Power grid distributed energy management platform based on cloud computing

Abstract

The invention discloses a power grid distributed energy management platform based on cloud computing, and relates to the technical field of energy management, and the platform comprises the steps: setting distributed energy management nodes in a power grid distributed energy management region, and constructing a network topology based on the set positions of the distributed energy management nodes, collecting historical energy data in the distributed energy management area of the power grid based on the constructed network topology, constructing a standard sample set, preprocessing the standard sample set, uploading the preprocessed historical energy data to the distributed energy management platform and analyzing the preprocessed historical energy data after the processing is completed, and meanwhile, determining the distributed energy management area of the power grid based on the analyzed historical energy data. According to the method, the user power consumption data is predicted through the neural network, distributed energy scheduling is predicted through a deep learning mode based on the user power consumption data obtained through prediction, the distributed energy scheduling result is obtained, energy scheduling execution is performed based on the obtained distributed energy scheduling result, and the real-time performance of distributed energy management is improved.

Description

Technical Field

[0001] The present invention relates to the field of energy management technology, and in particular to a distributed energy management platform for a power grid based on cloud computing. Background Art

[0002] With the continuous advancement of technology, the rise of renewable energy generation has significantly increased the intermittent and random nature of energy management problems. Therefore, an optimal energy management technology is needed to achieve a high level of system reliability and operational efficiency.

[0003] Prior art, such as the invention patent application with publication number CN112801379A, discloses a smart grid distributed energy management system based on cloud computing and big data. The method includes: processing the acquired big data using a data feature module to obtain characteristic information that facilitates analysis. A load forecasting module predicts this characteristic information and outputs the total power required for the current region. A resource allocation module initially allocates power according to an optimal allocation algorithm. An optimal convergence module then adjusts the allocated power to achieve the optimal point. Finally, a power resource allocation module dynamically allocates power based on the allocation results in real time.

[0004] From the above scheme, it can be seen that the current way of distributing electricity is often adjusted according to the regional electricity load, which lacks predictability and has certain limitations. Summary of the Invention

[0005] The purpose of the present invention is to provide a cloud computing-based distributed energy management platform for power grids, which solves the intermittent and random problems of energy management problems in the background technology.

[0006] To solve the above technical problems, the present invention adopts the following technical solution: The present invention provides a cloud computing-based distributed energy management platform for power grids, which specifically includes the following steps:

[0007] S1. Setting distributed energy management nodes based on the distributed energy management area of ​​the power grid, and constructing and saving the network topology based on the locations of the set distributed energy management nodes;

[0008] S2. Based on the constructed network topology, historical energy data in the distributed energy management area of ​​the power grid is collected, and a standard sample set is constructed. The historical energy data in the standard sample set is preprocessed to obtain preprocessed historical energy data.

[0009] The historical energy data includes: historical energy dispatch data and user historical electricity consumption data;

[0010] S3. Upload the pre-processed historical energy data to the distributed energy management platform and analyze it to obtain the analyzed historical energy data;

[0011] S31. Average computing power analysis of distributed energy management platform;

[0012] S32, Energy usage analysis;

[0013] S4. Based on the analyzed historical energy data, predict the user's electricity consumption data through a neural network;

[0014] S5. Based on the predicted user electricity consumption data, the distributed energy scheduling is predicted by deep learning to obtain the distributed energy scheduling result;

[0015] S6. Perform energy scheduling based on the obtained distributed energy scheduling results.

[0016] Preferably, the setting of distributed energy management nodes based on the distributed energy management area of ​​the power grid, and the construction and storage of the network topology based on the set positions of the distributed energy management nodes include the following steps:

[0017] Set the center point coordinates, and number and locate all set distributed energy management nodes based on the center point coordinates;

[0018] Set a triple to save the established network topology;

[0019] M=(A,N,D)

[0020] Where A represents the serial number of the distributed energy management node; N represents the coordinates of the corresponding serial number distributed energy management node; D represents the distance between the current distributed energy management node and other adjacent distributed energy management nodes in the established network topology;

[0021] The adjacent distributed energy management node is a monitoring point directly connected to the current distributed energy management node.

[0022] Preferably, the process of collecting historical energy data within the distributed energy management area of ​​the power grid based on the constructed network topology, constructing a standard sample set, and preprocessing the historical energy data in the standard sample set to obtain the preprocessed historical energy data comprises the following steps:

[0023] Set the user's historical electricity consumption data collected by the distributed energy management node to b a :

[0024]

[0025] Among them, n a Expressed as the number of types of electrical equipment on the ath power line, n e Expressed as the number of class e electrical equipment on the ath power line, c iRepresents the collected data of the i-th type of electrical equipment, d j Indicates the data collected from the jth electrical equipment in category i, b a It represents the historical electricity consumption data of users collected on the a-th power line;

[0026] Summarize the historical electricity consumption data collected from each power line and construct a sample set B = {b1, b2, ..., b a};

[0027] Where b1 represents the historical energy data collected on the first power line in the sample set;

[0028] Assume that the historical energy dispatch data collected by the distributed energy management node is g;

[0029]

[0030] Where G is the energy stored in the distributed energy management node;

[0031] The data standardization formula is as follows:

[0032]

[0033] Where X represents the historical energy data before standardization. Represents the historical energy data after standardization; X max Indicates the maximum value in historical energy data, X min Indicates the minimum value in historical energy data.

[0034] Preferably, the average computing power analysis of the distributed energy management platform includes the following steps:

[0035] Average computing power data = processing time of a set of historical energy data × the number of historical energy data after pre-processing;

[0036] Furthermore, the platform processing time is set based on the calculated average computing power data, and the amount of data processed by the platform is dynamically adjusted according to the set platform processing time.

[0037] Preferably, the energy usage analysis comprises the following steps:

[0038] Summarize the pre-processed historical energy data and construct an energy usage analysis model through Gaussian fitting;

[0039] The energy usage analysis model is as follows:

[0040]

[0041] Where n represents the number of historical energy data after preprocessing, bη represents the user's historical electricity consumption data in the nth group of pre-processed historical energy data; p η represents the average value of the historical energy data after preprocessing of the ηth group, k η represents the historical energy dispatch data in the nth group of pre-processed historical energy data, represents the fitted energy usage analysis model, It represents the average value of the user's historical electricity consumption data, and e is a natural constant;

[0042] The fitted energy usage analysis model and the preprocessed historical energy data are set as the analyzed historical energy data.

[0043] Preferably, the method of predicting user electricity consumption data through a neural network based on the analyzed historical energy data includes the following steps:

[0044] Based on the analyzed historical energy data, the energy usage analysis model is trained through the BP neural network to predict user electricity consumption data;

[0045] The structure of the BP neural network is set to include: input layer, hidden layer and output layer;

[0046] Set the input layer to be Y=[y1,y2,...,y j ,...,y m ], where y1 represents the first set of pre-processed historical energy data input, y m The output of the output layer is Z = [z1,z2,...,z r ,...,z s ], where z1 represents the first set of predicted user electricity consumption data output, z s Represents the output of the sth group of predicted user electricity consumption data;

[0047] The training process of the BP neural network is set to include two stages: the forward propagation of neural signals and the back propagation of errors;

[0048] S41, forward propagation;

[0049] S42. Back propagation of error.

[0050] Preferably, the forward propagation comprises the following steps:

[0051] The calculation formula for the forward propagation phase of the neural signal is as follows:

[0052] The calculation formula for transmitting historical energy data from the input layer to the hidden layer is:

[0053]

[0054] Among them, α h represents the input of the hth hidden layer neuron, v jh represents the weight of the j-th input to the h-th hidden layer neuron, m represents the number of input data in the input layer; y j represents the jth input of the input layer; l h represents the bias value of the hth hidden layer neuron;

[0055] The calculation formula for transmitting historical energy data from the hidden layer to the output layer is:

[0056]

[0057] Among them, β r represents the input of the rth output layer neuron, w hr represents the weight from the hth hidden layer neuron to the rth output layer neuron, q represents the number of hidden layer neurons; u h represents the output of the hth hidden layer neuron; l r Represents the bias value of the rth output layer neuron;

[0058] After receiving the historical energy data transmitted by the hidden layer, the output layer outputs the historical energy data received from the hidden layer through the energy usage analysis model;

[0059] Right now, z r It represents the output of the r-th output layer neuron and represents the r-th group of predicted historical energy data.

[0060] Preferably, the back propagation of the error comprises the following steps:

[0061] Calculate the error between the output layer and the expected value in real time, set the error threshold between the output layer and the expected value, and when the error between the output layer and the expected value is greater than or equal to the set error threshold, adjust the weight from the hidden layer to the output layer and the weight from the input layer to the hidden layer in turn through the back propagation stage of the error;

[0062] The error calculation formula is as follows:

[0063]

[0064] Where E is the error, T r is the expected value of the rth output, s is the number of output neurons;

[0065] The calculation formula for weight adjustment is as follows:

[0066] Δw=(λ)Ez;

[0067] Among them, Δw represents the weight adjustment value, λ represents the learning rate, and z represents the output of the output layer;

[0068] The BP neural network after weight adjustment is set as the trained energy usage analysis model, and the output result is the predicted user electricity consumption data.

[0069] Preferably, the method of predicting distributed energy scheduling based on the predicted user electricity consumption data by deep learning to obtain the distributed energy scheduling result includes the following steps:

[0070] Set Q as the predicted user electricity consumption data set, and set F as the energy scheduling action set;

[0071] The probability of executing distributed energy dispatch actions based on the predicted user electricity consumption data is as follows:

[0072] Q(Q t ,F t ,θ)≈Q * (Q t ,F t );

[0073] The loss function is as follows:

[0074]

[0075] Among them, Q t represents the user's electricity consumption data at time t, Q t+1 User electricity consumption data at time t+1, F t represents the user's electricity consumption data at time t, F t+1 The user's electricity consumption data at time t+1, θ represents the neural network parameters, T represents the expected value, L t (θ t ) represents the loss function of the neural network parameters at time t, γ1 represents the reward function, Indicates action parameters;

[0076] Based on the continuous iteration of the loss function, the probability of distributed energy scheduling action execution is:

[0077]

[0078] Among them, δ t represents the probability of executing the distributed energy dispatch action at time t.

[0079] The present invention also discloses a cloud computing-based distributed energy management platform for a power grid, which is used to implement a cloud computing-based distributed energy management method for a power grid. The platform includes: a data collection module, a data processing module, a data analysis module, a data prediction module, and a management execution module;

[0080] The data collection module is used to collect energy data in the distributed energy management area of ​​the power grid in real time;

[0081] The data processing module is used to process the energy data collected in real time within the distributed energy management area of ​​the power grid;

[0082] The data analysis module is used to analyze the processed energy data and build an energy usage analysis model;

[0083] The data prediction module is used to train the energy usage analysis model through a neural network and perform data prediction based on the trained energy usage analysis model;

[0084] The management execution module is used to perform energy management scheduling based on data prediction.

[0085] The beneficial effects of the present invention are:

[0086] (1) The present invention sets distributed energy management nodes in the distributed energy management area of ​​the power grid, and constructs a network topology based on the location of the set distributed energy management nodes. At the same time, based on the constructed network topology, historical energy data in the distributed energy management area of ​​the power grid is collected, and a standard sample set is constructed and preprocessed. After the processing is completed, the preprocessed historical energy data is uploaded to the distributed energy management platform and analyzed. At the same time, based on the analyzed historical energy data, the user electricity consumption data is predicted through a neural network, and then based on the predicted user electricity consumption data, the distributed energy scheduling is predicted through a deep learning method to obtain a distributed energy scheduling result, and energy scheduling is executed based on the obtained distributed energy scheduling result, thereby improving the real-time performance of distributed energy management.

[0087] (2) The present invention improves the accuracy of data by setting distributed energy management nodes to collect historical user electricity usage data and historical energy scheduling data, and processing the collected historical user electricity usage data and historical energy scheduling data through data processing methods.

[0088] (3) This invention improves the processing efficiency of the energy management platform and the accuracy of energy data by analyzing the average computing power and energy usage of the distributed energy management platform.

[0089] (4) The present invention predicts user electricity consumption data through a neural network based on the analyzed historical energy data, and makes accurate predictions through the forward propagation stage and error reverse control stage of the BP neural network, thereby improving the reliability of energy data prediction.

[0090] (5) The present invention uses a digital twin approach to predict the probability of distributed energy scheduling action execution based on the predicted user electricity consumption data, thereby improving the real-time performance of energy management. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0092] Figure 1 This is a flow chart of the cloud computing distributed energy management method for power grids according to the present invention. DETAILED DESCRIPTION

[0093] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0094] In a specific embodiment of the present invention,

[0095] Reference Figure 1 As shown, the present invention provides a cloud computing-based distributed energy management platform for a power grid, comprising the following steps:

[0096] S1. Setting distributed energy management nodes based on the distributed energy management area of ​​the power grid, and constructing and saving the network topology based on the locations of the set distributed energy management nodes;

[0097] S2. Based on the constructed network topology, historical energy data in the distributed energy management area of ​​the power grid is collected, and a standard sample set is constructed. The historical energy data in the standard sample set is preprocessed to obtain preprocessed historical energy data.

[0098] The historical energy data includes: historical energy dispatch data and user historical electricity consumption data;

[0099] S3. Upload the pre-processed historical energy data to the distributed energy management platform and analyze it to obtain the analyzed historical energy data;

[0100] S31. Average computing power analysis of distributed energy management platform;

[0101] S32, Energy usage analysis;

[0102] S4. Based on the analyzed historical energy data, predict the user's electricity consumption data through a neural network;

[0103] S5. Based on the predicted user electricity consumption data, the distributed energy scheduling is predicted by deep learning to obtain the distributed energy scheduling result;

[0104] S6. Perform energy dispatch based on the obtained distributed energy dispatch result;

[0105] Further, refer to Figure 1 As shown, the distributed energy management nodes are set based on the distributed energy management area of ​​the power grid, and based on the locations of the set distributed energy management nodes, the network topology is constructed and saved, including the following steps:

[0106] Set the center point coordinates, and number and locate all set distributed energy management nodes based on the center point coordinates;

[0107] Furthermore, a triplet is set to save the established network topology;

[0108] M=(A,N,D)

[0109] Where A represents the serial number of the distributed energy management node; N represents the coordinates of the corresponding serial number distributed energy management node; D represents the distance between the current distributed energy management node and other adjacent distributed energy management nodes in the established network topology;

[0110] The adjacent distributed energy management node is a monitoring point directly connected to the current distributed energy management node;

[0111] Further, refer to Figure 1 As shown, based on the constructed network topology, historical energy data in the distributed energy management area of ​​the power grid is collected, a standard sample set is constructed, and the historical energy data in the standard sample set is preprocessed to obtain the preprocessed historical energy data, including the following steps:

[0112] Set the user's historical electricity consumption data collected by the distributed energy management node to b a :

[0113]

[0114] Among them, n a Expressed as the number of types of electrical equipment on the ath power line, n e Expressed as the number of class e electrical equipment on the ath power line, c i Represents the collected data of the i-th type of electrical equipment, d j Indicates the data collected from the jth electrical equipment in category i, ba It represents the historical electricity consumption data of users collected on the a-th power line;

[0115] Summarize the historical electricity consumption data collected from each power line and construct a sample set B = {b1, b2, ..., b a};

[0116] Where b1 represents the historical energy data collected on the first power line in the sample set;

[0117] Assume that the historical energy dispatch data collected by the distributed energy management node is g;

[0118]

[0119] Where G is the energy stored in the distributed energy management node;

[0120] The data standardization formula is as follows:

[0121]

[0122] Where X represents the historical energy data before standardization. Represents the historical energy data after standardization; X max Indicates the maximum value in historical energy data, X min Indicates the minimum value in historical energy data;

[0123] Further, refer to Figure 1 As shown in the figure, the average computing power analysis of the distributed energy management platform includes the following steps:

[0124] Average computing power data = processing time of a set of historical energy data × the number of historical energy data after pre-processing;

[0125] Furthermore, the platform processing time is set based on the calculated average computing power data, and the amount of data processed by the platform is dynamically adjusted according to the set platform processing time;

[0126] Further, refer to Figure 1 As shown, energy use analysis includes the following steps:

[0127] Summarize the pre-processed historical energy data and construct an energy usage analysis model through Gaussian fitting;

[0128] The energy usage analysis model is as follows:

[0129]

[0130] Where n represents the number of historical energy data after preprocessing, b ηrepresents the user's historical electricity consumption data in the nth group of pre-processed historical energy data; p η represents the average value of the historical energy data after preprocessing of the ηth group, k η represents the historical energy dispatch data in the nth group of pre-processed historical energy data, represents the fitted energy usage analysis model, It represents the average value of the user's historical electricity consumption data, and e is a natural constant;

[0131] Furthermore, the fitted energy usage analysis model and the pre-processed historical energy data are set as the analyzed historical energy data;

[0132] Further, refer to Figure 1 As shown, based on the analyzed historical energy data, predicting user electricity consumption data through a neural network includes the following steps:

[0133] Based on the analyzed historical energy data, the energy usage analysis model is trained through the BP neural network to predict user electricity consumption data;

[0134] The structure of the BP neural network is set to include: input layer, hidden layer and output layer;

[0135] Set the input layer to be Y=[y1,y2,...,y j ,...,y m ], where y1 represents the first set of pre-processed historical energy data input, y m The output of the output layer is Z = [z1,z2,...,z r ,...,z s ], where z1 represents the first set of predicted user electricity consumption data output, z s Represents the output of the sth group of predicted user electricity consumption data;

[0136] Furthermore, the training process of the BP neural network is set to include two stages: the forward propagation of neural signals and the back propagation of errors;

[0137] The forward propagation phase of the neural signal indicates that historical energy data is input from the input layer, passes through the hidden layer, and finally reaches the output layer;

[0138] The back propagation phase of the error represents: historical energy data goes from the output layer to the hidden layer and finally to the input layer, and the weights from the hidden layer to the output layer and from the input layer to the hidden layer are adjusted in turn;

[0139] S41, forward propagation;

[0140] The calculation formula for the forward propagation phase of the neural signal is as follows:

[0141] The calculation formula for transmitting historical energy data from the input layer to the hidden layer is:

[0142]

[0143] Among them, α h represents the input of the hth hidden layer neuron, v jh represents the weight of the j-th input to the h-th hidden layer neuron, m represents the number of input data in the input layer; y j represents the jth input of the input layer; l h represents the bias value of the hth hidden layer neuron;

[0144] The calculation formula for transmitting historical energy data from the hidden layer to the output layer is:

[0145]

[0146] Among them, β r represents the input of the rth output layer neuron, w hr represents the weight from the hth hidden layer neuron to the rth output layer neuron, q represents the number of hidden layer neurons; u h represents the output of the hth hidden layer neuron; l r Represents the bias value of the rth output layer neuron;

[0147] After receiving the historical energy data transmitted by the hidden layer, the output layer outputs the historical energy data received from the hidden layer through the energy usage analysis model;

[0148] Right now, z r represents the output of the r-th output layer neuron, which represents the r-th group of predicted historical energy data;

[0149] S42. Back propagation of error

[0150] Furthermore, the error between the output layer and the expected value is calculated in real time, and a threshold value of the error between the output layer and the expected value is set. When the error between the output layer and the expected value is greater than or equal to the set threshold value, the weights from the hidden layer to the output layer and the weights from the input layer to the hidden layer are adjusted in sequence through the back propagation stage of the error.

[0151] The error calculation formula is as follows:

[0152]

[0153] Where E is the error, T r is the expected value of the rth output, s is the number of output neurons;

[0154] The calculation formula for weight adjustment is as follows:

[0155] Δw=(λ)Ez;

[0156] Among them, Δw represents the weight adjustment value, λ represents the learning rate, and z represents the output of the output layer;

[0157] Furthermore, the BP neural network after weight adjustment is set as the trained energy usage analysis model, and the output result is the predicted user electricity consumption data;

[0158] Further, refer to Figure 1 As shown, based on the predicted user electricity consumption data, distributed energy scheduling is predicted through deep learning, and the distributed energy scheduling results are obtained, which include the following steps:

[0159] Set Q as the predicted user electricity consumption data set, and set F as the energy scheduling action set;

[0160] Furthermore, the probability of executing distributed energy dispatch actions is predicted based on the predicted user electricity consumption data as follows:

[0161] Q(Q t ,F t ,θ)≈Q * (Q t ,F t );

[0162] The loss function is as follows:

[0163]

[0164] Among them, Q t represents the user's electricity consumption data at time t, Q t+1 User electricity consumption data at time t+1, F t represents the user's electricity consumption data at time t, F t+1 The user's electricity consumption data at time t+1, θ represents the neural network parameters, T represents the expected value, L t (θ t ) represents the loss function of the neural network parameters at time t, γ1 represents the reward function, Indicates action parameters;

[0165] Based on the continuous iteration of the loss function, the probability of distributed energy scheduling action execution is:

[0166]

[0167] Among them, δ t represents the probability of executing distributed energy dispatch action at time t;

[0168] In a specific embodiment, the cloud computing-based grid distributed energy management platform is used to implement a cloud computing-based grid distributed energy management method, and the platform includes: a data collection module, a data processing module, a data analysis module, a data prediction module, and a management execution module;

[0169] The data collection module is used to collect energy data in the distributed energy management area of ​​the power grid in real time;

[0170] The data processing module is used to process the energy data collected in real time within the distributed energy management area of ​​the power grid;

[0171] The data analysis module is used to analyze the processed energy data and build an energy usage analysis model;

[0172] The data prediction module is used to train the energy usage analysis model through a neural network and perform data prediction based on the trained energy usage analysis model;

[0173] The management execution module is used to perform energy management scheduling based on data prediction.

[0174] It should be noted that

[0175] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should all fall within the scope of protection of the present invention.

Claims

1. A cloud computing-based distributed energy management platform for power grids, characterized by: The following steps are involved: S1. Setting distributed energy management nodes based on the distributed energy management area of ​​the power grid, and constructing and saving the network topology based on the locations of the set distributed energy management nodes; S2. Based on the constructed network topology, historical energy data in the distributed energy management area of ​​the power grid is collected, and a standard sample set is constructed. The historical energy data in the standard sample set is preprocessed to obtain preprocessed historical energy data. The historical energy data includes: historical energy dispatch data and user historical electricity consumption data; S3. Upload the pre-processed historical energy data to the distributed energy management platform and analyze it to obtain the analyzed historical energy data; S31. Average computing power analysis of distributed energy management platform; S32, Energy usage analysis; S4. Based on the analyzed historical energy data, predict the user's electricity consumption data through a neural network; S5. Based on the predicted user electricity consumption data, the distributed energy scheduling is predicted by deep learning to obtain the distributed energy scheduling result; S6. Perform energy scheduling based on the obtained distributed energy scheduling results.

2. The cloud computing-based distributed energy management platform for power grids according to claim 1, characterized in that: The method of setting a distributed energy management node based on the distributed energy management area of ​​the power grid, and constructing and saving a network topology based on the location of the set distributed energy management node comprises the following steps: Set the center point coordinates, and number and locate all set distributed energy management nodes based on the center point coordinates; Set a triple to save the established network topology; M=(A,N,D) Where A represents the serial number of the distributed energy management node; N represents the coordinates of the corresponding serial number distributed energy management node; D represents the distance between the current distributed energy management node and other adjacent distributed energy management nodes in the established network topology; The adjacent distributed energy management node is a monitoring point directly connected to the current distributed energy management node.

3. The cloud computing-based distributed energy management platform for power grids according to claim 1, characterized in that: The method of collecting historical energy data in the distributed energy management area of ​​the power grid based on the constructed network topology, constructing a standard sample set, and preprocessing the historical energy data in the standard sample set to obtain the preprocessed historical energy data includes the following steps: Set the user's historical electricity consumption data collected by the distributed energy management node to b a : Among them, n a Expressed as the number of types of electrical equipment on the ath power line, n e Expressed as the number of class e electrical equipment on the ath power line, c i Represents the collected data of the i-th type of electrical equipment, d j Indicates the data collected from the jth electrical equipment in category i, b a It represents the historical electricity consumption data of users collected on the a-th power line; Summarize the historical electricity consumption data collected from each power line and construct a sample set B = {b1, b2, ..., b a }; Where b1 represents the historical energy data collected on the first power line in the sample set; Assume that the historical energy dispatch data collected by the distributed energy management node is g; Where G is the energy stored in the distributed energy management node; The data standardization formula is as follows: Where X represents the historical energy data before standardization. Represents the historical energy data after standardization; X max Indicates the maximum value in historical energy data, X min Indicates the minimum value in historical energy data.

4. The cloud computing-based distributed energy management platform for power grids according to claim 1, characterized in that: The average computing power analysis of the distributed energy management platform includes the following steps: Average computing power data = processing time of a set of historical energy data × the number of historical energy data after pre-processing; The platform processing time is set based on the calculated average computing power data, and the amount of data processed by the platform is dynamically adjusted according to the set platform processing time.

5. The cloud computing-based distributed energy management platform for power grids according to claim 1, characterized in that: The energy usage analysis includes the following steps: Summarize the pre-processed historical energy data and construct an energy usage analysis model through Gaussian fitting; The energy usage analysis model is as follows: Where n represents the number of historical energy data after preprocessing, b η represents the user's historical electricity consumption data in the nth group of pre-processed historical energy data; p η represents the average value of the historical energy data after preprocessing of the ηth group, k η represents the historical energy dispatch data in the nth group of pre-processed historical energy data, represents the fitted energy usage analysis model, Represents the average value of the user's historical electricity consumption data, and e is a natural constant; The fitted energy usage analysis model and the preprocessed historical energy data are set as the analyzed historical energy data.

6. The cloud computing-based distributed energy management platform for power grids according to claim 1, characterized in that: The method of predicting user electricity consumption data based on the analyzed historical energy data through a neural network includes the following steps: Based on the analyzed historical energy data, the energy usage analysis model is trained through the BP neural network to predict user electricity consumption data; The structure of the BP neural network is set to include: input layer, hidden layer and output layer; Set the input layer to be Y=[y1,y2,...,y j ,...,y m ], where y1 represents the first set of pre-processed historical energy data input, y m The output of the output layer is Z = [z1,z2,...,z r ,...,z s ], where z1 represents the first set of predicted user electricity consumption data output, z s Represents the output of the sth group of predicted user electricity consumption data; The training process of the BP neural network is set to include two stages: the forward propagation of neural signals and the back propagation of errors; S41, forward propagation; S42. Back propagation of error.

7. The cloud computing-based distributed energy management platform for power grids according to claim 6, characterized in that: The forward propagation includes the following steps: The calculation formula for the forward propagation stage of the neural signal is as follows: The calculation formula for transmitting historical energy data from the input layer to the hidden layer is: Among them, α h represents the input of the hth hidden layer neuron, v jh represents the weight of the j-th input to the h-th hidden layer neuron, m represents the number of input data in the input layer; y j represents the jth input of the input layer; l h represents the bias value of the hth hidden layer neuron; The calculation formula for transmitting historical energy data from the hidden layer to the output layer is: Among them, β r represents the input of the rth output layer neuron, w hr represents the weight from the hth hidden layer neuron to the rth output layer neuron, q represents the number of hidden layer neurons; u h represents the output of the hth hidden layer neuron; l r Represents the bias value of the rth output layer neuron; After receiving the historical energy data transmitted by the hidden layer, the output layer outputs the historical energy data received from the hidden layer through the energy usage analysis model; Right now, It represents the output of the r-th output layer neuron and represents the r-th group of predicted historical energy data.

8. The cloud computing-based distributed energy management platform for power grids according to claim 6, characterized in that: The back propagation of the error includes the following steps: Calculate the error between the output layer and the expected value in real time, set the error threshold between the output layer and the expected value, and when the error between the output layer and the expected value is greater than or equal to the set error threshold, adjust the weight from the hidden layer to the output layer and the weight from the input layer to the hidden layer in turn through the back propagation stage of the error; The error calculation formula is as follows: Where E is the error, T r is the expected value of the rth output, s is the number of output neurons; The calculation formula for weight adjustment is as follows: Δw=(λ)Ez; Among them, Δw represents the weight adjustment value, λ represents the learning rate, and z represents the output of the output layer; The BP neural network after weight adjustment is set as the trained energy usage analysis model, and the output result is the predicted user electricity consumption data.

9. The cloud computing-based distributed energy management platform for power grids according to claim 1, characterized in that: The method of predicting distributed energy scheduling based on the predicted user electricity consumption data by deep learning to obtain the distributed energy scheduling result includes the following steps: Set Q as the predicted user electricity consumption data set, and set F as the energy scheduling action set; The probability of executing distributed energy dispatch actions based on the predicted user electricity consumption data is as follows: Q(Q t ,F t ,θ)≈Q * (Q t ,F t ); The loss function is as follows: Among them, Q t represents the user's electricity consumption data at time t, Q t+1 User electricity consumption data at time t+1, F t represents the user's electricity consumption data at time t, F t+1 The user's electricity consumption data at time t+1, θ represents the neural network parameters, T represents the expected value, L t (θ t ) represents the loss function of the neural network parameters at time t, γ1 represents the reward function, Indicates action parameters; Based on the continuous iteration of the loss function, the probability of distributed energy scheduling action execution is: Among them, δ t represents the probability of executing the distributed energy dispatch action at time t.

10. A cloud computing-based distributed energy management platform for a power grid according to any one of claims 1 to 9, characterized in that: The platform includes: a data collection module, a data processing module, a data analysis module, a data prediction module and a management execution module; The data collection module is used to collect energy data in the distributed energy management area of ​​the power grid in real time; The data processing module is used to process the energy data collected in real time within the distributed energy management area of ​​the power grid; The data analysis module is used to analyze the processed energy data and build an energy usage analysis model; The data prediction module is used to train the energy usage analysis model through a neural network and perform data prediction based on the trained energy usage analysis model; The management execution module is used to perform energy management scheduling based on data prediction.

Images