Power grid carbon flow simulation analysis method driven by NLP large model of power system

Abstract

The invention discloses a power system NLP large model-driven power grid carbon flow simulation analysis method, which comprises the following steps of: collecting generating capacity data and power consumption data of a transformer substation by using an NLP large model technology, constructing a neural network model, and performing feature engineering processing on the collected generating capacity data and power consumption data of the transformer substation by the neural network model to obtain a power grid carbon flow simulation analysis result. Training a neural network model according to a relationship between learning features and data, constructing a power grid carbon flow simulation model based on a training result, covering generation, transmission and consumption links, and performing simulation calculation by using the model to obtain a carbon emission condition of a power grid. The average emission factor of the power grid is calculated according to the carbon emission condition of the power grid to evaluate the carbon emission condition of the power grid, so that the carbon emission condition of the power grid can be comprehensively understood while the efficiency and accuracy of data processing and analysis are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power simulation analysis, and in particular to a power grid carbon flow simulation analysis method driven by a large NLP model of a power system. Background Art

[0002] With the severe challenges of global climate change, the low-carbon transformation of power systems has become an industry consensus. However, traditional power grid carbon flow analysis methods have certain limitations and it is difficult to fully consider the complexity and dynamics of the power system, resulting in low efficiency and limited accuracy of the analysis results. At the same time, the existing power system simulation technology lacks in-depth analysis and optimization suggestions for power grid carbon flow, making it difficult to meet the needs of low-carbon operation of the power system.

[0003] The Chinese patent publication number: CN 106055730 B discloses a method for carbon flow simulation and analysis of a power system suitable for a large-scale power grid. The core technical solution is to simulate the carbon flow of the power grid through the power grid carbon flow simulation and analysis method, so as to optimize and adjust the power grid. The method includes the steps of power grid carbon flow simulation, power grid carbon flow analysis and node carbon potential calculation. The power grid carbon flow simulation refers to simulating all nodes in the power grid, including generators, loads and DC systems, to determine the carbon flow at each node in the power grid. The power grid carbon flow analysis is carried out on the basis of the power grid carbon flow simulation. The performance of the power grid is evaluated through the node carbon potential, branch carbon flow rate, branch network loss carbon flow rate, power generation carbon flow rate and load carbon flow rate indicators. The node carbon potential is calculated by numbering the branch connected by the node with the power flowing into the node as a standard to obtain the node carbon potential vector. Finally, according to these indicators, other indicators of the power grid carbon flow simulation analysis can be calculated, including the total power generation-total load-total network loss carbon flow rate, the average node carbon potential of the partition, the average node carbon potential of the voltage level, the section carbon flow rate and the section definition.

[0004] When processing complex and dynamically changing power system data, existing technologies fail to make full use of data processing and machine learning technologies, and fail to use natural language processing or neural network technologies to improve model performance, resulting in low analysis efficiency and affected accuracy. In addition, they fail to conduct in-depth analysis of the specific situation of carbon flow in the power grid and are unable to provide detailed data support for the low-carbon operation of the power system.

[0005] The present invention proposes a solution to the above shortcomings: by adopting NLP big model technology to process substation text data, extracting power generation data and power consumption data to build a neural network model, using recurrent neural network to train the model, and combining feature engineering and outlier processing, a simulation model is constructed to deeply calculate the carbon emissions of each node in the power grid, and at the same time, the average emission factor of the power grid is calculated based on the carbon emissions of each node to evaluate the carbon emissions of the power grid. While effectively improving the efficiency and accuracy of data processing and analysis, it can more comprehensively understand the carbon emissions of the power grid. Summary of the invention

[0006] In order to solve the above-mentioned technical problems, the present invention provides a power grid carbon flow simulation analysis method driven by a large NLP model of a power system.

[0007] The technical solution of the present invention is implemented as follows: a power grid carbon flow simulation analysis method driven by a large NLP model of a power system, comprising:

[0008] S1. Collect power generation data and power consumption data of the substation, and build a neural network model based on the collected power generation data and power consumption data of the substation;

[0009] S2. The neural network model performs feature engineering processing on the collected power generation data and power consumption data of the substation;

[0010] S3. Train the neural network model based on the relationship between the learning features and the data to obtain the output results;

[0011] S4. Based on the output results of the neural network model, a power grid carbon flow simulation model is constructed, and the simulation model includes the generation link, transmission link and consumption link of the power grid carbon flow;

[0012] S5. Use the simulation model to simulate the carbon flow of the power grid, obtain the carbon emissions of each node in the power grid, calculate the carbon emissions of the total nodes in the power grid, calculate the average emission factor of the power grid based on the calculated carbon emissions, and evaluate the carbon emissions of the power grid based on the average emission factor of the power grid.

[0013] Furthermore, in the step S1, the power generation data and power consumption data of the substation are collected, and a neural network model is constructed based on the collected power generation data and power consumption data of the substation:

[0014] The NLP big model technology is used to extract the substation data from the substation text data, including the power generation and power consumption, and the neural network model is constructed by using a recurrent neural network to construct the neural network model;

[0015] The power generation data P and power consumption data Q of the substation extracted from the substation text data are represented by a data sequence T1, where T1 = {(P1, Q1), (P2, Q2), ... (P n , Q n )}.

[0016] Furthermore, the neural network model in step S2 performs feature engineering processing on the collected power generation data and power consumption data of the substation, including the power generation and power consumption of the substation:

[0017] Performing outlier processing on the collected data, detecting and processing outliers in the substation data extracted by the NLP big model;

[0018] Furthermore, outliers are handled as follows:

[0019] Perform outlier processing on the collected data, clean the collected data, including the power generation data and power consumption data of the substation, calculate the power generation data and power consumption data using the z-score function, set the outlier threshold interval, and compare the output value when it is not within the outlier threshold interval. Take a processing method to process the outlier;

[0020] The collected data is processed for outliers, and the collected data, including power generation data and power consumption data, is cleaned. The power generation data and power consumption data are calculated using the z-score function. The calculation formula is:

[0021] Power generation data:

[0022]

[0023] Among them, z1 is the standardized value of the power generation data, x1 is the original value of the power generation data, μ1 is the average value of the power generation data, and σ1 is the standard deviation of the power generation data;

[0024] Power consumption data:

[0025]

[0026] Among them, z2 is the standardized value of electricity consumption data, x2 is the original value of electricity consumption data, μ2 is the average value of electricity consumption data, and σ2 is the standard deviation of electricity consumption data;

[0027] Define |Z|>3 as an outlier, then the normal range of Z value is (-3, 3). When the output Z value is not in the normal range, it is an outlier. The outlier is processed, including deleting or replacing the outlier;

[0028] After processing the outliers, we get the data sequence T2, where T2 = {(P1', Q1'), (P2', Q1'), ... (P n ', Q n ')}.

[0029] Furthermore, in the step S3, the neural network model is trained according to the relationship between the learning features and the data to obtain the output result:

[0030] The power generation data and power consumption data of the substation after outlier processing are input into the constructed neural network model as input sequences, and the model is trained using the stochastic gradient descent algorithm;

[0031] Furthermore, the specific steps of the training process are:

[0032] S301, initializing parameters, using random numbers to initialize the weights and biases of the recurrent neural network;

[0033] S302, forward propagation, according to the current weight and bias, forward propagation calculation is performed to obtain the predicted value, the data sequence T2 is input as the input sequence, the hidden layer state is obtained through the recurrent neural network, and the output sequence is calculated based on the calculated hidden layer state;

[0034] The hidden layer state of the recurrent neural network is calculated as:

[0035] h t =σ(Wh t-1 +Ux t +b)

[0036] Among them, h t is the hidden layer state, h t-1 is the hidden layer state of the previous time step, x t is the input sequence, W and U are weight matrices, b is the bias vector, and σ is the activation function;

[0037] The output sequence calculation formula is:

[0038] y t =Vh t +c

[0039] Among them, y t is the output sequence, V is the weight matrix, h t is the hidden layer state, c is the bias vector;

[0040] S303, calculate the loss, use the mean square error to forward propagate the output data y t Compare the predicted value with the input data T2 as the true value and calculate the value of the loss function;

[0041] The formula for calculating the mean square error is:

[0042]

[0043] Among them, T2 is the true value, y t is the model prediction value, n is the number of samples;

[0044] S304, back propagation, calculating the gradient of the loss function with respect to each parameter, starting from the output layer, and back propagating the gradient to the hidden layer and the input layer;

[0045] Calculate the gradient of the weight matrix W, U and bias vector b in the loss function recurrent neural network model. The gradient formula is:

[0046] Weight matrix W:

[0047]

[0048] in, is the gradient of the loss function with respect to the weight matrix W, E is the loss function, and W is the weight matrix;

[0049] Weight matrix U:

[0050]

[0051] in, is the gradient of the loss function with respect to the weight matrix U, E is the loss function, and U is the weight matrix;

[0052] Bias vector b:

[0053]

[0054] in, is the gradient of the loss function with respect to the bias vector b, E is the loss function, and b is the bias vector;

[0055] S305, updating parameters, updating the parameters of the network according to the calculated gradient and learning rate;

[0056] The parameter update settings are:

[0057] Weight matrix W:

[0058]

[0059] Among them, W1 is the updated weight matrix, η is the learning rate, is the calculated gradient;

[0060] Weight matrix U

[0061]

[0062] Among them, U1 is the updated weight matrix, η is the learning rate, is the calculated gradient;

[0063] Bias vector b

[0064]

[0065] Among them, b1 is the updated weight matrix, η is the learning rate, is the calculated gradient;

[0066] S306, repeat the iteration, repeat steps S302 to S305 until the loss function reaches the maximum number of iterations, and then the iteration is completed.

[0067] Furthermore, in step S4, based on the output result of the neural network model, a power grid carbon flow simulation model is constructed, and the simulation model includes the generation link, transmission link and consumption link of the power grid carbon emissions:

[0068] Combined with the model output data trained in step S3, a power grid carbon flow simulation model is constructed. The power grid carbon flow simulation model is calculated by superimposing the carbon emissions in the generation link, the carbon emissions in the transmission link, and the carbon emissions in the consumption link. The specific process is as follows:

[0069] The trained model output data in step S3 is T3 = {(P1", Q1"), (P2", Q2"), ... (P n ”,Q n ”)}, combined with data T3, a power grid carbon flow simulation model F is constructed. The power grid carbon flow simulation model F is calculated by superimposing the carbon emissions F1 in the generation link, the carbon emissions F2 in the transmission link, and the carbon emissions F3 in the consumption link. Specifically, it is:

[0070] Generation stage:

[0071] F1=αP"

[0072] Among them, α is the carbon emission coefficient in the power generation process, P” is the output power generation, and F1 is the carbon emission generated in the generation process;

[0073] Transmission link:

[0074]

[0075] Where β is the loss coefficient during transmission, is the additional carbon emission coefficient in the transmission process, and F2 is the carbon emission generated in the transmission process;

[0076] Consumption link:

[0077] F3=δQ"

[0078] Among them, δ is the loss coefficient generated in the process of electricity consumption, Q” is the output power consumption, and F3 is the carbon emissions generated in the consumption link.

[0079] Furthermore, in step S5, the simulation model is used to simulate and calculate the carbon flow of the power grid, obtain the carbon emissions of each node in the power grid, calculate the carbon emissions of the total nodes in the power grid, calculate the average emission factor of the power grid based on the calculated carbon emissions, and evaluate the carbon emissions of the power grid based on the average emission factor of the power grid:

[0080] The total power generation and node carbon emissions of the entire power grid are calculated, and the total carbon emissions data of the entire power grid are obtained through the calculated total power generation and node carbon emissions of the entire power grid, so as to calculate the average emission factor of the power grid, and the carbon emissions of the power grid are evaluated according to the average emission factor of the power grid. The specific process is as follows:

[0081] In the preset power grid simulation model, the power grid consists of n nodes, where the power generation of each node is P i , the power consumption is Q i , calculate the total power generation of the entire power grid as:

[0082]

[0083] Where n is the number of nodes, P i is the amount of electricity generated in a node;

[0084] Carbon emissions of the node F i Calculated as:

[0085] F i =F1+F2+F3

[0086] Among them, F1 is the carbon emissions generated by the generation phase of the node, F2 is the carbon emissions generated by the transmission phase of the node, and F3 is the carbon emissions generated by the consumption phase of the node;

[0087] The total carbon emissions of the entire power grid are calculated as:

[0088]

[0089] Where n is the number of nodes, F i is the carbon emissions generated in a node;

[0090] By calculating the total carbon emissions and total power generation of the entire power grid, the average emission factor O of the power grid is calculated. The calculation formula is:

[0091]

[0092] Among them, F 总 is the total carbon emissions of the power grid, P 总is the total power generation of the grid;

[0093] Based on the calculated average grid emission factor O, the carbon emission intensity of the regional grid is determined.

[0094] Beneficial effects:

[0095] In view of the shortcoming of the prior art that the accuracy of data cannot be guaranteed during the simulation and analysis process, the present invention utilizes the NLP big model technology to collect the power generation data and power consumption data of the substation, and constructs a neural network model. The neural network model performs feature engineering on the collected power generation data and power consumption data of the substation, trains the neural network model according to the relationship between the learning features and the data, and constructs a power grid carbon flow simulation model based on the training results, covering the generation, transmission and consumption links. The model is used to perform simulation calculations to obtain the carbon emissions of the power grid, and the average emission factor of the power grid is calculated based on the carbon emissions of the power grid to evaluate the carbon emissions of the power grid. While effectively improving the efficiency and accuracy of data processing and analysis, a more comprehensive understanding of the carbon emissions of the power grid can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 It is a structural block diagram of a power grid carbon flow simulation analysis method driven by a large NLP model of a power system in an embodiment of the present invention;

[0097] Figure 2 A flowchart of a power grid carbon flow simulation analysis method driven by a large NLP model of a power system in an embodiment of the present invention;

[0098] Figure 3 The present invention is a flowchart of the steps of training a neural network model of a power grid carbon flow simulation analysis method driven by a large NLP model of a power system in an embodiment of the present invention. DETAILED DESCRIPTION

[0099] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0100] The preferred implementation methods of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these implementation methods are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0101] When used herein, the singular forms "a", "an", and "said / the" may also include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "include / comprise" or "have" and the like specify the presence of stated features, wholes, steps, operations, components, parts, or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof. At the same time, the term "and / or" used in this specification includes any and all combinations of the relevant listed items.

[0102] See also Figure 1-Figure 3 As shown, a power grid carbon flow simulation analysis method driven by a large NLP model of a power system includes:

[0103] S1. Collect power generation data and power consumption data of the substation, and build a neural network model based on the collected power generation data and power consumption data of the substation;

[0104] S2. The neural network model performs feature engineering processing on the collected power generation data and power consumption data of the substation;

[0105] S3. Train the neural network model based on the relationship between the learning features and the data to obtain the output results;

[0106] S4. Based on the output results of the neural network model, a power grid carbon flow simulation model is constructed, and the simulation model includes the generation link, transmission link and consumption link of the power grid carbon emissions;

[0107] S5. Use the simulation model to simulate the carbon flow of the power grid, obtain the carbon emissions of each node in the power grid, calculate the carbon emissions of the total nodes in the power grid, and analyze them.

[0108] like Figure 2 As shown, in step S1, the power generation data and power consumption data of the substation are collected, and a neural network model is constructed based on the collected power generation data and power consumption data of the substation.

[0109] Specifically, the NLP big model technology is used to extract the substation data in the substation text data, including the power generation and power consumption, and the neural network model is constructed by using a recurrent neural network (RNN) to construct the neural network model;

[0110] The power generation data P and power consumption data Q of the substation extracted from the substation text data are represented by a data sequence T1, where T1 = {(P1, Q1), (P2, Q2), ... (P n , Q n )}.

[0111] like Figure 2As shown, the neural network model performs feature engineering processing on the collected power generation data and power consumption data of the substation.

[0112] Specifically, outliers are handled as follows:

[0113] Perform outlier processing on the collected data, clean the collected data, including the power generation data and power consumption data of the substation, calculate the power generation data and power consumption data using the z-score function, set the outlier threshold interval, and compare the output value when it is not within the outlier threshold interval. Take a processing method to process the outlier;

[0114] The collected data is processed for outliers, and the collected data, including power generation data and power consumption data, is cleaned. The power generation data and power consumption data are calculated using the z-score function. The calculation formula is:

[0115] Power generation data:

[0116]

[0117] Among them, z1 is the standardized value of the power generation data, x1 is the original value of the power generation data, μ1 is the average value of the power generation data, and σ1 is the standard deviation of the power generation data;

[0118] Power consumption data:

[0119]

[0120] Among them, z2 is the standardized value of electricity consumption data, x2 is the original value of electricity consumption data, μ2 is the average value of electricity consumption data, and σ2 is the standard deviation of electricity consumption data;

[0121] Define |Z|>3 as an outlier, then the normal range of Z value is (-3, 3). When the output Z value is within the normal range, it is an outlier. The outlier is processed, including deleting or replacing the outlier.

[0122] After processing the outliers, we get the data sequence T2, where T2 = {(P1', Q1'), (P2', Q1'), ... (P n ', Q n ')}.

[0123] like Figure 2 As shown, in step S3, the neural network model is trained according to the relationship between the learning features and the data to obtain the output results.

[0124] Specifically, the power generation data and power consumption data of the substation after outlier processing are input into the constructed neural network model as input sequences, and the model is trained using a stochastic gradient descent algorithm;

[0125] Furthermore, the specific steps of the training process are:

[0126] S301, initialization parameters, using random numbers to initialize the weights and biases of the RNN;

[0127] S302, forward propagation, according to the current weight and bias, forward propagation calculation is performed to obtain the predicted value, the data sequence T2 is input as the input sequence, the hidden layer state is obtained through the RNN, and the output sequence is calculated based on the calculated hidden layer state;

[0128] The RNN hidden layer state is calculated as:

[0129] h t =σ(Wh t-1 +Ux t +b)

[0130] Among them, h t is the hidden layer state, h t-1 is the hidden layer state of the previous time step, x t is the input sequence, W and U are weight matrices, b is the bias vector, and σ is the activation function;

[0131] The output sequence calculation formula is:

[0132] y t =Vh t +c

[0133] Among them, y t is the output sequence, V is the weight matrix, h t is the hidden layer state, c is the bias vector;

[0134] S303, calculate the loss, use the mean square error (MSE) to convert the output data y obtained by forward propagation t Compare the predicted value with the input data T2 as the true value and calculate the value of the loss function;

[0135] The MSE calculation formula is:

[0136]

[0137] Among them, T2 is the true value, y t is the model prediction value, n is the number of samples;

[0138] S304, back propagation, calculating the gradient of the loss function with respect to each parameter, starting from the output layer, and back propagating the gradient to the hidden layer and the input layer;

[0139] Calculate the gradient of the weight matrix W, U and bias vector b in the loss function RNN model. The gradient formula is:

[0140] Weight matrix W:

[0141]

[0142] in, is the gradient of the loss function with respect to the weight matrix W, E is the loss function, and W is the weight matrix;

[0143] Weight matrix U:

[0144]

[0145] in, is the gradient of the loss function with respect to the weight matrix U, E is the loss function, and U is the weight matrix;

[0146] Bias vector b:

[0147]

[0148] in, is the gradient of the loss function with respect to the bias vector b, E is the loss function, and b is the bias vector;

[0149] S305, updating parameters, updating the parameters of the network according to the calculated gradient and learning rate;

[0150] The parameter update settings are:

[0151] Weight matrix W:

[0152]

[0153] Among them, W1 is the updated weight matrix, η is the learning rate, is the calculated gradient;

[0154] Weight matrix U

[0155]

[0156] Among them, U1 is the updated weight matrix, η is the learning rate, is the calculated gradient;

[0157] Bias vector b

[0158]

[0159] Among them, b1 is the updated weight matrix, η is the learning rate, is the calculated gradient;

[0160] S306, repeat the iteration, repeating steps S302 to S305 until the loss function reaches the maximum number of iterations, and the iteration is completed;

[0161] Among them, forward propagation is to calculate the output of the neural network through input data, and then compare it with the true label to calculate the loss function value. Backward propagation is to feed back the currently calculated loss function value to the previous layers of the neural network and adjust the weights of the network parameters of each layer.

[0162] like Figure 2 As shown, in step S4, based on the output results of the neural network model, a power grid carbon flow simulation model is constructed, and the simulation model includes the generation link, transmission link and consumption link of the power grid carbon emission intensity.

[0163] Specifically, combined with the model output data trained in step S3, a power grid carbon flow simulation model is constructed. The power grid carbon flow simulation model is calculated by superimposing the carbon emissions in the generation link, the carbon emissions in the transmission link, and the carbon emissions in the consumption link. The specific process is as follows:

[0164] The trained model output data in step S3 is T3 = {(P1", Q1"), (P2", Q2"), ... (P n ”,Q n ”)}, combined with data T3, a power grid carbon flow simulation model F is constructed. The power grid carbon flow simulation model F is calculated by superimposing the carbon emissions F1 in the generation link, the carbon emissions F2 in the transmission link, and the carbon emissions F3 in the consumption link. Specifically, it is:

[0165] Generation stage:

[0166] F1=αP"

[0167] Among them, α is the carbon emission coefficient in the power generation process, P” is the output power generation, and F1 is the carbon emission generated in the generation process;

[0168] Transmission link:

[0169]

[0170] Where β is the loss coefficient during transmission, is the additional carbon emission coefficient in the transmission process, and F2 is the carbon emission generated in the transmission process;

[0171] Consumption link:

[0172] F3=δQ"

[0173] Among them, δ is the loss coefficient generated in the process of electricity consumption, Q” is the output power consumption, and F3 is the carbon emissions generated in the consumption link.

[0174] like Figure 2 As shown, in step S5, the simulation model is used to simulate the carbon flow of the power grid, the carbon emissions of each node in the power grid are obtained, the carbon emissions of the total nodes in the power grid are calculated, the average emission factor of the power grid is calculated based on the calculated carbon emissions, and the carbon emissions of the power grid are evaluated based on the average emission factor of the power grid.

[0175] Specifically, the total power generation and node carbon emissions of the entire power grid are calculated, and the total carbon emissions data of the entire power grid are obtained through the calculated total power generation and node carbon emissions of the entire power grid, so as to calculate the average emission factor of the power grid, and evaluate the carbon emissions of the power grid according to the average emission factor of the power grid. The specific process is as follows:

[0176] In the preset power grid simulation model, the power grid consists of n nodes, where the power generation of each node is P i , the power consumption is Q i , calculate the total power generation of the entire power grid as:

[0177]

[0178] Where n is the number of nodes, P i is the amount of electricity generated in a node;

[0179] Carbon emissions of the node F i Calculated as:

[0180] F i =F1+F2+F3

[0181] Among them, F1 is the carbon emissions generated by the generation phase of the node, F2 is the carbon emissions generated by the transmission phase of the node, and F3 is the carbon emissions generated by the consumption phase of the node;

[0182] The total carbon emissions of the entire power grid are calculated as:

[0183]

[0184] Where n is the number of nodes, F i is the carbon emissions generated in a node;

[0185] By calculating the total carbon emissions and total power generation of the entire power grid, the average emission factor O of the power grid is calculated. The calculation formula is:

[0186]

[0187] Among them, F 总 is the total carbon emissions of the power grid, P总 is the total power generation of the grid;

[0188] The grid carbon flow is evaluated based on the average grid emission factor O. When the calculated O is greater than the national average of 0.5703t, it indicates that the carbon emission intensity of the regional grid is high, and the simulation staff is notified to handle it.

[0189] Among them, the national average value of the average grid emission factor of 0.5703t comes from the "Notice on Doing a Good Job in the Management of Greenhouse Gas Emission Reports of Power Generation Enterprises from 2023 to 2025" issued by the Ministry of Ecology and Environment of the People's Republic of China, and the unit is CO2 / MWh.

[0190] like Figure 1-Figure 3 As shown, the present invention works as follows in the application scenario of solving the disadvantage that the prior art cannot guarantee the accuracy of data during the simulation analysis process:

[0191] Collect power generation data and power consumption data of the substation, and build a neural network model based on the collected power generation data and power consumption data of the substation;

[0192] The neural network model performs feature engineering processing on the collected power generation data and power consumption data of the substation;

[0193] According to the relationship between learning features and data, the neural network model is trained to obtain output results;

[0194] Based on the output results of the neural network model, a power grid carbon flow simulation model is constructed. The simulation model includes the generation, transmission and consumption of power grid carbon emissions.

[0195] The simulation model is used to simulate the carbon flow of the power grid, obtain the carbon emissions of each node in the power grid, calculate the carbon emissions of the total nodes in the power grid, calculate the average emission factor of the power grid based on the calculated carbon emissions, and evaluate the carbon emissions of the power grid based on the average emission factor of the power grid.

[0196] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0197] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and rules of the present invention shall be included in the protection scope of the present invention.

Claims

1. A power grid carbon flow simulation analysis method driven by a large NLP model of a power system, characterized by: S1. Collect power generation data and power consumption data of the substation, and build a neural network model based on the collected power generation data and power consumption data of the substation; S2. The neural network model performs feature engineering processing on the collected power generation data and power consumption data of the substation; S3. Train the neural network model based on the relationship between the learning features and the data to obtain the output results; S4. Based on the output results of the neural network model, a power grid carbon flow simulation model is constructed, and the simulation model includes the generation link, transmission link and consumption link of the power grid carbon emissions; S5. Use the simulation model to perform model simulation calculations to obtain the carbon emissions of each node in the power grid, calculate the carbon emissions of the total nodes in the power grid, calculate the average emission factor of the power grid based on the calculated carbon emissions, and evaluate the carbon emissions of the power grid based on the average emission factor of the power grid.

2. According to claim 1, the method for carbon flow simulation analysis of power grid driven by NLP large model of power system is characterized by: In step S1, the power generation data and power consumption data of the substation are collected. Based on the collected power generation data and power consumption data of the substation, a neural network model is constructed as follows: The NLP big model technology is used to extract the substation data in the substation text data, including the power generation and power consumption, and the neural network model is constructed by using a recurrent neural network.

3. According to claim 1, a method for simulating and analyzing carbon flow in a power grid driven by a large NLP model of a power system, characterized in that: The neural network model in step S2 performs feature engineering processing on the collected power generation data and power consumption data of the substation as follows: The collected data is processed for outliers, and the collected data is cleaned, including the power generation data and power consumption data of the substation. The power generation data and power consumption data are calculated using the z-score function, and the outlier threshold interval is set. When the output value is not within the outlier threshold interval, the output value is an outlier, and the outlier is processed.

4. According to claim 1, a method for simulating and analyzing carbon flow in a power grid driven by a large NLP model of a power system, characterized in that: In the step S3, the neural network model is trained according to the relationship between the learning features and the data, and the output result is: The power generation data and power consumption data of the substation after outlier processing are input into the constructed neural network model as input sequences, and the model is trained using the stochastic gradient descent algorithm.

5. According to claim 1, a method for simulating and analyzing carbon flow in a power grid driven by a large NLP model of a power system, characterized in that: In step S4, based on the output result of the neural network model, a power grid carbon flow simulation model is constructed. The simulation model includes the generation link, transmission link and consumption link of the power grid carbon emissions: Combined with the model output data trained in step S3, a power grid carbon flow simulation model is constructed. The power grid carbon flow simulation model is calculated by superimposing the carbon emissions in the generation link, the carbon emissions in the transmission link, and the carbon emissions in the consumption link.

6. According to claim 1, a method for simulating and analyzing carbon flow in a power grid driven by a large NLP model of a power system, characterized in that: In the step S5, the simulation model is used to simulate and calculate the carbon flow of the power grid, obtain the carbon emissions of each node in the power grid, calculate the carbon emissions of the total nodes in the power grid, calculate the average emission factor of the power grid based on the calculated carbon emissions, and evaluate the carbon emissions of the power grid based on the average emission factor of the power grid: The total power generation and node carbon emissions of the entire power grid are calculated, and the total carbon emissions data of the entire power grid are obtained through the calculated total power generation and node carbon emissions of the entire power grid, thereby calculating the average emission factor of the power grid, and evaluating the carbon emissions of the power grid based on the average emission factor of the power grid.

7. The method for simulating and analyzing carbon flow in a power grid driven by a large NLP model of a power system according to claim 4 is characterized by: The neural network model training process is: S301, initializing parameters, using random numbers to initialize the weights and biases of the recurrent neural network; S302, forward propagation, according to the current weight and bias, forward propagation calculation is performed to obtain the predicted value, the data sequence output after feature engineering processing is used as the input sequence, the hidden layer state is obtained through the recurrent neural network, and the output sequence is calculated based on the calculated hidden layer state; S303, calculating the loss, using the mean square error to compare the output data obtained by forward propagation as the predicted value with the input data as the true value, and calculating the value of the loss function; S304, back propagation, calculating the gradient of the loss function with respect to each parameter, starting from the output layer, and back propagating the gradient to the hidden layer and the input layer; S305, updating parameters, updating the parameters of the network according to the calculated gradient and learning rate; S306, repeat the iteration, repeat steps S302 to S305 until the loss function reaches the maximum number of iterations, and then the iteration is completed.

8. The method for carbon flow simulation analysis of a power grid driven by a large NLP model of a power system according to claim 5 is characterized by: The grid carbon flow simulation model superposition calculation is: Generation stage: F1= α P” in, α is the carbon emission coefficient in the power generation process, P ” is the output power generation, and F1 is the carbon emission generated in the generation process; Transmission link: Where β is the loss coefficient during transmission, is the additional carbon emission coefficient in the transmission process, and F2 is the carbon emission generated in the transmission process; Consumption link: F3= δ Q” in, δ is the loss coefficient generated during the power consumption process, Q ” is the output electricity consumption, and F3 is the carbon emissions generated in the consumption process.

9. The method for simulating and analyzing carbon flow in a power grid driven by a large NLP model of a power system according to claim 6, characterized in that: The grid average emission factor is calculated as: Where O is the average emission factor of the power grid, F 总 is the total carbon emissions of the power grid, P 总 is the total power generation of the grid.

Images