Real-time carbon emission estimation method for high energy-consuming enterprises based on big data mining technology

Abstract

The present invention discloses a real-time carbon emission estimation method for high-energy-consuming enterprises based on big data mining technology. For direct emissions in the production process, the collected load data of high-energy-consuming enterprises are input into a multi-task learning GRU neural network, the network bottom layer shares parameters, and the existing load state categories and corresponding powers are used as model training supervision quantities to perform equipment status monitoring and power decomposition. According to the carbon emission intensity of the equipment under different states, direct carbon emissions are calculated. For indirect carbon emissions, based on the collected load data of high-energy-consuming enterprises, the power consumption for each estimated time interval is calculated; the carbon emissions on the power supply side are obtained by solving the DC optimal power flow, and the marginal carbon emission factor of the grid node to which the enterprise is connected is obtained according to the unit increment method; indirect carbon emissions are calculated using the load data of high-energy-consuming enterprises and the marginal carbon emission factor. The results obtained by direct emissions and indirect emissions are added together to achieve real-time and accurate carbon emissions estimation.

Description

Technical Field

[0001] The present invention belongs to the field of integrated energy and relates to a real-time carbon emission estimation method for high-energy-consuming enterprises based on big data mining technology. Specifically, direct carbon emissions are calculated through big data load decomposition, and indirect carbon emissions are estimated through node marginal carbon emission factors and DC optimal power flow, thereby estimating the real-time emissions of high-energy-consuming enterprises. Background Art

[0002] Carbon emission estimation is the basis for achieving environmental goals such as climate change mitigation and carbon neutrality, and can effectively promote the transition to a low-carbon economy. Industry is one of the main sources of carbon emissions, and high-energy-consuming enterprises such as steel are major carbon emitters in the manufacturing industry, accounting for about 15% of the world's total carbon emissions, and their energy consumption accounts for a large proportion. Estimating the carbon emissions of high-energy-consuming enterprises can obtain specific emissions, thereby providing support for emission reduction. At present, most of the research related to corporate carbon footprint estimation only considers direct carbon emissions in the production process of the enterprise, without considering indirect carbon emissions caused by its electricity consumption. In addition, current research estimates carbon emissions mainly based on information such as corporate annual reports. The estimated time granularity is too coarse, and the estimation accuracy is low, which cannot meet the real-time and accurate carbon emission reduction needs. In this regard, the present invention discloses a real-time carbon emission estimation method for high-energy-consuming enterprises based on big data mining technology, specifically, direct carbon emissions are calculated through big data mining technology, and indirect carbon emissions are estimated through node marginal carbon emission factors and DC optimal power flow, thereby estimating the real-time emissions of high-energy-consuming enterprises. Summary of the Invention

[0003] The purpose of the present invention is to provide a real-time carbon emission estimation method for high-energy-consuming enterprises based on big data mining technology. The load data of high-energy-consuming enterprises is input into a multi-task learning GRU neural network (Multitask Learning-GateRecurrentUnit, MTL-GRU). The network bottom layer shares parameters, and the existing load state categories and corresponding power are used as model training supervision quantities to perform equipment status monitoring and power decomposition. Based on the carbon emission intensity of the equipment under different states, direct carbon emissions are calculated. The marginal carbon emission factor of the grid node to which the enterprise is connected is calculated based on the DC optimal power flow and the unit increment method. Indirect carbon emissions are thus calculated. The real-time carbon emission estimate is the sum of direct emissions and indirect emissions.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] A real-time carbon emissions estimation method for high-energy-consuming enterprises based on big data mining technology is proposed. Real-time carbon emissions estimation for high-energy-consuming enterprises is divided into direct emissions from the production process and indirect emissions caused by electricity consumption. For direct emissions from the production process, the collected load data of high-energy-consuming enterprises is input into a multitask learning-gate recurrent unit (MTL-GRU) neural network. The underlying network shares parameters, uses existing load state categories and corresponding power as model training supervision, and performs equipment state monitoring and power decomposition. Direct carbon emissions are calculated based on the carbon emission intensity of different equipment states. For indirect carbon emissions, the collected load data of high-energy-consuming enterprises is first used to calculate the electricity consumption for each estimation time interval. Then, the power supply-side carbon emissions are calculated by solving the DC optimal power flow (DC-OPF). The marginal carbon emission factor of the grid node to which the enterprise is connected is obtained using the unit increment method. Finally, indirect carbon emissions are calculated using the load data and marginal carbon emission factor of the high-energy-consuming enterprise. The results of direct and indirect emissions are added together to achieve real-time and accurate carbon emissions estimation.

[0006] A method for estimating real-time carbon emissions from high-energy-consuming enterprises based on big data mining technology. The specific steps and processes are as follows:

[0007] Step 1: Obtain load data of high-energy-consuming enterprises and perform data cleaning, including missing value filling and outlier processing;

[0008] Step 2: Build a multi-task learning GRU neural network. Use the results from step 1 as the input of the neural network. The underlying network shares parameters. Use the existing load status categories and corresponding power as model training supervision to perform equipment status monitoring and power decomposition.

[0009] Step 3: Based on the equipment status and power decomposition results of step 2, calculate the direct carbon emissions E within a period of time according to the carbon emission intensity of the equipment in different states. direct .

[0010] Step 4: Using the unit increment method, calculate the DC optimal power flow (DC-OPF) before and after a one-unit load increase. From this, calculate the power supply-side carbon emissions before and after the one-unit load increase. Difference the carbon emissions before and after the one-unit load increase to obtain the marginal carbon emission factor.

[0011] Step 5: Calculate indirect carbon emissions E using the load data of high-energy-consuming enterprises and marginal carbon emission factors indirect .

[0012] Step 6: Direct carbon emissions E direct and indirect carbon emissions E indirectThe results are added together to estimate the real-time carbon emissions of high-energy-consuming enterprises.

[0013] The estimation object of the real-time carbon emission estimation method is high-energy-consuming enterprises.

[0014] Furthermore, the multi-task learning GRU neural network method for improving decomposition accuracy has the following specific process:

[0015] (1) One-dimensional convolution module

[0016]

[0017] (2) GRU module

[0018] Reset Gate r t , update gate z t , hidden state h t It can be expressed by the following equation:

[0019] r t =σ(W r ·[h t-1 ,x t ])

[0020] z t =σ(W z ·[h t-1 ,x t ])

[0021]

[0022]

[0023] Where W r , W z , is the weight matrix, x t Input for the current moment, h t-1 is the hidden state at time t-1, and σ() is the Sigmoid activation function.

[0024] (3) Parameter sharing

[0025] The bottom layer of the network uses shared parameters with hard parameter constraints.

[0026] Furthermore, the direct carbon emissions E over a period of time can be calculated direct , the specific calculation process is as follows:

[0027]

[0028] Among them, S i,j is the state of device i in time period j, is the carbon emission intensity of equipment i in time period j. T is the time interval, b is the number of time periods, and d is the total number of equipment in the enterprise.

[0029] Furthermore, the unit increment method is used to calculate the carbon emissions on the power supply side before and after a one-unit load increase based on the DC optimal power flow (DC-OPF), thereby obtaining the marginal carbon emission factor α of the grid node to which the enterprise is connected. The specific implementation process is as follows:

[0030] DC Optimal Power Flow (DC-OPF):

[0031]

[0032] st

[0033] B T θ=P

[0034]

[0035]

[0036] in, represents the active power of generator i, represents the cost function of generator i. B represents the admittance matrix, θ represents the node voltage phase angle, and P represents the node injection power. ij represents the voltage phase angle difference between nodes i and j, Indicates the upper bound of the active power flow through line j. PG i max and PG i min Indicates the upper and lower bounds of the generator active power. L represents the number of lines, and l represents the number of units.

[0037] Calculate the carbon emissions E based on the unit output before and after the load increases by one unit. G and E G ′, thus obtaining the marginal carbon emission factor α=E G ′-E G Then we can calculate the indirect carbon emissions E over a period of time. indirect :

[0038]

[0039] Among them, α j represents the marginal carbon emission factor of the node in time period j, E j It represents the electricity consumption of high energy-consuming enterprises in time period j, T is the time interval, and b is the number of time periods.

[0040] Furthermore, carbon emissions include direct emissions during the production process and indirect emissions caused by electricity consumption. The specific calculation process is as follows:

[0041] E total =E direct +E indirect .

[0042] The present invention provides a real-time carbon emission estimation method for high-energy-consuming enterprises based on big data mining technology. The load data of high-energy-consuming enterprises is input into a multi-task learning-gate recurrent unit (MTL-GRU) neural network. The network bottom layer shares parameters, and the existing load state categories and corresponding power are used as model training supervision quantities to perform equipment status monitoring and power decomposition. Direct carbon emissions are calculated based on the carbon emission intensity of the equipment under different states. The marginal carbon emission factor of the grid node to which the enterprise is connected is calculated based on the DC optimal power flow and the unit increment method. Indirect carbon emissions are thus calculated. The real-time carbon emission estimate is the sum of direct emissions and indirect emissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of a method for estimating real-time carbon emissions of high-energy-consuming enterprises based on big data mining technology of the present invention;

[0044] Figure 2 This is a diagram of the multi-task learning GRU neural network structure. DETAILED DESCRIPTION

[0045] The following is a further description of specific embodiments of the present invention in conjunction with the accompanying drawings. It should be noted that the description of these embodiments is intended to facilitate understanding of the present invention and does not constitute a limitation of the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0046] The present invention proposes a method for estimating carbon emissions in real time for high energy-consuming enterprises based on big data mining technology. The corresponding flow chart of the method for estimating carbon emissions in real time for high energy-consuming enterprises based on big data mining technology is as follows: Figure 1 shown.

[0047] A method for estimating real-time carbon emissions from high-energy-consuming enterprises based on big data mining technology. The specific steps and processes are as follows:

[0048] Step 1: Obtain load data of high-energy-consuming enterprises and perform data cleaning, including missing value filling and outlier processing;

[0049] Step 2: Build a multi-task learning GRU neural network. Use the results from step 1 as the input of the neural network. The underlying network shares parameters. Use the existing load status categories and corresponding power as model training supervision to perform equipment status monitoring and power decomposition.

[0050] Step 3: Based on the equipment status and power decomposition results of step 2, calculate the direct carbon emissions E within a period of time according to the carbon emission intensity of the equipment in different states. direct .

[0051] Step 4: Using the unit increment method, calculate the DC optimal power flow (DC-OPF) before and after a one-unit load increase. From this, calculate the power supply-side carbon emissions before and after the one-unit load increase. Difference the carbon emissions before and after the one-unit load increase to obtain the marginal carbon emission factor.

[0052] Step 5: Calculate indirect carbon emissions E using the load data of high-energy-consuming enterprises and marginal carbon emission factors indirect .

[0053] Step 6: Direct carbon emissions E direct and indirect carbon emissions E indirect The results are added together to estimate the real-time carbon emissions of high-energy-consuming enterprises.

[0054] The estimation object of the real-time carbon emission estimation method is high-energy-consuming enterprises.

[0055] The specific process of the multi-task learning GRU neural network method to improve decomposition accuracy is as follows:

[0056] Multi-task learning GRU neural network structure Figure 2 shown.

[0057] (1) One-dimensional convolution module

[0058]

[0059] (2) GRU module

[0060] Reset Gate r t , update gate z t , hidden state h t It can be expressed by the following equation:

[0061] r t =σ(W r ·[h t-1 ,x t ])

[0062] z t =σ(W z ·[h t-1 ,xt ])

[0063]

[0064]

[0065] Where W r , W z , is the weight matrix, x t Input for the current moment, h t-1 is the hidden state at time t-1, and σ() is the Sigmoid activation function.

[0066] (3) Parameter sharing

[0067] The bottom layer of the network uses shared parameters with hard parameter constraints.

[0068] The direct carbon emissions E over a period of time can be calculated direct , the specific calculation process is as follows:

[0069]

[0070] Among them, S i,j is the state of device i in time period j, is the carbon emission intensity of equipment i in time period j. T is the time interval, b is the number of time periods, and d is the total number of equipment in the enterprise.

[0071] Using the unit increment method and based on the DC optimal power flow (DC-OPF), we calculate the carbon emissions on the power supply side before and after a one-unit load increase, thereby obtaining the marginal carbon emission factor α of the grid node to which the enterprise is connected. The specific implementation process is as follows:

[0072] DC Optimal Power Flow (DC-OPF):

[0073]

[0074] st

[0075] B T θ=P

[0076]

[0077]

[0078] in, represents the active power of generator i, represents the cost function of generator i. B represents the admittance matrix, θ represents the node voltage phase angle, and P represents the node injection power. ij represents the voltage phase angle difference between nodes i and j, Indicates the upper bound of the active power flow through line j. PG i max and PG i min Indicates the upper and lower bounds of the generator active power. L represents the number of lines, and l represents the number of units.

[0079] Calculate the carbon emissions E based on the unit output before and after the load increases by one unit. G and E G ′, thus obtaining the marginal carbon emission factor α=E G ′-E G Then we can calculate the indirect carbon emissions E over a period of time. indirect :

[0080]

[0081] Among them, α j represents the marginal carbon emission factor of the node in time period j, E j It represents the electricity consumption of high energy-consuming enterprises in time period j, T is the time interval, and b is the number of time periods.

[0082] Carbon emissions include direct emissions from the production process and indirect emissions caused by electricity consumption. The specific calculation process is as follows:

[0083] E total =E direct +E indirect

[0084] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0085] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and modifications fall within the scope of the invention as claimed.

Claims

1. A real-time carbon emissions estimation method for high-energy-consuming enterprises based on big data mining technology, characterized by: The specific steps of the method are as follows: Step 1: Obtain load data of high-energy-consuming enterprises and perform data cleaning, including missing value filling and outlier processing; Step 2: Build a multi-task learning GRU neural network. Use the results from step 1 as the input of the neural network. The underlying network shares parameters. Use the existing load status categories and corresponding power as model training supervision to perform equipment status monitoring and power decomposition. Step 3: Based on the equipment status and power decomposition results of step 2, calculate the direct carbon emissions E within a period of time according to the carbon emission intensity of the equipment in different states. direct ; The specific calculation process is as follows: Among them, S i,j is the state of device i in time period j, is the carbon emission intensity of equipment i in time period j; T is the time interval, b is the number of time periods, and d is the total number of equipment in the enterprise; Step 4: Use the unit increment method to calculate the DC optimal power flow (DC-OPF) before and after a one-unit load increase. From this, calculate the carbon emissions on the power supply side before and after the one-unit load increase. Difference the carbon emissions before and after the one-unit load increase to obtain the marginal carbon emission factor. The specific implementation process is as follows: DC Optimal Power Flow DC-OPF: st B T θ=P in, represents the active power of generator i, represents the cost function of generator i, B represents the admittance matrix, θ represents the node voltage phase angle, P represents the node injection power, θ ij represents the voltage phase angle difference between nodes i and j, represents the upper bound of the active power flow through line j, PG i max and PG i min Indicates the upper and lower bounds of the generator active power, L indicates the number of lines, and l indicates the number of units; Calculate the carbon emissions E based on the unit output before and after the load increases by one unit. G and E G ′, thus obtaining the marginal carbon emission factor α=E G ′-E G , and then the indirect carbon emissions E over a period of time can be calculated indirect : Among them, α j represents the marginal carbon emission factor of the node in time period j, E j represents the electricity consumption of high energy-consuming enterprises in time period j, T is the time interval, and b is the number of time periods; Step 5: Calculate indirect carbon emissions E using the load data of high-energy-consuming enterprises and marginal carbon emission factors indirect ; Step 6: Direct carbon emissions E direct and indirect carbon emissions E indirect The results are added together to estimate the real-time carbon emissions of high-energy-consuming enterprises.

2. The method for estimating real-time carbon emissions of high-energy-consuming enterprises based on big data mining technology according to claim 1 is characterized in that: In step 2, the multi-task learning GRU neural network improves the decomposition accuracy method. The specific process is as follows: (1) One-dimensional convolution module (2) GRU module Reset Gate r t , update gate z t , hidden state h t It can be expressed by the following equation: r t =σ(W r ·[h t-1 ,x t ]) z t =σ(W z ·[h t-1 ,x t ]) Where W r , W z , is the weight matrix, x t Input for the current moment, h t-1 is the hidden state at time t-1, σ() is the Sigmoid activation function; (3) Parameter sharing The bottom layer of the network uses shared parameters with hard parameter constraints.

3. The method for estimating real-time carbon emissions of high-energy-consuming enterprises based on big data mining technology according to claim 1 is characterized in that: Carbon emissions include direct emissions from the production process and indirect emissions caused by electricity consumption. The specific calculation process is as follows: AND total =And direct +E indirect 。

Images