Carbon emission model prediction method based on time sequence, related method and device

Abstract

The invention discloses a carbon emission model prediction method based on a time sequence, a related method and a related device. The method comprises the following steps: acquiring a carbon emission prediction model based on the time sequence; inputting the obtained historical carbon emission data of the oil and gas field in the preset time period into the carbon emission prediction model to obtain a carbon emission prediction result; the carbon emission prediction model based on the time sequence is obtained through the following steps that multiple sets of historical carbon emission data of an oil and gas field are obtained; decomposing the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an eigenmode function; obtaining an initial carbon emission prediction model; inputting data in the eigenmode function as training data into an attention mechanism module to obtain weighted representation; enabling the weighted representation and the training data to pass through an LSTM layer to obtain a carbon emission predicted value; updating the initial carbon emission prediction model according to the carbon emission prediction value and the obtained real carbon emission value; and the training process is repeated until a preset condition is met, and a carbon emission prediction model is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas field exploration and development, and in particular to a carbon emission model prediction method based on time series, and related methods and devices. Background Art

[0002] Under the background of the "dual carbon" development strategy, the oil industry is facing a severe situation of carbon reduction, especially the oil and gas extraction industry, which consumes a lot of energy. It is of great practical significance to explore the relationship between the development of the oil and gas industry and carbon emissions and its influencing factors. Under the new energy security strategy, the oil and gas industry needs to fully implement the requirements of increasing reserves and production. Attaching importance to innovation, energy conservation and consumption reduction, and optimizing the energy structure are the most realistic and effective ways to achieve carbon emission reduction in the oil and gas extraction industry. The high-energy-consuming equipment for exploration and development in oil and gas fields and key oil and gas production process links include: machines (pumping units), pumps (water injection pumps, oil pumps), and furnaces (boilers, heating furnaces). The electricity consumption of these three types of equipment accounts for about 80% of the production electricity consumption, and the energy consumption accounts for 87% of the total production energy consumption. Through the energy consumption analysis and carbon emission analysis and prediction of the above key links, energy can be saved, energy efficiency can be improved, and effective carbon emission reduction can be achieved.

[0003] Prediction based on time series data is a common prediction method, which predicts future values ​​or trends by analyzing and modeling the trends, periodicity and seasonality in time series data. Common methods include: moving average method, which smoothes data and predicts future trends by calculating the sliding average of time series data; exponential smoothing method, which smoothes and predicts time series data based on the exponential weighted average method; autoregressive integrated moving average model (ARIMA), which considers the autocorrelation and moving average of time series data; and the latest neural network model, which uses recurrent neural network (RNN) or long short-term memory network (LSTM) to model and predict time series data. Summary of the invention

[0004] In order to obtain more accurate prediction results of carbon emissions from oil and gas fields, an embodiment of the present invention provides a carbon emission model prediction method based on time series, and related methods and devices.

[0005] In a first aspect, an embodiment of the present invention provides a carbon emission model prediction method based on a time series, the method comprising:

[0006] Obtain a carbon emission prediction model based on time series;

[0007] Inputting the acquired historical carbon emission data of the oil and gas field for a preset time period into the carbon emission prediction model to obtain a carbon emission prediction result;

[0008] The carbon emission prediction model based on time series is obtained in the following way:

[0009] Acquire multiple sets of historical carbon emission data of oil and gas fields; the historical carbon emission data include multiple variables and carbon emission values ​​corresponding to multiple time points;

[0010] Decomposing the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function;

[0011] Acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module; the attention mechanism module includes an influencing factor attention module and a time attribute attention module;

[0012] Inputting the data of a preset time period in the eigenmode function as training data into the attention mechanism module to obtain a weighted representation;

[0013] Passing the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value;

[0014] updating the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value;

[0015] The above training process is repeated until the preset conditions are met to obtain the carbon emission prediction model.

[0016] In one or some optional implementations of the embodiments of the present application, the step of inputting the data of the preset time period in the eigenmode function as training data into the attention mechanism module to obtain a weighted representation includes:

[0017] Inputting the training data into an influencing factor attention module to obtain an influencing factor weight matrix;

[0018] Inputting the training data into a time attribute attention module to obtain a time attribute weight matrix;

[0019] The training data is weightedly calculated based on the influencing factor weight matrix and the time attribute weight matrix to obtain a weighted representation.

[0020] In one or some optional implementations of the embodiments of the present application, the inputting the training data into the influencing factor attention module to obtain the influencing factor weight matrix includes:

[0021] Based on the following formula 1, the training data is split along the variable dimension to obtain the intrinsic mode function components corresponding to the multiple variables;

[0022]

[0023] In the formula, IMF i represents the eigenmode function component corresponding to the i-th variable, Indicates the intrinsic mode function component of variable i corresponding to time step t, where t is the preset time length;

[0024] Based on the following formula 2, the variable weights at all times are calculated separately:

[0025]

[0026] In the formula, Represents the weight of variable i corresponding to time step j;

[0027] According to the variable weights at all times, based on the following formula 3, the weight matrix of influencing factors is obtained:

[0028]

[0029] In the formula, A is the weight matrix of influencing factors, A i is the weight of the influencing factor corresponding to variable i, and m is the total number of variables.

[0030] In one or some optional implementations of the embodiments of the present application, the step of inputting the training data into a time attribute attention module to obtain a time attribute weight matrix includes:

[0031] Based on the following formula 4, the training data is split along the time dimension to obtain eigenmode function components corresponding to multiple time steps;

[0032]

[0033] In the formula, IMF i represents the eigenmode function component corresponding to the j time step, represents the intrinsic mode function component of variable m corresponding to time step j, m is the total number of variables, and T represents transpose;

[0034] Based on the following formula 5, the time attribute weights of all variables are calculated separately:

[0035]

[0036] In the formula, Represents the weight of variable j corresponding to time step i;

[0037] According to the time attribute weights of all variables, the time attribute weight matrix is ​​obtained based on the following formula 6:

[0038]

[0039] Where B is the time attribute weight matrix, and t is the preset time length.

[0040] In one or some optional implementations of the embodiments of the present application, the weighted calculation of the training data based on the influencing factor weight matrix and the time attribute weight matrix to obtain the weighted representation includes:

[0041] According to the influencing factor weight matrix, the time attribute weight matrix and the training data, a weighted calculation is performed based on the following formula 7 to obtain a weighted representation:

[0042]

[0043] In the formula, r is the weighted representation, IMF i represents the eigenmode function component corresponding to the i-th variable, A i is the weight of the influencing factor corresponding to variable i, B is the time attribute weight matrix, λ is the carbon emission prediction model parameter, ⊙ represents the dot multiplication operation, m is the total number of variables, and T represents transposition.

[0044] In one or some optional implementations of the embodiment of the present application, the decomposing the historical carbon emission data based on the ensemble empirical mode decomposition method to obtain the intrinsic mode function includes:

[0045] Different random noises are introduced multiple times into each variable of the historical carbon emission data to obtain multiple noisy data:

[0046] Performing empirical mode decomposition on the multiple noisy data respectively, and obtaining multiple eigenmode function component sets based on the following formula 8;

[0047]

[0048] In the formula, IMF i,j It represents the intrinsic mode function component corresponding to the i-th variable obtained by the empirical mode decomposition after adding random noise for the jth time, E i is the noise data corresponding to the i-th variable, r i,j is the residual function;

[0049] The plurality of intrinsic mode function component sets are averaged to obtain a plurality of intrinsic mode function components based on the following formula 9:

[0050]

[0051] In the formula, IMF i represents the intrinsic mode function component corresponding to the i-th variable, and M is the total number of variables;

[0052] The multiple intrinsic mode function components are concatenated to obtain an intrinsic mode function.

[0053] In one or some optional implementations of the embodiment of the present application, the step of passing the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value includes:

[0054] Perform multi-layer LSTM operation on the training data to obtain a cell state and a first hidden state;

[0055] The weighted representation and the first hidden state are input into the LSTM layer, and the second hidden state is obtained based on the following formula 10:

[0056] o t =σ(w·[h t-1 ,r])+b Formula 10;

[0057] In the formula, o t is the second hidden state, σ represents the LSTM layer operation, and h t-1 is the output of the previous layer of the carbon emission prediction model, r is the weighted representation, w and b are the weight and bias values ​​respectively;

[0058] According to the cell state and the second hidden state, the carbon emission prediction value is obtained based on the following formula 11:

[0059] output=o t ·tanh(c t-1 ) Formula 11;

[0060] In the formula, output is the predicted value of carbon emissions, c t-1 The cell state.

[0061] In a second aspect, an embodiment of the present invention provides a carbon emission model training method based on time series, the method comprising:

[0062] Acquire multiple sets of historical carbon emission data of oil and gas fields; the historical carbon emission data include multiple variables and carbon emission values ​​corresponding to multiple time points;

[0063] Decomposing the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function;

[0064] Obtaining an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module;

[0065] Inputting the data of a preset time period in the eigenmode function as training data into the attention mechanism module to obtain a weighted representation;

[0066] Passing the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value;

[0067] updating the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value;

[0068] The above training process is repeated until the preset conditions are met to obtain the carbon emission prediction model.

[0069] In a third aspect, an embodiment of the present invention provides a carbon emission model prediction device based on a time series, the device comprising:

[0070] A first acquisition module is used to acquire a carbon emission prediction model based on a time series;

[0071] The first prediction module is used to input the historical carbon emission data of the oil and gas field in a preset time period into the carbon emission prediction model to obtain a carbon emission prediction result; wherein the carbon emission prediction model based on time series is obtained by the following method:

[0072] The second acquisition module is used to acquire multiple groups of historical carbon emission data of oil and gas fields; the historical carbon emission data includes multiple variables and carbon emission values ​​corresponding to multiple time points;

[0073] A first processing module, configured to decompose the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function;

[0074] A third acquisition module is used to acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module;

[0075] A second processing module is used to input the data of a preset time period in the eigenmode function as training data into an attention mechanism module to obtain a weighted representation;

[0076] A third processing module is used to pass the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value;

[0077] The first updating module is used to update the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; repeat the above training process until the preset conditions are met to obtain the carbon emission prediction model.

[0078] In a fourth aspect, an embodiment of the present invention provides a carbon emission model training device based on time series, the device comprising:

[0079] A first acquisition module is used to acquire multiple groups of historical carbon emission data of oil and gas fields; the historical carbon emission data includes multiple variables and carbon emission values ​​corresponding to multiple time points;

[0080] A first processing module, configured to decompose the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function;

[0081] A second acquisition module is used to acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module;

[0082] A second processing module is used to input the data of a preset time period in the eigenmode function as training data into an attention mechanism module to obtain a weighted representation;

[0083] A third processing module is used to pass the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value;

[0084] The first updating module is used to update the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; repeat the above training process until the preset conditions are met to obtain the carbon emission prediction model.

[0085] In a fifth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned time series-based carbon emission model prediction method, and / or the above-mentioned time series-based carbon emission model training method.

[0086] In a sixth aspect, an embodiment of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the above-mentioned time series-based carbon emission model prediction method, and / or the above-mentioned time series-based carbon emission model training method.

[0087] In the seventh aspect, an embodiment of the present invention provides a computer program product comprising instructions. When the computer program product is run on a computer device, the computer device executes the above-mentioned time series-based carbon emission model prediction method and / or the above-mentioned time series-based carbon emission model training method.

[0088] In an eighth aspect, an embodiment of the present invention provides a chip, comprising a processor and a communication interface, wherein the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement the above-mentioned time series-based carbon emission model prediction method, and / or the above-mentioned time series-based carbon emission model training method.

[0089] The beneficial effects of the above technical solution provided by the embodiment of the present invention include at least:

[0090] The carbon emission model prediction method based on time series provided in an embodiment of the present invention obtains historical carbon emission data of oil and gas fields, processes the historical carbon emission data using a set empirical mode decomposition method, inputs the carbon emission prediction model based on time series for prediction, establishes a carbon emission prediction model by combining an LSTM layer and an attention mechanism module, obtains a weighted representation of the input data, and obtains a carbon emission prediction value, wherein the set empirical mode decomposition can effectively reduce data fluctuations and improve stability for carbon emission nonlinear time series data, and the attention mechanism module extracts weight representations of influencing factors and time attributes respectively, so that the weight representation includes the importance of carbon emission variables and time attributes to carbon emission values, respectively, so that the carbon emission variables are organically combined with model predictions, effectively improving the prediction accuracy, and making the prediction results of the carbon emission prediction model more interpretable.

[0091] The time series-based carbon emission model prediction method provided in the embodiment of the present application can obtain carbon emission prediction results with higher accuracy and interpretability, provide a theoretical basis for carbon emission early warning and emission reduction in the petroleum and chemical industry, help enterprises understand carbon emissions, formulate reasonable carbon emission management strategies, reduce carbon emission costs, reduce greenhouse gas emissions, and improve environmental sustainability and the economic benefits of enterprises.

[0092] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0093] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0095] Figure 1 It is a flow chart of a carbon emission model prediction method based on time series provided in an embodiment of the present application;

[0096] Figure 2 It is a schematic diagram of the structure of the carbon emission prediction model provided in the embodiment of the present application;

[0097] Figure 3 is a schematic diagram of the carbon emission model training process provided in an embodiment of the present application;

[0098] Figure 4 It is a flowchart of a carbon emission model training method based on time series provided in an embodiment of the present application;

[0099] Figure 5 A schematic diagram of the structure of a carbon emission model prediction device based on time series provided in an embodiment of the present application;

[0100] Figure 6 A schematic diagram of the structure of a time series-based carbon emission model training device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0101] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0102] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0103] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0104] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0105] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0106] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0107] It should be understood that the size of the serial numbers of the steps in the following embodiments 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.

[0108] In order to illustrate the technical solution of the present application, a specific embodiment is provided below for illustration.

[0109] The inventors have discovered that, currently, various industries are gradually conducting carbon emission forecasts, but there is still a lack of relevant forecasts for oil and gas fields. How to effectively utilize historical time series data and realize intelligent forecasts of carbon emissions from oil and gas fields through relevant algorithms is an urgent problem to be solved in the current research on carbon emissions from oil and gas fields. At the same time, in the prior art, other technologies for carbon emission forecasting are usually based on deep learning algorithms. The black box characteristics of the model lead to a lack of interpretability in the carbon emission forecast results. The historical data of carbon emissions are nonlinear time series data, and the relevant influencing factors, i.e., variables, are numerous and have large fluctuations, which will affect the accuracy of model training. Based on this, the inventors have made the present invention after further research and development, providing a carbon emission model prediction method based on time series, related methods and devices.

[0110] Embodiment 1

[0111] The embodiment of the present invention provides a carbon emission model prediction method based on time series, referring to Figure 1 As shown, the method includes:

[0112] S101: Obtain a carbon emission prediction model based on time series.

[0113] In the embodiment of the present application, in the above step S101, a carbon emission prediction model based on a time series is obtained, wherein the carbon emission prediction model based on a time series is obtained by:

[0114] Acquire multiple sets of historical carbon emission data of oil and gas fields; the historical carbon emission data include multiple variables and carbon emission values ​​corresponding to multiple time points;

[0115] Decomposing the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function;

[0116] Acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module; the attention mechanism module includes an influencing factor attention module and a time attribute attention module;

[0117] Inputting the data of a preset time period in the eigenmode function as training data into the attention mechanism module to obtain a weighted representation;

[0118] Passing the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value;

[0119] updating the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value;

[0120] The above training process is repeated until the preset conditions are met to obtain the carbon emission prediction model.

[0121] In the embodiment of the present application, the steps of obtaining the carbon emission prediction model based on the time series include:

[0122] The first step is to obtain multiple sets of historical carbon emission data of oil and gas fields; the historical carbon emission data include multiple variables and carbon emission values ​​corresponding to multiple time points.

[0123] Obtain historical carbon emission data of oil and gas fields. The main links of oil and gas production involve water injection wells and oil pumping wells. After the oil is extracted from the wellhead of the oil well, it is gathered in the metering room, the metering room is gathered in the oil transfer station, and the oil transfer station is gathered in the joint station (gathering and transportation station). The water injection well is input from the heating furnace to the metering room and then to the wellhead for water mixing and heating. The metering room measures the oil, water and gas of several wells and then gathers them to the oil transfer station. The oil transfer station transports oil and gas through a three-in-one three-phase separator (oil, gas and water separation) and a two-in-one heating furnace (heating and gathering and transportation). The energy-consuming equipment in the whole process is the heating furnace and the pump, which serve as the original data of historical carbon emissions. The original data of historical carbon emissions include daily electricity consumption data (water injection station, pump number, hourly rate, unit consumption, electricity consumption), daily gas consumption data (station name, liquid production, oil production, daily gas consumption), among which the input variables of the carbon emission prediction model are hourly rate, unit consumption, electricity consumption, liquid production, oil production, and daily gas consumption, and the output of the carbon emission prediction model is the carbon emission value. The above-mentioned multiple sets of historical carbon emission data contain 6 variables and carbon emission values, constituting a historical carbon emission data set. The present invention does not limit the number and type of input variables used in detail.

[0124] The step of obtaining the historical carbon emission data of oil and gas fields also includes preprocessing the historical carbon emission data. The original data is processed and corrected through data cleaning to remove erroneous data, handle missing values, and solve duplicate data; the historical carbon emission data is normalized and scaled to a smaller range to facilitate subsequent model training and prediction; the time series data is converted into the input and output form of supervised learning problems through data conversion processing, such as taking multiple variables in the historical carbon emission data of the current time step as input and the carbon emission value of the next time step as output to construct a supervised learning data set, and the historical carbon emission data is divided into a training set (80%), a validation set (10%), and a test set (10%).

[0125] In the second step, the historical carbon emission data is decomposed based on the ensemble empirical mode decomposition method to obtain the intrinsic mode function.

[0126] In the embodiment of the present application, the step of decomposing the historical carbon emission data based on ensemble empirical mode decomposition (EEMD) to obtain an intrinsic mode function (IMF) includes:

[0127] Different random noises are introduced multiple times into each variable of the historical carbon emission data to obtain multiple noisy data;

[0128] Performing empirical mode decomposition on the multiple noisy data respectively, and obtaining multiple eigenmode function component sets based on the following formula 8;

[0129]

[0130] In the formula, IMF i,j It represents the intrinsic mode function component corresponding to the i-th variable obtained by the empirical mode decomposition after adding random noise for the jth time, E i is the noise data corresponding to the i-th variable, r i,j is the residual function;

[0131] The plurality of intrinsic mode function component sets are averaged to obtain a plurality of intrinsic mode function components based on the following formula 9:

[0132]

[0133] In the formula, IMF i represents the intrinsic mode function component corresponding to the i-th variable, and M is the total number of variables;

[0134] The multiple intrinsic mode function components are concatenated to obtain an intrinsic mode function.

[0135] In the embodiment of the present application, the training set X in the historical carbon emission data set obtained above is subjected to EEMD decomposition, which can be expressed as: The data X of the i-th variable i =[x i,1 ,x i,2 ,...,x i,t ] 1×t , where X c is the carbon emission value, x i,t represents the value of the i-th variable at time step t. The steps of performing EEMD decomposition on the training set X include: introducing different random noises multiple times on each variable of the historical carbon emission data to obtain multiple noisy data, wherein the data of each variable is processed as follows: in the data X of the i-th variable i Add random noise n i (such as standard normal distribution of Gaussian white noise), as shown in formula 12, we get a noisy data E i :

[0136] E i =X i +n i Formula 12;

[0137] The multiple noisy data are subjected to empirical mode decomposition respectively to obtain multiple sets of intrinsic mode function components, wherein for each noisy data, EMD (Empirical Mode Decomposition) decomposition is performed based on the above formula 8 to obtain the corresponding intrinsic mode function component IMF i,j ; Perform an average operation on the multiple intrinsic mode function component sets to obtain multiple intrinsic mode function components, wherein each intrinsic mode function component set is averaged based on the above formula 9 to obtain the intrinsic mode function component corresponding to each variable i; concatenate the multiple intrinsic mode function components to obtain the intrinsic mode function, the training data set in the original historical carbon emission data set Now we concatenate the multiple intrinsic mode function components and obtain the intrinsic mode function as a new training data set. Here T represents the total time series length of the training data.

[0138] The third step is to obtain the initial carbon emission prediction model. The initial carbon emission prediction model includes an LSTM layer and an attention mechanism module. The attention mechanism module includes an influencing factor attention module and a time attribute attention module.

[0139] In the embodiment of the present application, the carbon emission prediction model structure diagram is shown in Figure 2As shown, the historical carbon emission data set is first decomposed using EEMD technology to obtain the intrinsic mode function IMF as training data, and then the carbon emission prediction model is used to map the input data. The carbon emission prediction model includes multiple LSTM layers and attention mechanism modules, wherein the attention mechanism module includes an influencing factor attention module and a time attribute attention module, and finally outputs the carbon emission prediction value. The present invention does not limit the specific number of LSTM layers used.

[0140] In the fourth step, the data of the preset time period in the eigenmode function is input into the attention mechanism module as training data to obtain a weighted representation.

[0141] In the embodiment of the present application, the step of inputting the data of the preset time period in the eigenmode function as training data into the attention mechanism module to obtain the weighted representation includes:

[0142] Input the training data into the influencing factor attention module to obtain the influencing factor weight matrix;

[0143] Input the training data into the temporal attribute attention module to obtain the temporal attribute weight matrix;

[0144] The training data is weighted based on the influencing factor weight matrix and the time attribute weight matrix to obtain a weighted representation.

[0145] The training data is input into the influencing factor attention module to obtain the influencing factor weight matrix, including:

[0146] Based on the following formula 1, the training data is split along the variable dimension to obtain the intrinsic mode function components corresponding to the multiple variables;

[0147]

[0148] In the formula, IMF i represents the eigenmode function component corresponding to the i-th variable, Indicates the intrinsic mode function component of variable i corresponding to time step t, where t is the preset time length;

[0149] Based on the following formula 2, the variable weights at all times are calculated separately:

[0150]

[0151] In the formula, Represents the weight of variable i corresponding to time step j;

[0152] According to the variable weights at all times, based on the following formula 3, the weight matrix of influencing factors is obtained:

[0153]

[0154] In the formula, A is the weight matrix of influencing factors, A i is the weight of the influencing factor corresponding to variable i, and m is the total number of variables.

[0155] The training data is input into the time attribute attention module to obtain a time attribute weight matrix, including:

[0156] Based on the following formula 4, the training data is split along the time dimension to obtain the eigenvalue function components corresponding to multiple time steps;

[0157]

[0158] In the formula, IMF i represents the eigenmode function component corresponding to the j time step, represents the intrinsic mode function component of variable m corresponding to time step j, m is the total number of variables, and T represents transpose;

[0159] Based on the following formula 5, the time attribute weights of all variables are calculated separately:

[0160]

[0161] In the formula, Represents the weight of variable j corresponding to time step i;

[0162] According to the time attribute weights of all variables, the time attribute weight matrix is ​​obtained based on the following formula 6:

[0163]

[0164] Where B is the time attribute weight matrix, and t is the preset time length.

[0165] Among them, the training data is weighted based on the influencing factor weight matrix and the time attribute weight matrix to obtain a weighted representation, including:

[0166] According to the influencing factor weight matrix, time attribute weight matrix and training data, weighted calculation is performed based on the following formula 7 to obtain the weighted representation:

[0167]

[0168] In the formula, r is the weighted representation, IMF i represents the eigenmode function component corresponding to the i-th variable, A i is the weight of the influencing factor corresponding to variable i, B is the time attribute weight matrix, λ is the carbon emission prediction model parameter, ⊙ represents the dot multiplication operation, m is the total number of variables, and T represents transposition.

[0169] In an embodiment of the present application, the intrinsic mode function obtained in the second step above is input as training data into the attention mechanism module, where one sample data contains multiple variable values ​​of a preset time period length, that is, the input of the carbon emission prediction model is an m×t two-dimensional matrix, m is the total number of variables, and t is the total number of time steps contained in the preset time period. The specific calculation steps after a sample data is input into the attention mechanism module include: the first part, inputting the sample data into the influencing factor attention module, including: based on the above formula 1, splitting the sample data along the variable dimension to obtain the intrinsic mode function components IMF corresponding to multiple variables i ; Based on the above formula 2, the variable weights at all times are calculated separately, where Then, based on the above formula 3, the variable weights of all moments in the sample data are combined and spliced ​​to obtain the weight matrix of the influencing factors of the sample data; the second part is to input the sample data into the time attribute attention module, including: based on the above formula 4, the sample data is split along the time dimension to obtain the intrinsic mode function components IMF corresponding to multiple time steps j ; Based on the above formula 5, the time attribute weights of all variables are calculated separately, where Then, based on the above formula 6, the time attribute weights of all variables in the sample data are combined and spliced ​​to obtain the time attribute weight matrix of the sample data; in the third part, based on the above formula 7, the obtained influencing factor weight matrix and the time attribute weight matrix are weighted calculated on the sample data to obtain the weighted representation corresponding to the sample data.

[0170] In the fifth step, the weighted representation and training data are passed through the LSTM layer to obtain the carbon emission prediction value.

[0171] In the embodiment of the present application, the weighted representation and the training data are passed through the LSTM layer to obtain the carbon emission prediction value, including:

[0172] Perform multi-layer LSTM operations on the training data to obtain the cell state and the first hidden state;

[0173] The weighted representation and the first hidden state are input into the LSTM layer, and the second hidden state is obtained based on the following formula 10:

[0174] o t =σ(w·[h t-1 ,r])+b Formula 10;

[0175] In the formula, o t is the hidden state, σ represents the LSTM layer operation, and h t-1 is the output of the previous layer of the carbon emission prediction model, r is the weighted representation, w and b are the weight and bias values ​​respectively;

[0176] According to the cell state and the second hidden state, the carbon emission prediction value is obtained based on the following formula 11:

[0177] output=o t ·tanh(c t-1 ) Formula 11;

[0178] In the formula, output is the predicted value of carbon emissions, c t-1 The cell state.

[0179] In the embodiment of the present application, the sample data in the training data is firstly processed by the multi-layer LSTM operation before the attention mechanism module to obtain the cell state c t-1 And the hidden vector output by the multi-layer LSTM, that is, the first hidden state. Then, the weighted representation obtained in the fourth step and the first hidden state are input into the next LSTM layer for calculation, as shown in the above formula 10, to obtain the hidden state of the current LSTM layer, that is, the second hidden state. According to the cell state c obtained above t-1 The carbon emission prediction value is obtained by the second hidden state, as shown in Formula 11.

[0180] The sixth step is to update the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; repeat the above training process until the preset conditions are met to obtain the carbon emission prediction model.

[0181] In the embodiment of the present application, the carbon emission model training process refers to Figure 3 As shown, the input sample data of the carbon emission prediction model includes multiple variable values ​​of a preset time period length, and the variables include multiple preset input parameters that affect the carbon emissions of oil and gas fields. The model is trained by sliding data from the far end of the timeline to the near end of the timeline. According to the carbon emission prediction value and the actual carbon emission value obtained in the next time step, the model parameters are adjusted, and the above steps are repeated to perform iterative forward propagation to calculate the loss, back propagation and update the tree structure until the preset conditions are met to stop the iterative training. The preset conditions can be set to reach a predetermined total number of iterations or the difference between the accuracy of two adjacent iterations does not exceed a preset threshold, etc. Determine the settings of specific hyperparameters and functions for model training, set the activation function (such as ReLU activation function), optimizer (such as Adam optimizer), learning rate, loss function (such as root mean square error RMSE) of the real-time drilling speed prediction model, and obtain the trained real-time drilling speed prediction model.

[0182] S102: Inputting the acquired historical carbon emission data of the oil and gas field in a preset time period into the carbon emission prediction model to obtain a carbon emission prediction result.

[0183] In an embodiment of the present application, in the above step S102, historical carbon emission data of a preset time period of the oil and gas field to be predicted will be obtained. According to the time series-based carbon emission prediction model obtained in step S101, the carbon emission data containing multiple variables will be input into the carbon emission prediction model to obtain the carbon emission prediction result.

[0184] In the embodiment of the present application, the carbon emission and its influencing factors (variables) data are a batch of time series data that change over time and are affected by multiple factors and present nonlinear changes. Therefore, in the process of carbon emission prediction, it is necessary not only to consider the changes in carbon emissions in the time dimension, but also to consider the connection between the input variable parameters. The carbon emission prediction based on time series proposed by the present invention uses EEMD to decompose the carbon emission related data into corresponding intrinsic mode functions, thereby effectively suppressing the generation of modal aliasing and reducing data fluctuations. Then, the LSTM network is trained by data features, and the influencing factor attention module and the time attribute attention module are added on the basis of the LSTM model. The data is automatically assigned the corresponding influencing factor and time attribute attention weight information through the attention module. The weight can explain the importance of different time attributes and different influencing factors to carbon emissions, so that the prediction results are interpretable. Thereby, the carbon emission prediction model based on time series can well complete the carbon emission prediction, solving the problem of lack of organic combination of carbon emission influencing factors and prediction models.

[0185] Embodiment 2

[0186] Based on the same inventive concept, the embodiment of the present invention also provides a carbon emission model training method based on time series, referring to Figure 4 As shown, the method includes:

[0187] S201: Acquire multiple groups of historical carbon emission data of oil and gas fields; the historical carbon emission data include multiple variables and carbon emission values ​​corresponding to multiple time points;

[0188] S202: Decomposing the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function;

[0189] S203: Obtain an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module;

[0190] S204: inputting the data of the preset time period in the eigenmode function as training data into the attention mechanism module to obtain a weighted representation;

[0191] S205: Passing the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value;

[0192] S206: updating the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; repeating the above training process until the preset conditions are met to obtain the carbon emission prediction model.

[0193] Embodiment 3

[0194] Based on the same inventive concept, the embodiment of the present invention also provides a carbon emission model prediction device based on time series, referring to Figure 5 As shown, the device comprises:

[0195] A first acquisition module 101 is used to acquire a carbon emission prediction model based on a time series;

[0196] The first prediction module 102 is used to input the historical carbon emission data of the oil and gas field in a preset time period into the carbon emission prediction model to obtain a carbon emission prediction result; wherein the carbon emission prediction model based on time series is obtained by:

[0197] The second acquisition module 103 is used to acquire multiple groups of historical carbon emission data of oil and gas fields; the historical carbon emission data includes multiple variables and carbon emission values ​​corresponding to multiple time points;

[0198] A first processing module 104 is used to decompose the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function;

[0199] The third acquisition module 105 is used to acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module;

[0200] A second processing module 106 is used to input the data of a preset time period in the eigenmode function as training data into an attention mechanism module to obtain a weighted representation;

[0201] The third processing module 107 is used to pass the weighted representation and the training data through the LSTM layer to obtain a carbon emission prediction value;

[0202] The first updating module 108 is used to update the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; repeat the above training process until the preset conditions are met to obtain the carbon emission prediction model.

[0203] Embodiment 4

[0204] Based on the same inventive concept, the embodiment of the present invention also provides a carbon emission model training device based on time series, referring to Figure 6 As shown, the device comprises:

[0205] The first acquisition module 201 is used to acquire multiple groups of historical carbon emission data of oil and gas fields; the historical carbon emission data includes multiple variables and carbon emission values ​​corresponding to multiple time points;

[0206] A first processing module 202 is used to decompose the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function;

[0207] The second acquisition module 203 is used to acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module;

[0208] A second processing module 204 is used to input the data of a preset time period in the eigenmode function as training data into an attention mechanism module to obtain a weighted representation;

[0209] The third processing module 205 is used to pass the weighted representation and the training data through the LSTM layer to obtain a carbon emission prediction value;

[0210] The first updating module 206 is used to update the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; repeat the above training process until the preset conditions are met to obtain the carbon emission prediction model.

[0211] Embodiment 5

[0212] Based on the same inventive concept, an embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, it implements the carbon emission model prediction method based on time series as described in the above embodiment one, and / or the carbon emission model training method based on time series as described in the above embodiment two.

[0213] Embodiment 6

[0214] Based on the same inventive concept, an embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the carbon emission model prediction method based on time series as described in the above embodiment 1, and / or the carbon emission model training method based on time series as described in the above embodiment 2.

[0215] Embodiment 7

[0216] Based on the same inventive concept, an embodiment of the present invention also provides a computer program product comprising instructions. When the computer program product is run on a computer device, the computer device executes the carbon emission model prediction method based on time series as described in the above embodiment one, and / or the carbon emission model training method based on time series as described in the above embodiment two.

[0217] Embodiment 8

[0218] Based on the same inventive concept, an embodiment of the present invention also provides a chip, the chip includes a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run a computer program or instruction to implement the carbon emission model prediction method based on time series as described in the above embodiment one, and / or the carbon emission model training method based on time series as described in the above embodiment two.

[0219] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0220] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0221] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0222] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0223] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A carbon emission model prediction method based on time series, characterized in that: include: Obtain a carbon emission prediction model based on time series; Inputting the acquired historical carbon emission data of the oil and gas field for a preset time period into the carbon emission prediction model to obtain a carbon emission prediction result; The carbon emission prediction model based on time series is obtained in the following way: Acquire multiple sets of historical carbon emission data of oil and gas fields; the historical carbon emission data include multiple variables and carbon emission values ​​corresponding to multiple time points; Decomposing the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function; Acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module; the attention mechanism module includes an influencing factor attention module and a time attribute attention module; Inputting the data of a preset time period in the eigenmode function as training data into the attention mechanism module to obtain a weighted representation; Passing the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value; updating the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; The above training process is repeated until the preset conditions are met to obtain the carbon emission prediction model.

2. The method according to claim 1, characterized in that The step of inputting the data of the preset time period in the eigenmode function as training data into the attention mechanism module to obtain a weighted representation includes: Inputting the training data into an influencing factor attention module to obtain an influencing factor weight matrix; Inputting the training data into a time attribute attention module to obtain a time attribute weight matrix; The training data is weightedly calculated based on the influencing factor weight matrix and the time attribute weight matrix to obtain a weighted representation.

3. The method according to claim 2, characterized in that The step of inputting the training data into an influencing factor attention module to obtain an influencing factor weight matrix comprises: Based on the following formula 1, the training data is split along the variable dimension to obtain the intrinsic mode function components corresponding to the multiple variables; In the formula, IMF i represents the eigenmode function component corresponding to the i-th variable, Indicates the intrinsic mode function component of variable i corresponding to time step t, where t is the preset time length; Based on the following formula 2, the variable weights at all times are calculated separately: In the formula, Represents the weight of variable i corresponding to time step j; According to the variable weights at all times, based on the following formula 3, the weight matrix of influencing factors is obtained: In the formula, A is the weight matrix of influencing factors, A i is the weight of the influencing factor corresponding to variable i, and m is the total number of variables.

4. The method according to claim 2, characterized in that The step of inputting the training data into a time attribute attention module to obtain a time attribute weight matrix includes: Based on the following formula 4, the training data is split along the time dimension to obtain eigenmode function components corresponding to multiple time steps; In the formula, IMF i represents the eigenmode function component corresponding to the j time step, represents the intrinsic mode function component of variable m corresponding to time step j, m is the total number of variables, and T represents transpose; Based on the following formula 5, the time attribute weights of all variables are calculated separately: In the formula, Represents the weight of variable j corresponding to time step i; According to the time attribute weights of all variables, the time attribute weight matrix is ​​obtained based on the following formula 6: Where B is the time attribute weight matrix, and t is the preset time length.

5. The method according to claim 4, characterized in that The weighted calculation of the training data based on the influencing factor weight matrix and the time attribute weight matrix to obtain a weighted representation includes: According to the influencing factor weight matrix, the time attribute weight matrix and the training data, a weighted calculation is performed based on the following formula 7 to obtain a weighted representation: In the formula, r is the weighted representation, IMF i represents the eigenmode function component corresponding to the i-th variable, A i is the weight of the influencing factor corresponding to variable i, B is the time attribute weight matrix, λ is the carbon emission prediction model parameter, ⊙ represents the dot multiplication operation, m is the total number of variables, and T represents transposition.

6. The method according to claim 1, characterized in that The method of decomposing the historical carbon emission data based on the ensemble empirical mode decomposition method to obtain the intrinsic mode function includes: Different random noises are introduced multiple times into each variable of the historical carbon emission data to obtain multiple noisy data: Performing empirical mode decomposition on the multiple noisy data respectively, and obtaining multiple eigenmode function component sets based on the following formula 8; In the formula, IMF i,j It represents the intrinsic mode function component corresponding to the i-th variable obtained by the empirical mode decomposition after adding random noise for the jth time, E i is the noise data corresponding to the i-th variable, r i,j is the residual function; The plurality of intrinsic mode function component sets are averaged to obtain a plurality of intrinsic mode function components based on the following formula 9: In the formula, IMF i represents the intrinsic mode function component corresponding to the i-th variable, and M is the total number of variables; The multiple intrinsic mode function components are concatenated to obtain an intrinsic mode function.

7. The method according to claim 1, characterized in that The step of passing the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value includes: Perform multi-layer LSTM operation on the training data to obtain a cell state and a first hidden state; The weighted representation and the first hidden state are input into the LSTM layer, and the second hidden state is obtained based on the following formula 10: o t = σ(w · [h t-1 , r]) + b Equation 10; In the formula, o t is the second hidden state, σ represents the LSTM layer operation, and h t-1 is the output of the previous layer of the carbon emission prediction model, r is the weighted representation, w and b are the weight and bias values ​​respectively; According to the cell state and the second hidden state, the carbon emission prediction value is obtained based on the following formula 11: output=o t ·tanh(c t-1 ) Formula 11; In the formula, output is the predicted value of carbon emissions, c t-1 The cell state.

8. A carbon emission model training method based on time series, characterized in that: include: Acquire multiple sets of historical carbon emission data of oil and gas fields; the historical carbon emission data include multiple variables and carbon emission values ​​corresponding to multiple time points; Decomposing the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function; Obtaining an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module; Inputting the data of a preset time period in the eigenmode function as training data into the attention mechanism module to obtain a weighted representation; Passing the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value; updating the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; The above training process is repeated until the preset conditions are met to obtain the carbon emission prediction model.

9. A carbon emission model prediction device based on time series, characterized in that: include: A first acquisition module is used to acquire a carbon emission prediction model based on a time series; The first prediction module is used to input the historical carbon emission data of the oil and gas field in a preset time period into the carbon emission prediction model to obtain a carbon emission prediction result; wherein the carbon emission prediction model based on time series is obtained by the following method: The second acquisition module is used to acquire multiple groups of historical carbon emission data of oil and gas fields; the historical carbon emission data includes multiple variables and carbon emission values ​​corresponding to multiple time points; A first processing module, configured to decompose the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function; A third acquisition module is used to acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module; A second processing module is used to input the data of a preset time period in the eigenmode function as training data into an attention mechanism module to obtain a weighted representation; A third processing module is used to pass the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value; The first updating module is used to update the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; repeat the above training process until the preset conditions are met to obtain the carbon emission prediction model.

10. A carbon emission model training device based on time series, characterized in that: include: A first acquisition module is used to acquire multiple groups of historical carbon emission data of oil and gas fields; the historical carbon emission data includes multiple variables and carbon emission values ​​corresponding to multiple time points; A first processing module, configured to decompose the historical carbon emission data based on an ensemble empirical mode decomposition method to obtain an intrinsic mode function; A second acquisition module is used to acquire an initial carbon emission prediction model; the initial carbon emission prediction model includes an LSTM layer and an attention mechanism module; A second processing module is used to input the data of a preset time period in the eigenmode function as training data into an attention mechanism module to obtain a weighted representation; A third processing module is used to pass the weighted representation and the training data through an LSTM layer to obtain a carbon emission prediction value; The first updating module is used to update the initial carbon emission prediction model according to the carbon emission prediction value and the acquired actual carbon emission value; repeat the above training process until the preset conditions are met to obtain the carbon emission prediction model.

11. A computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal, the terminal executes the carbon emission model prediction method based on time series as described in any one of claims 1 to 7, and / or the carbon emission model training method based on time series as described in claim 8.

12. A computer device, characterized in that: It includes a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the carbon emission model prediction method based on time series as described in any one of claims 1 to 7, and / or the carbon emission model training method based on time series as described in claim 8.

13. A computer program product comprising instructions, which, when executed on a computer device, enables the computer device to execute the carbon emission model prediction method based on time series as described in any one of claims 1 to 7, and / or the carbon emission model training method based on time series as described in claim 8.

14. A chip, comprising a processor and a communication interface, wherein the communication interface and the processor are coupled, and the processor is used to run a computer program or instruction to implement the carbon emission model prediction method based on time series as described in any one of claims 1 to 7, and / or the carbon emission model training method based on time series as described in claim 8.

Images