13, preprocessing the collected operating data, including removing abnormal values and filling missing values.
[0094] In an embodiment of the present invention, it is preferred to use a statistical method or a machine learning algorithm to identify outliers, such as using the 3σ principle to identify outliers, and the outlier judgment formula is:
[0095] |x-μ|>3σ
[0096] Where x is the data point, μ is the mean, and σ is the standard deviation.
[0097] In the embodiment of the present invention, for missing values, linear interpolation or a method based on similar days is preferably used to fill the missing values, and the linear interpolation formula is:
[0098]
[0099] Among them, y is the interpolation result, y1 and y2 are adjacent known data points, t1 and t2 are corresponding time points, and t is the time point to be interpolated.
[0100] The regulatory potential assessment modeling constructed with the long short-term memory network model can handle dependencies in long time series data, accurately capture the dynamic changes of resource operation data, provide rich and accurate original data for subsequent modeling, and ensure that the model can reflect the actual operation of resources.
[0101] In the above step S2, the following steps are specifically included:
[0102] In an embodiment of the present invention, step S21 designs the structure of the autoencoder, including the number of layers and neurons of the encoder and decoder. The encoder gradually compresses the data dimension, and the decoder attempts to reconstruct the original data. The output formula of the encoder is:
[0103] h=σ(W enc *x+b enc )
[0104] Among them, h is the encoded output, W enc is the weight matrix of the encoder, x is the original data, b enc is the bias vector.
[0105] The output formula of the decoder is:
[0106] x′=σ(W dec *x+b dec )
[0107] Among them, x′ is the reconstructed data after decoding, σ is the standard deviation, and W dec is the weight matrix of the decoder, b dec is the bias vector.
[0108] Step S22, a convolution layer is introduced into the encoding layer of the autoencoder, and convolution kernels of different sizes and numbers are used to extract multi-scale features. The output formula of the convolution layer is:
[0109]
[0110] Among them, s(i, j) represents the value of the output feature map, x(i+m, j+n) represents the value of the input image, and k(m, n) represents the value of the convolution kernel.
[0111] It should be noted that in order to increase the diversity of features, multiple convolution kernels are used in parallel, that is, S = [S1, S2, S3, ..., S n ], where S is a feature vector composed of the outputs of multiple convolution kernels.
[0112] Step S23, use L1 and L2 regularization to prevent the model from overfitting. The formula of L1 regularization term is:
[0113]
[0114] The formula for the L2 regularization term is:
[0115]
[0116] Among them, λ is the regularization parameter, w i are the parameters of the model.
[0117] It should be noted that step S2 also includes step S24, that is, using the early stopping method to stop the training in advance according to the performance of the verification set.
[0118] In the early stopping method, according to the loss function value L of the validation set val To decide whether to stop training, when L val The training is stopped when the value no longer decreases within a number of consecutive epochs. An epoch refers to the process in which the entire training data set is completely traversed by the neural network.
[0119] The autoencoder can learn the compressed representation of data, and the convolutional neural network can extract local features. The combination of the two improves the data processing efficiency and the learning ability of the model, and more accurately extracts the multi-dimensional regulatory properties of resources.
[0120] In the above step S3, the following steps are specifically included:
[0121] Step S31, determining the parameters of the cloud model to describe the characteristics of uncertainty. The parameters of the cloud model include the expected E x and entropy E n , the generation formula of the cloud model is:
[0122] x=E x +E n *randn
[0123] Among them, x is the generated cloud drop value, and randn is a random number from the normal distribution.
[0124] Step S32, constructing a Bayesian network structure, determining the connection relationship between nodes based on the domain knowledge and the correlation of data. That is, the dependency relationship between nodes is described by calculating the conditional probability table (CPT), and the expression for calculating the conditional probability table is P(Xi|Pa(X i )), where X i is a node, Pa(X i ) as its parent node.
[0125] Step S33, using a gradient-based optimization algorithm to learn the parameters of the Bayesian network.
[0126] On the basis of the above embodiments, in an embodiment of the present invention, the gradient-based optimization algorithm is a stochastic gradient descent algorithm, and the parameter update formula of the Bayesian network is:
[0127]
[0128] Among them, θ t is the current parameter, α is the learning rate, is the gradient of the loss function.
[0129] Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. A modeling method for evaluating the regulation potential of distributed resources within a microgrid, characterized in that: include: Step S1, real-time collection of operating data of distributed resources in the microgrid and construction of a long short-term memory network model; Step S2, using an autoencoder and a convolutional neural network to perform feature extraction and dimensionality reduction processing on the collected operation data, and to mine patterns and features of the operation data; Step S3, based on the patterns and characteristics of the operating data, a regulation potential assessment model is constructed using a combination of the cloud model and the Bayesian network.
2. The modeling method for evaluating the regulation potential of distributed resources within a microgrid according to claim 1 is characterized in that: Constructing the long short-term memory network model includes processing and predicting the operation data sorted in a time series mode. The prediction formula of the long short-term memory network model is as follows: y t =σ(W f *[h t-1 ,x t ]+b f )*h t-1 +σ(W i *[h t-1 ,x t ]+b i )*tanh(W c *[h t-1 ,x t ]+b c ) Among them, y t represents the output of the long short-term memory network model at time t, h t-1 represents the hidden state at time t-1, x t represents the input of the long short-term memory network model at time t, W f , W i , W c is the weight matrix, b f 、b i 、b c is the bias vector, σ represents the sigmoid function, and tanh represents the hyperbolic tangent function.
3. The modeling method for evaluating the regulation potential of distributed resources within a microgrid according to claim 2 is characterized in that: The step S1 specifically includes the following steps: Step S11, deploying photoelectric sensors on various distributed resources in the microgrid, the formula for measuring light intensity by the photoelectric sensors is: I=k*E Among them, I is the light intensity, k is the sensitivity coefficient of the photoelectric sensor, and E is the received light energy; The calculation is performed by taking the average value of multiple measurements. The calculation formula is as follows: Among them, I avg is the average light intensity, n is the number of measurements, I i is the light intensity measured at the i-th time; Step S12, establish a data transmission network to transmit the collected operation data to the processor in a timely manner. The data transmission rate is calculated as follows: R=B*log2(1+S / N) Where R is the data transmission rate, B is the channel bandwidth, S is the signal power, and N is the noise power; Step S13, preprocessing the collected operating data, wherein the preprocessing includes removing abnormal values and filling missing values.
4. The modeling method for evaluating the regulation potential of distributed resources within a microgrid according to claim 3 is characterized in that: In step S13, the 3σ principle is used to identify the abnormal value, and the abnormal value judgment formula is: |x-μ|>3σ Where x is the data point, μ is the mean, and σ is the standard deviation.
5. The method for evaluating and modeling the potential of distributed resource regulation within a microgrid according to claim 3, characterized in that: In step S13, linear interpolation or a method based on similar days is used to fill in the missing values. The linear interpolation formula is: Among them, y is the interpolation result, y1 and y2 are adjacent known data points, t1 and t2 are corresponding time points, and t is the time point to be interpolated.
6. The method for evaluating the regulation potential of distributed resources within a microgrid according to claim 1, characterized in that: The use of the autoencoder and the convolutional neural network includes fusing the convolutional neural network with the encoding part of the autoencoder to jointly extract features of the operating data, and then reconstructing the data through the decoding part. The convolution operation formula of the convolutional neural network is: Among them, s(i,j) represents the value of the output feature map, x(i+m,j+n) represents the value of the input image, and k(m,n) represents the value of the convolution kernel.
7. The method for evaluating and modeling the regulation potential of distributed resources within a microgrid according to claim 6, characterized in that: The step S2 specifically includes the following steps: Step S21, designing the structure of the autoencoder, including the number of layers and neurons of the encoder and decoder. The encoder gradually compresses the data dimension, and the decoder attempts to reconstruct the original data. The output formula of the encoder is: h=σ(W enc *x+b enc ) Among them, h is the encoded output, W enc is the weight matrix of the encoder, x is the original data, b enc is the bias vector; The output formula of the decoder is: x′=σ(W dec *x+b dec ) Among them, x′ is the reconstructed data after decoding, σ is the standard deviation, and W dec is the weight matrix of the decoder, b dec is the bias vector; Step S22, introducing a convolution layer into the encoding layer of the autoencoder, using convolution kernels of different sizes and numbers to extract multi-scale features, the output formula of the convolution layer is: Among them, s(i,j) represents the value of the output feature map, x(i+m,j+n) represents the value of the input image, and k(m,n) represents the value of the convolution kernel; Step S23, using L1 and L2 regularization to prevent model overfitting, the formula of the L1 regularization term is: The formula for the L2 regularization term is: Among them, λ is the regularization parameter, w i are the parameters of the model.
8. The method for evaluating and modeling the regulation potential of distributed resources within a microgrid according to claim 1, characterized in that: The cloud model is combined with the Bayesian network. It includes determining the nodes and edges of the Bayesian network, estimating the probability distribution of the nodes based on prior knowledge and data, and using the output of the cloud model as the input of the Bayesian network for reasoning. The probability reasoning formula of the Bayesian network is: Among them, P(A|B) represents the probability of event A occurring under the condition that event B occurs.
9. The method for evaluating and modeling the regulation potential of distributed resources within a microgrid according to claim 8, characterized in that: The step S3 specifically includes the following steps: Step S31, determining the parameters of the cloud model to describe the characteristics of uncertainty, the parameters of the cloud model include expected E x and entropy E n , the generation formula of the cloud model is: x=E x +E n *randn Among them, x is the generated cloud drop value, and randn is a random number of normal distribution; Step S32, constructing the Bayesian network structure, and determining the connection relationship between nodes according to the domain knowledge and the relevance of the data; Step S33, using a gradient-based optimization algorithm to learn the parameters of the Bayesian network.
10. The modeling method for evaluating the regulation potential of distributed resources within a microgrid according to claim 9, characterized in that: The gradient-based optimization algorithm is a stochastic gradient descent algorithm, and the parameter update formula of the Bayesian network is: Among them, θ t is the current parameter, α is the learning rate, is the gradient of the loss function.