A method and system for calculating dynamic carbon emissions of power grid based on machine learning

Abstract

The present invention relates to the technical field of power systems, and in particular to a method and system for calculating dynamic carbon emissions in power grids based on machine learning, including extracting features from power load data to obtain load distribution features in each region; simulating the power flow distribution in each region of the power grid at different time periods based on power grid operation status data and load distribution features in each region using the Newton-Raphson method to obtain dynamic power flow distribution data of the power grid; obtaining power grid power flow correction data based on the dynamic power flow distribution data of the power grid and inter-regional power exchange data using a machine learning algorithm; obtaining the carbon emission transfer amount of each region based on the inter-regional power exchange data using an inter-regional carbon emission transfer model; obtaining the dynamic carbon emissions of the power grid based on the carbon emission transfer amount and power grid power flow correction data. The present invention dynamically calculates and corrects the carbon emissions of the power grid through a machine learning algorithm and a carbon emission transfer model, thereby achieving accurate quantification of the dynamic carbon emissions of the power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method and system for calculating dynamic carbon emissions of a power grid based on machine learning. Background Art

[0002] In the context of the current energy transformation, the low-carbon demand response mechanism has become an important part of the operation of the power system. Among them, the accurate and timely acquisition of the dynamic carbon emission factors of the power grid is the key to achieving efficient and low-carbon dispatch. However, the existing carbon emission factor calculation methods mainly rely on historical data and statistical models. These methods usually require a long time of data accumulation and a long processing cycle, resulting in the calculation results often lagging behind the actual operation status of the power grid, and failing to fully capture the volatility and uncertainty of the output of various types of power sources in the power grid (such as thermal power, hydropower, wind power, photovoltaics, etc.), as well as the real-time distribution of power grid trends. These dynamic factors lead to large differences in carbon emission factors in different regions and different time periods. This static calculation method based on historical data cannot capture these dynamic changes, resulting in deviations between the calculation results of the carbon emission factors and the actual situation.

[0003] Secondly, the electricity load data and carbon emission information of each load node are usually scattered in different data sources. The lack of effective data collection and transmission mechanism makes it difficult to quickly summarize and distribute these data, which not only affects the timeliness of low-carbon demand response, but also further limits the accuracy and real-time nature of the carbon emission factor. At the same time, in the calculation process of the carbon emission factor, the impact of grid losses and cross-regional exchange of electricity on the carbon emission factor cannot be ignored. How to reasonably allocate the carbon emissions corresponding to grid losses and cross-regional exchange of electricity is also a complex issue. Existing calculation methods often fail to fully consider these factors, but instead use a simple allocation model, resulting in errors in the calculation results of the carbon emission factor.

[0004] In summary, there are many problems with the existing methods for calculating dynamic carbon emissions in power grids, which seriously restrict the effective implementation of low-carbon demand response mechanisms. Therefore, there is an urgent need to provide a method for calculating dynamic carbon emissions that can accurately and timely obtain the information. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method and system for calculating dynamic carbon emissions of a power grid based on machine learning.

[0006] In a first aspect, the present invention provides a method for calculating dynamic carbon emissions of a power grid based on machine learning, the method comprising the following steps:

[0007] Collecting grid operation status data and power load data of each load node, and extracting features of the power load data to obtain load distribution features of each area;

[0008] Based on the grid operation status data and the load distribution characteristics of each region, the Newton-Raphson method is used to simulate the power flow distribution of each region of the grid at different times to obtain the dynamic power flow distribution data of the grid;

[0009] According to the dynamic power flow distribution data of the power grid and the inter-regional power exchange data, a machine learning algorithm is used to obtain power flow quantity correction data of the power grid;

[0010] Construct an inter-regional carbon emission transfer model, and use the inter-regional carbon emission transfer model to obtain the carbon emission transfer amount of each region based on the inter-regional power exchange data;

[0011] The dynamic carbon emissions of the power grid are obtained according to the carbon emission transfer amount and the power grid flow quantity correction data.

[0012] In a further embodiment, the step of simulating the power flow distribution of each area of ​​the power grid at different time periods using the Newton-Raphson method based on the power grid operation status data and the load distribution characteristics of each area to obtain the power grid dynamic power flow distribution data includes:

[0013] The power flow distribution model of the power grid is established based on the power grid operation status data, and the load distribution characteristics of each region are analyzed through the time series analysis method to obtain the load fluctuation data of each region in different periods of time;

[0014] Dynamically correcting the power grid power flow distribution model according to the load fluctuation data to obtain a power grid power flow distribution correction model;

[0015] According to the grid topology data, the Newton-Raphson method is used to calculate the power flow of the power grid power flow distribution correction model to obtain the dynamic power flow distribution data of the power grid.

[0016] In a further embodiment, the step of obtaining power grid flow quantity correction data using a machine learning algorithm based on the power grid dynamic flow distribution data and the inter-regional power exchange data includes:

[0017] Based on the power output data of the power system, the load response prediction value is obtained using the machine learning algorithm;

[0018] Obtaining inter-regional power exchange data of each region at different time periods, and obtaining power grid flow data based on the inter-regional power exchange data and load response prediction values; the inter-regional power exchange data includes the power exchange amount, power exchange direction and power exchange time between different regions;

[0019] According to the load distribution characteristics of each area and the dynamic power flow distribution data of the power grid, the power grid loss calculation method is used to calculate the power grid loss in each area;

[0020] According to the direction of power exchange, the power loss of the power grid is allocated to the power flow data of each area according to the power supply ratio, and the power flow correction data of the power grid is obtained.

[0021] In a further embodiment, the step of obtaining a load response prediction value using a machine learning algorithm based on the power output data of the power system includes:

[0022] Obtain the power output data of the power system, analyze the power output data using the autoregressive integral moving average model, and obtain the power output fluctuation rate;

[0023] The average carbon emission intensity change rate is calculated based on the ratio of the difference between the carbon emission intensity data of adjacent periods and the carbon emission intensity data of the previous period;

[0024] Taking the average carbon emission intensity change rate and power output fluctuation rate as incentive factors, a load response incentive model is established using a machine learning algorithm.

[0025] According to the meteorological factors of the forecast day, the load response incentive model is used to analyze the relationship between the load response mode and the incentive factor, and the future load response forecast value of each region is obtained.

[0026] In a further embodiment, the step of obtaining power grid flow data based on the inter-regional power exchange data and the load response prediction value includes:

[0027] The inter-regional power exchange data and the load response prediction value are fuzzified to obtain a fuzzy input value; the fuzzy input value includes a power exchange amount fuzzy set and a load response fuzzy set;

[0028] constructing a fuzzy relationship matrix between the inter-regional power exchange data and the load response prediction value using a fuzzy C-means algorithm;

[0029] According to the fuzzy relationship matrix and the fuzzy input value, a fuzzy reasoning method is used to obtain a fuzzy output value of a power grid power flow state;

[0030] The maximum membership method is used to convert the power grid flow state fuzzy output value into power grid flow quantity data.

[0031] In a further embodiment, the machine learning algorithm is a support vector machine algorithm.

[0032] In a further embodiment, the grid loss calculation method is an equivalent resistance method.

[0033] In a further embodiment, the step of constructing an inter-regional carbon emission transfer model and obtaining the carbon emission transfer amount of each region using the inter-regional carbon emission transfer model according to the inter-regional power exchange data includes:

[0034] Obtaining the carbon emission factor of each region, and constructing a carbon emission feature vector based on the inter-regional power exchange data and the carbon emission factor;

[0035] A long short-term memory network is used to construct an inter-regional carbon emission transfer model, and according to the carbon emission characteristic vector, the inter-regional carbon emission transfer model is used to obtain the carbon emission transfer amount caused by power exchange in each region at different time periods.

[0036] In a further embodiment, the step of obtaining the dynamic carbon emissions of the power grid according to the carbon emission transfer amount and the power grid flow power correction data comprises:

[0037] Determine the carbon emission coefficient of different energy sources in each region based on the energy structure and power generation methods of each region;

[0038] Calculate the regional basic carbon emissions of each region based on the power grid flow electricity correction data and the carbon emission coefficient;

[0039] Allocate the carbon emission transfer amount to the regional basic carbon emissions of each region according to the direction and proportion of power exchange to obtain the regional comprehensive carbon emissions of each region;

[0040] The regional comprehensive carbon emissions of all regions are added together to obtain the dynamic carbon emissions of the power grid.

[0041] In a second aspect, the present invention provides a system for calculating dynamic carbon emissions of a power grid based on machine learning, the system comprising:

[0042] The data acquisition module is used to collect the power grid operation status data and the power load data of each load node, and perform feature extraction on the power load data to obtain the load distribution characteristics of each area;

[0043] The power flow analysis module is used to simulate the power flow distribution of each area of ​​the power grid at different times based on the power grid operation status data and the load distribution characteristics of each area using the Newton-Raphson method to obtain the dynamic power flow distribution data of the power grid;

[0044] A data processing module, used to obtain power grid power flow quantity correction data using a machine learning algorithm according to the power grid dynamic power flow distribution data and inter-regional power exchange data;

[0045] The transfer analysis module is used to construct an inter-regional carbon emission transfer model and obtain the carbon emission transfer amount of each region using the inter-regional carbon emission transfer model based on the inter-regional power exchange data;

[0046] The carbon emission determination module is used to obtain the dynamic carbon emission of the power grid according to the carbon emission transfer amount and the power grid flow power correction data.

[0047] The present invention provides a method and system for calculating dynamic carbon emissions of a power grid based on machine learning. The method collects power grid operation status data and power load data of each load node, and extracts features of the power load data to obtain load distribution characteristics of each region; based on the power grid operation status data and the load distribution characteristics of each region, the Newton-Raphson method is used to simulate the power flow distribution of each region of the power grid at different time periods to obtain dynamic power flow distribution data of the power grid; according to the dynamic power flow distribution data of the power grid and the power exchange data between regions, the machine learning algorithm is used to obtain the power flow power correction data; an inter-regional carbon emission transfer model is constructed, and according to the power exchange data between regions, the carbon emission transfer amount of each region is obtained by using the inter-regional carbon emission transfer model; according to the carbon emission transfer amount and the power flow power correction data of the power grid, the dynamic carbon emissions of the power grid are obtained. Compared with the prior art, the method can realize real-time monitoring and analysis of the power grid operation status, accurately simulate the power grid power flow distribution, and dynamically calculate and correct the carbon emissions of the power grid through the machine learning algorithm and the carbon emission transfer model, so that the dynamic carbon emissions of the power grid can be accurately quantified, providing a scientific basis for the carbon emission management of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of a flow chart of a method for calculating dynamic carbon emissions of a power grid based on machine learning provided by an embodiment of the present invention;

[0049] Figure 2 It is a block diagram of a power grid dynamic carbon emission calculation system based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following specifically illustrates the implementation mode of the present invention in conjunction with the accompanying drawings. The embodiments are provided for illustrative purposes only and cannot be understood as limiting the present invention. The accompanying drawings are provided for reference and illustration only and do not constitute a limitation on the scope of patent protection of the present invention, because many changes can be made to the present invention without departing from the spirit and scope of the present invention.

[0051] refer to Figure 1 , the embodiment of the present invention provides a method for calculating the dynamic carbon emissions of a power grid based on machine learning, such as Figure 1 As shown, the method comprises the following steps:

[0052] S1. Collecting grid operation status data and power load data of each load node, and extracting features from the power load data to obtain load distribution features of each region.

[0053] This embodiment comprehensively collects real-time operation status data of the power grid to form power grid operation status data. The power grid operation status data includes but is not limited to key parameters such as current, voltage, power factor of each transmission line, and frequency and stability indicators of the power grid. At the same time, for each load node in the power grid, this embodiment uses smart meters and remote monitoring systems to record their power load data. These data cover the power consumption in different time periods. Then, this embodiment extracts features from these power load data. For example, time series analysis, frequency domain analysis, and statistical learning methods can be used to extract key features that can reflect the law of load changes and obtain load distribution characteristics of each region, such as load peak, valley, volatility, periodicity, etc. The load distribution characteristics can reflect the power consumption pattern of the power grid in different regions. This is basic data for subsequent power flow simulation and carbon emission calculation, and provides strong data support for subsequent dynamic carbon emission calculation of the power grid.

[0054] S2. Based on the grid operation status data and the load distribution characteristics of each region, the Newton-Raphson method is used to simulate the power flow distribution of each region of the grid at different time periods to obtain the dynamic power flow distribution data of the grid.

[0055] In this embodiment, the steps of simulating the power flow distribution of each area of ​​the power grid at different time periods using the Newton-Raphson method based on the power grid operation status data and the load distribution characteristics of each area to obtain the power grid dynamic power flow distribution data include:

[0056] The power flow distribution model of the power grid is established based on the power grid operation status data, and the load distribution characteristics of each region are analyzed through the time series analysis method to obtain the load fluctuation data of each region in different periods of time;

[0057] Dynamically correcting the power grid power flow distribution model according to the load fluctuation data to obtain a power grid power flow distribution correction model;

[0058] According to the grid topology data, the Newton-Raphson method is used to calculate the power flow of the power grid power flow distribution correction model to obtain the dynamic power flow distribution data of the power grid.

[0059] Specifically, this embodiment uses power system analysis software (such as PSASP, ETAP, etc.) to establish a power grid flow distribution model that can reflect the current operation status of the power grid based on the power grid operation status data, such as key parameters such as voltage amplitude, phase angle, line impedance, etc. of each node in the power grid, as well as real-time operation data such as generator output and load demand. The power grid flow distribution model will simulate the flow and distribution of electricity in the power grid. The model construction can use machine learning methods, such as neural networks, to capture the complex nonlinear relationship in the power grid operation status data. In order to more accurately capture the characteristics of the load changes in each region over time and provide dynamic input for the power grid flow distribution model, this embodiment uses a time series analysis method to analyze the load distribution characteristics of each region. Time series analysis can reveal the periodicity, trend and randomness components in the load data, thereby obtaining the load fluctuation situation. Then, based on the above load fluctuation data, this embodiment uses the time and space of the load The changing characteristics are integrated into the power flow distribution model of the power grid, and the power flow distribution model is dynamically corrected to reflect the dynamic response of the power grid in actual operation. For example, by adjusting the load node parameters in the power flow distribution model of the power grid to match the load fluctuation data, a power flow distribution correction model that is closer to the actual operating state is obtained. Finally, the power flow calculation is performed using the Newton-Raphson method. The Newton-Raphson method is an iterative algorithm that can approximate the real power flow distribution by continuously correcting the power imbalance. It is particularly suitable for dealing with power flow calculation problems in large-scale power systems. On the basis of the power flow distribution correction model, this embodiment sets the iteration initial value and convergence condition according to the topological structure data of the power grid (including node connection relationship, line parameters, etc.), and starts the calculation process of the Newton-Raphson method. This embodiment sets the initial voltage amplitude and phase angle, and constructs the Jacobian matrix (Jacobian) according to the topological structure and power flow equation of the power grid. Matrix) and power imbalance vector, and update the voltage amplitude and phase angle by solving a group of linear equations until the power imbalance satisfies the given convergence condition. After a series of iterative steps, the algorithm converges to a stable solution, and outputs the results of the power flow calculation (including the voltage amplitude and phase angle of each node, the power flow of each line, etc.) as the dynamic power flow distribution data of the power grid. The dynamic power flow distribution data of the power grid is specific and quantified power grid power flow data, which is used to provide an accurate description of the power grid power flow state and provides the necessary basic data for the subsequent calculation of carbon emissions. Through the implementation of the above steps, this embodiment can simulate the power flow distribution of each area of ​​the power grid at different time periods based on the power grid operation state data and the load distribution characteristics of each area, using the Newton-Raphson method, which can not only obtain the dynamic power flow distribution data of the power grid under different time periods and different load conditions, but also provide accurate and reliable input information for the subsequent carbon emission calculation, thereby effectively supporting the formulation and optimization of the low-carbon operation strategy of the power grid.

[0060] S3. Based on the dynamic power flow distribution data of the power grid and the inter-regional power exchange data, a machine learning algorithm is used to obtain power flow quantity correction data of the power grid.

[0061] In this embodiment, the step of obtaining power grid power flow quantity correction data by using a machine learning algorithm according to the power grid dynamic power flow distribution data and the inter-regional power exchange data includes:

[0062] Based on the power output data of the power system, the load response prediction value is obtained using the machine learning algorithm;

[0063] Obtaining inter-regional power exchange data of each region at different time periods, and obtaining power grid flow data based on the inter-regional power exchange data and load response prediction values; the inter-regional power exchange data includes the power exchange amount, power exchange direction and power exchange time between different regions;

[0064] According to the load distribution characteristics of each area and the dynamic power flow distribution data of the power grid, the power grid loss calculation method is used to calculate the power grid loss in each area;

[0065] According to the direction of power exchange, the power loss of the power grid is allocated to the power flow data of each area according to the power supply ratio, and the power flow correction data of the power grid is obtained.

[0066] Specifically, this embodiment uses a machine learning algorithm to analyze historical power output data of the power system, including but not limited to real-time output power and change trends of various power sources such as thermal power generation, wind power generation, and solar power generation, so that it can accurately capture the complex relationship between power output and load demand, as well as the delay and inertia characteristics of load response, to obtain load response prediction values. These load response prediction values ​​not only take into account the changes in power output, but also incorporate the dynamic characteristics of load demand, providing a reliable basis for subsequent power grid flow analysis. In this embodiment, the step of obtaining load response prediction values ​​using a machine learning algorithm based on the power output data of the power system includes:

[0067] Obtain the power output data of the power system, analyze the power output data using the autoregressive integral moving average model, and obtain the power output fluctuation rate;

[0068] The average carbon emission intensity change rate is calculated based on the ratio of the difference between the carbon emission intensity data of adjacent periods and the carbon emission intensity data of the previous period;

[0069] Taking the average carbon emission intensity change rate and power output fluctuation rate as incentive factors, a load response incentive model is established using a machine learning algorithm; the machine learning algorithm is a support vector machine algorithm;

[0070] According to the meteorological factors of the forecast day, the load response incentive model is used to analyze the relationship between the load response mode and the incentive factor, and the future load response forecast value of each region is obtained.

[0071] Then, this embodiment collects and integrates the inter-regional power exchange data of each region at different time periods. These data record the specific value of power exchange between different regions, the clear power flow direction (that is, which region transmits power to which region) and the specific time point when the power exchange occurs, so as to comprehensively reflect the cross-regional power flow of the power grid. While obtaining the inter-regional power exchange data, combined with the load response prediction value, this embodiment uses a fuzzy control algorithm to calculate the power grid flow data such as voltage, current and power of each node in the entire power grid in different time periods, which describes the flow path and value of power in the power grid. After obtaining the power grid flow data, in order to accurately evaluate the loss during the operation of the power grid, this embodiment comprehensively considers the load distribution characteristics of each region (such as load density, load type, etc.) and the dynamic power flow distribution data of the power grid, and uses the equivalent resistance method and other power grid loss calculation methods to accurately calculate the power grid loss of each region in different time periods. This step not only considers direct losses (such as line resistance heating losses), but also considers indirect losses (such as transformer iron losses, reactive compensation equipment losses, etc.), so as to determine The comprehensiveness and accuracy of the loss calculation are ensured, wherein the proportional allocation method is defined as allocating the grid loss according to the proportion of the node power generation or load active power size, and the equivalent resistance method is defined as regarding the active power loss of the distribution network through the element as the active power loss generated by the equivalent resistance of the distribution network line and the equivalent resistance of the distribution transformer; finally, according to the specific direction of power exchange, this embodiment reasonably allocates the calculated grid loss power to the grid flow power data of each region, for example, the allocated loss power is added to the grid flow power data of each region to obtain the grid flow power correction data of each region, and the allocation principle follows the power supply proportion method, that is, the loss power is allocated according to the contribution of each region in the power exchange (that is, the proportion of its power supply to the total power supply). Through this step, this embodiment obtains more accurate grid flow power correction data of each region at different time periods. These data not only reflect the actual flow of power in the grid, but also include the effective power reduced due to grid loss, which provides more accurate and comprehensive data for subsequent carbon emission calculations, and realizes accurate calculation of dynamic carbon emissions of the grid.

[0072] In this embodiment, the step of obtaining power grid flow data according to the inter-regional power exchange data and the load response prediction value includes:

[0073] The inter-regional power exchange data and the load response prediction value are fuzzified to obtain a fuzzy input value; the fuzzy input value includes a power exchange amount fuzzy set and a load response fuzzy set;

[0074] constructing a fuzzy relationship matrix between the inter-regional power exchange data and the load response prediction value using a fuzzy C-means algorithm;

[0075] According to the fuzzy relationship matrix and the fuzzy input value, a fuzzy reasoning method is used to obtain a fuzzy output value of a power grid power flow state;

[0076] The maximum membership method is used to convert the power grid flow state fuzzy output value into power grid flow quantity data.

[0077] Specifically, according to the needs of power grid flow calculation, this embodiment fuzzifies the inter-regional power exchange data and load response prediction values, defines a power exchange quantity fuzzy set and a load response fuzzy set, and defines membership functions for the power exchange quantity fuzzy set and the load response fuzzy set to describe the fuzziness of these fuzzy sets, and obtains fuzzy input values, which include a power exchange quantity fuzzy set and a load response fuzzy set; then, cluster analysis is performed on the inter-regional power exchange data and the load response prediction values ​​by using fuzzy clustering methods such as the fuzzy C-means algorithm to determine the correlation between the power exchange quantity and the load response prediction value, and a fuzzy relationship matrix is ​​constructed based on these correlations. The elements of the fuzzy relationship matrix represent the fuzzy relationship strength between the two fuzzy sets, which can be expressed as Similarity calculation methods (such as cosine similarity, Pearson correlation coefficient, etc.) are used to calculate the element values ​​of the fuzzy relationship matrix; according to the fuzzy relationship matrix and the fuzzy input value, a fuzzy reasoning method (such as Mamdani reasoning or Sugeno reasoning) is used to obtain the fuzzy output value of the power grid flow state, and the maximum membership method is used to convert the fuzzy output value of the power grid flow state into specific power grid flow electricity data. This embodiment is based on real-time inter-regional power exchange data and load response prediction values, and the fuzzy output value of the power grid flow state is predicted by a fuzzy control algorithm. It should be noted that the fuzzy output value of the power grid flow state is the result of fuzzy reasoning, which describes the fuzzy characteristics of the power grid flow state, is used to provide a qualitative description of the power grid flow state, and is an estimate of the future state of the power grid.

[0078] It should be noted that the inter-regional power exchange data and the load response prediction value jointly affect the power flow state of the power grid. The power exchange data directly affects the power supply of the power grid, while the load response prediction value affects the power demand of the power grid. The power grid power flow data is the balance result between power supply and power consumption. In the power system, due to the influence of various factors, there is a certain uncertainty in power exchange and load demand. If this uncertainty is ignored, the result of the power grid power flow calculation will deviate from the actual situation. Therefore, this embodiment introduces a fuzzy relationship matrix. The fuzzy relationship matrix provides the correlation degree or similarity information between the inter-regional power exchange data and the load response prediction value, and can handle the uncertainty and ambiguity in the power grid data. The power grid power flow calculation is performed according to the fuzzy relationship matrix. The power flow calculation is an important link in the power system analysis, which is used to determine the voltage, current and power distribution of each node in the power grid. Through the power flow calculation, the power flow data of each region in the power grid can be obtained.

[0079] S4. Construct an inter-regional carbon emission transfer model, and use the inter-regional carbon emission transfer model to obtain the carbon emission transfer amount of each region based on the inter-regional electricity exchange data.

[0080] In this embodiment, the steps of constructing an inter-regional carbon emission transfer model and obtaining the carbon emission transfer amount of each region using the inter-regional carbon emission transfer model according to the inter-regional power exchange data include:

[0081] Obtaining the carbon emission factor of each region, and constructing a carbon emission feature vector based on the inter-regional power exchange data and the carbon emission factor;

[0082] A long short-term memory network is used to construct an inter-regional carbon emission transfer model, and according to the carbon emission characteristic vector, the inter-regional carbon emission transfer model is used to obtain the carbon emission transfer amount caused by power exchange in each region at different time periods.

[0083] Specifically, this embodiment can obtain the carbon emission factors of each region from the database. These carbon emission factors are usually calculated based on the carbon emission intensity of different types of energy (such as coal, natural gas, nuclear energy, wind energy, etc.). At the same time, the inter-regional power exchange data can be obtained from the power grid dispatching center. These inter-regional power exchange data should include the timestamp of the power exchange, the exchange volume, the involved regions and the direction of power transmission for subsequent analysis. Then, the collected carbon emission factors and inter-regional power exchange data are cleaned and pre-processed to remove outliers, process missing values, and ensure the consistency of the data format. The data is normalized or standardized, and the normalized inter-regional power exchange volume and the corresponding carbon emissions of each region are used to construct a multi-dimensional carbon emission feature vector. The vector can fully reflect the carbon emission characteristics of the power grid. Then, this embodiment can select a long short-term memory network (LSTM) to construct an inter-regional carbon emission transfer model. LSTM TM can capture long-term dependencies in time series data. The LSTM model can obtain the carbon emission transfer amount based on the carbon emission feature vector. Specifically, this embodiment uses a training set to train the LSTM model. During the training process, cross-validation and other techniques can be used to prevent overfitting, and adjust model parameters (such as learning rate, number of hidden layers, number of neurons, etc.) to optimize model performance. After the model training is completed, this embodiment uses the preprocessed carbon emission feature vector as input, and the carbon emission transfer amount of each region in different time periods as output. The carbon emission feature vector is used to calculate the carbon emission transfer amount caused by power exchange in each region in different time periods. It should be noted that since power exchange will lead to the transfer of carbon emission responsibilities, the model needs to consider the amount of power exchange and the carbon emission intensity of the two parties to the exchange. Therefore, it is necessary to construct an inter-regional carbon emission transfer model to analyze the carbon emission transfer amount caused by power exchange.

[0084] S5. Obtain the dynamic carbon emissions of the power grid according to the carbon emission transfer amount and the power grid flow power correction data.

[0085] In this embodiment, the step of obtaining the dynamic carbon emissions of the power grid according to the carbon emission transfer amount and the power grid flow power correction data includes:

[0086] Determine the carbon emission coefficient of different energy sources in each region based on the energy structure and power generation methods of each region;

[0087] Calculate the regional basic carbon emissions of each region based on the power grid flow electricity correction data and the carbon emission coefficient;

[0088] Allocate the carbon emission transfer amount to the regional basic carbon emissions of each region according to the direction and proportion of power exchange to obtain the regional comprehensive carbon emissions of each region;

[0089] The regional comprehensive carbon emissions of all regions are added together to obtain the dynamic carbon emissions of the power grid.

[0090] Specifically, this embodiment determines the carbon emission coefficients of different energy sources in a region based on the energy structure and power generation mode of each region. These coefficients will be used to convert the grid-corrected electricity data set into carbon emissions. According to the grid flow electricity correction data and the carbon emission coefficient of each region, the direct carbon emissions generated by each region due to its own electricity consumption, that is, the regional basic carbon emissions of each region, are calculated. For example, the electricity of each region in the grid flow electricity correction data can be multiplied by the carbon emission coefficient of the corresponding region to obtain the regional basic carbon emissions of each region. Then, according to the direction and proportion of power exchange, the inter-regional carbon emission transfer amount is allocated to the regional basic carbon emissions of each region to obtain the regional comprehensive carbon emissions of each region. For example, if area A transmits a certain amount of electricity to area B, the carbon emissions of area A should be reduced by the corresponding part, while the carbon emissions of area B should be increased by the corresponding part. This embodiment adjusts the regional basic carbon emissions to reflect the impact of inter-regional electricity exchange on carbon emissions. During the allocation process, this embodiment needs to ensure that the comprehensive carbon emissions of each region include the carbon emissions generated by its own electricity consumption and the carbon emission transfer caused by electricity exchange. Finally, the comprehensive carbon emission data of each region are summarized and added to obtain the dynamic carbon emissions of the entire power grid. This embodiment combines the power grid flow electricity correction data and the impact of inter-regional carbon emission transfers, and can more accurately reflect the carbon emissions of the power grid.

[0091] The embodiment of the present invention provides a method for calculating dynamic carbon emissions of a power grid based on machine learning. The method collects power grid operation status data and power load data of each load node, and extracts features of the power load data to obtain load distribution characteristics of each region; based on the power grid operation status data and the load distribution characteristics of each region, the Newton-Raphson method is used to simulate the power flow distribution of each region of the power grid at different time periods to obtain dynamic power flow distribution data of the power grid; according to the dynamic power flow distribution data of the power grid and the power exchange data between regions, the machine learning algorithm is used to obtain the power flow power correction data; an inter-regional carbon emission transfer model is constructed, and according to the power exchange data between regions, the carbon emission transfer amount of each region is obtained by using the inter-regional carbon emission transfer model; according to the carbon emission transfer amount and the power flow power correction data of the power grid, the dynamic carbon emissions of the power grid are obtained. Compared with the prior art, the method can realize real-time monitoring and analysis of the power grid operation status, accurately simulate the power grid power flow distribution, and dynamically calculate and correct the carbon emissions of the power grid through the machine learning algorithm and the carbon emission transfer model, thereby improving the accuracy and efficiency of the power grid carbon emission calculation, and providing a scientific basis for the carbon emission management of the power grid.

[0092] It should be noted that the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0093] In one embodiment, Figure 2 As shown, an embodiment of the present invention provides a system for calculating dynamic carbon emissions of a power grid based on machine learning, the system comprising:

[0094] The data collection module 101 is used to collect the power grid operation status data and the power load data of each load node, and perform feature extraction on the power load data to obtain the load distribution characteristics of each area;

[0095] The power flow analysis module 102 is used to simulate the power flow distribution of each area of ​​the power grid at different time periods using the Newton-Raphson method based on the power grid operation status data and the load distribution characteristics of each area, so as to obtain the dynamic power flow distribution data of the power grid;

[0096] The data processing module 103 is used to obtain power grid power flow quantity correction data using a machine learning algorithm according to the power grid dynamic power flow distribution data and the inter-regional power exchange data;

[0097] The transfer analysis module 104 is used to construct an inter-regional carbon emission transfer model, and obtain the carbon emission transfer amount of each region using the inter-regional carbon emission transfer model based on the inter-regional power exchange data;

[0098] The carbon emission determination module 105 is used to obtain the dynamic carbon emission of the power grid according to the carbon emission transfer amount and the power grid flow power correction data.

[0099] For the specific limitations of a power grid dynamic carbon emission calculation system based on machine learning, please refer to the above-mentioned limitations on a power grid dynamic carbon emission calculation method based on machine learning, which will not be repeated here. A person of ordinary skill in the art will appreciate that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0100] The embodiment of the present invention provides a dynamic carbon emission calculation system for a power grid based on machine learning. The system collects power grid operation status data and power load data of each load node through a data acquisition module, and extracts features of the power load data to obtain load distribution characteristics of each region; the power flow analysis module simulates the power flow distribution of each region of the power grid at different time periods based on the power grid operation status data and the load distribution characteristics of each region using the Newton-Raphson method to obtain dynamic power flow distribution data of the power grid; the data processing module obtains power flow power correction data of the power grid using a machine learning algorithm according to the dynamic power flow distribution data of the power grid and the power exchange data between regions; the transfer analysis module constructs an inter-regional carbon emission transfer model, and obtains the carbon emission transfer amount of each region according to the inter-regional power exchange data using the inter-regional carbon emission transfer model; the carbon emission determination module obtains the dynamic carbon emission of the power grid according to the carbon emission transfer amount and the power flow power correction data of the power grid. Compared with the prior art, the system can realize real-time monitoring and analysis of the power grid operation status, accurately simulate the power grid power flow distribution, and dynamically calculate and correct the carbon emissions of the power grid through machine learning algorithms and carbon emission transfer models, thereby improving the accuracy and efficiency of the power grid carbon emission calculation, and providing a scientific basis for the carbon emission management of the power grid.

[0101] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A method for calculating dynamic carbon emissions of a power grid based on machine learning, characterized in that: The following steps are involved: Collecting grid operation status data and power load data of each load node, and extracting features of the power load data to obtain load distribution features of each area; Based on the grid operation status data and the load distribution characteristics of each region, the Newton-Raphson method is used to simulate the power flow distribution of each region of the grid at different times to obtain the dynamic power flow distribution data of the grid; According to the dynamic power flow distribution data of the power grid and the inter-regional power exchange data, a machine learning algorithm is used to obtain power flow quantity correction data of the power grid; Construct an inter-regional carbon emission transfer model, and use the inter-regional carbon emission transfer model to obtain the carbon emission transfer amount of each region based on the inter-regional power exchange data; Obtaining the dynamic carbon emissions of the power grid according to the carbon emission transfer amount and the power grid flow quantity correction data; The step of obtaining power grid power flow quantity correction data by using a machine learning algorithm according to the power grid dynamic power flow distribution data and the inter-regional power exchange data includes: Based on the power output data of the power system, the load response prediction value is obtained using the machine learning algorithm; Obtaining inter-regional power exchange data of each region at different time periods, and obtaining power grid flow data based on the inter-regional power exchange data and load response prediction values; the inter-regional power exchange data includes the power exchange amount, power exchange direction and power exchange time between different regions; According to the load distribution characteristics of each area and the dynamic power flow distribution data of the power grid, the power grid loss calculation method is used to calculate the power grid loss in each area; According to the direction of power exchange, the power loss of the power grid is allocated to the power flow data of each area according to the power supply ratio, and the power flow correction data of the power grid is obtained; The step of obtaining a load response prediction value using a machine learning algorithm according to the power output data of the power system comprises: Obtain the power output data of the power system, analyze the power output data using the autoregressive integral moving average model, and obtain the power output fluctuation rate; The average carbon emission intensity change rate is calculated based on the ratio of the difference between the carbon emission intensity data of adjacent periods and the carbon emission intensity data of the previous period; Taking the average carbon emission intensity change rate and power output fluctuation rate as incentive factors, a load response incentive model is established using a machine learning algorithm. According to the meteorological factors of the forecast day, the load response incentive model is used to analyze the relationship between the load response mode and the incentive factor, and the future load response forecast value of each region is obtained.

2. A method for calculating dynamic carbon emissions of a power grid based on machine learning as claimed in claim 1, characterized in that: The steps of simulating the power flow distribution of each area of ​​the power grid at different time periods using the Newton-Raphson method based on the power grid operation status data and the load distribution characteristics of each area to obtain the power grid dynamic power flow distribution data include: The power flow distribution model of the power grid is established based on the power grid operation status data, and the load distribution characteristics of each region are analyzed through the time series analysis method to obtain the load fluctuation data of each region in different periods of time; Dynamically correcting the power grid power flow distribution model according to the load fluctuation data to obtain a power grid power flow distribution correction model; According to the grid topology data, the Newton-Raphson method is used to calculate the power flow of the power grid power flow distribution correction model to obtain the dynamic power flow distribution data of the power grid.

3. The method for calculating dynamic carbon emissions of a power grid based on machine learning according to claim 1, characterized in that: The step of obtaining power grid flow data based on the inter-regional power exchange data and the load response prediction value includes: The inter-regional power exchange data and the load response prediction value are fuzzified to obtain a fuzzy input value; the fuzzy input value includes a power exchange amount fuzzy set and a load response fuzzy set; constructing a fuzzy relationship matrix between the inter-regional power exchange data and the load response prediction value using a fuzzy C-means algorithm; According to the fuzzy relationship matrix and the fuzzy input value, a fuzzy reasoning method is used to obtain a fuzzy output value of a power grid power flow state; The maximum membership method is used to convert the power grid flow state fuzzy output value into power grid flow quantity data.

4. The method for calculating dynamic carbon emissions of a power grid based on machine learning according to claim 1, characterized in that: The machine learning algorithm is a support vector machine algorithm.

5. The method for calculating dynamic carbon emissions of a power grid based on machine learning according to claim 1, characterized in that: The grid loss calculation method is the equivalent resistance method.

6. A method for calculating dynamic carbon emissions of a power grid based on machine learning as claimed in claim 1, characterized in that: The steps of constructing an inter-regional carbon emission transfer model and obtaining the carbon emission transfer amount of each region using the inter-regional carbon emission transfer model according to the inter-regional power exchange data include: Obtaining the carbon emission factor of each region, and constructing a carbon emission feature vector based on the inter-regional power exchange data and the carbon emission factor; A long short-term memory network is used to construct an inter-regional carbon emission transfer model, and according to the carbon emission characteristic vector, the inter-regional carbon emission transfer model is used to obtain the carbon emission transfer amount caused by power exchange in each region at different time periods.

7. The method for calculating dynamic carbon emissions of a power grid based on machine learning according to claim 1, characterized in that: The step of obtaining the dynamic carbon emissions of the power grid according to the carbon emissions transfer amount and the power grid flow power correction data comprises: Determine the carbon emission coefficient of different energy sources in each region based on the energy structure and power generation methods of each region; Calculate the regional basic carbon emissions of each region based on the power grid flow electricity correction data and the carbon emission coefficient; Allocate the carbon emission transfer amount to the regional basic carbon emissions of each region according to the direction and proportion of power exchange to obtain the regional comprehensive carbon emissions of each region; The regional comprehensive carbon emissions of all regions are added together to obtain the dynamic carbon emissions of the power grid.

8. A machine learning-based power grid dynamic carbon emission calculation system, characterized in that: The system comprises: The data acquisition module is used to collect the power grid operation status data and the power load data of each load node, and perform feature extraction on the power load data to obtain the load distribution characteristics of each area; The power flow analysis module is used to simulate the power flow distribution of each area of ​​the power grid at different times based on the power grid operation status data and the load distribution characteristics of each area using the Newton-Raphson method to obtain the dynamic power flow distribution data of the power grid; A data processing module, used to obtain power grid power flow quantity correction data using a machine learning algorithm according to the power grid dynamic power flow distribution data and inter-regional power exchange data; The transfer analysis module is used to construct an inter-regional carbon emission transfer model and obtain the carbon emission transfer amount of each region using the inter-regional carbon emission transfer model based on the inter-regional power exchange data; A carbon emission determination module, configured to obtain a dynamic carbon emission amount of a power grid according to the carbon emission transfer amount and the power grid flow quantity correction data; Wherein, the data processing module is specifically used for: Based on the power output data of the power system, the load response prediction value is obtained using the machine learning algorithm; Obtaining inter-regional power exchange data of each region at different time periods, and obtaining power grid flow data based on the inter-regional power exchange data and load response prediction values; the inter-regional power exchange data includes the power exchange amount, power exchange direction and power exchange time between different regions; According to the load distribution characteristics of each area and the dynamic power flow distribution data of the power grid, the power grid loss calculation method is used to calculate the power grid loss in each area; According to the direction of power exchange, the power loss of the power grid is allocated to the power flow data of each area according to the power supply ratio, and the power flow correction data of the power grid is obtained; The method of obtaining a load response prediction value by using a machine learning algorithm based on the power output data of the power system specifically includes: Obtain the power output data of the power system, analyze the power output data using the autoregressive integral moving average model, and obtain the power output fluctuation rate; The average carbon emission intensity change rate is calculated based on the ratio of the difference between the carbon emission intensity data of adjacent periods and the carbon emission intensity data of the previous period; Taking the average carbon emission intensity change rate and power output fluctuation rate as incentive factors, a load response incentive model is established using a machine learning algorithm. According to the meteorological factors of the forecast day, the load response incentive model is used to analyze the relationship between the load response mode and the incentive factor, and the future load response forecast value of each region is obtained.

Images