Industrial multistage carbon flow real-time monitoring and leakage prediction method

Abstract

The invention relates to the technical field of industrial carbon flow monitoring and early warning, and discloses an industrial multi-stage carbon flow real-time monitoring and leakage prediction method, which comprises the following steps of: arranging a plurality of sensors in a carbon emission pipeline at the joint of power equipment and the carbon emission pipeline; historical electricity consumption of the power equipment and carbon emission data collected by each sensor at the same time are collected; constructing a carbon emission prediction model, and training the carbon emission prediction model by taking the historical electricity consumption of the power equipment and the carbon emission data collected by each sensor as a training set; and obtaining the electricity consumption of each power device in a certain period of time, and inputting the electricity consumption into the trained carbon emission prediction model to obtain the predicted carbon emission collected by each sensor in the period of time. According to the method, the carbon emission prediction model is obtained through training based on the electricity consumption and all sensor data at the same moment, and electricity-carbon associated data are fully fused, so that the trained model can perform linkage analysis and prediction on the electric power data and the carbon emission data.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial carbon flow monitoring and early warning, and particularly relates to a method for real-time monitoring of industrial multi-stage carbon flow and leakage prediction. Background Art

[0002] With the increasing global attention to climate change issues, the power industry, as one of the main sources of carbon emissions, it is crucial to accurately monitor the carbon flow direction and timely detect carbon emission leakage. The existing industrial carbon flow monitoring technologies have the following problems:

[0003] 1. Insufficient coverage of the sensor network: The traditional solution relies on a single gas sensor, fails to form a multi-sensor collaborative monitoring network, and lacks an adaptive sampling mechanism, resulting in weak ability to capture early leakage characteristics;

[0004] 2. Disconnection between electricity and carbon data: In the power industry, traditional electricity meters only collect power parameters such as current and voltage, without jointly analyzing power data and carbon emission data, and are unable to establish an electricity-carbon correlation model, resulting in large errors in carbon emission accounting;

[0005] 3. Lack of time synchronization mechanism: Due to differences in sampling frequencies and clock sources between sensor and electricity meter data, the time axis of cross-device data drifts severely, affecting the causal analysis of leakage events. Summary of the Invention

[0006] The purpose of the present invention is to improve the deficiencies in the existing technology and provide a method for real-time monitoring of industrial multi-stage carbon flow and leakage prediction.

[0007] To achieve the above-mentioned invention purpose, the embodiments of the present invention provide the following technical solutions:

[0008] A method for real-time monitoring of industrial multi-stage carbon flow and leakage prediction includes the following steps:

[0009] Step 1, set a number of sensors at the connection between power equipment and carbon emission pipelines and inside the carbon emission pipelines; collect the historical electricity consumption of the power equipment and the carbon emission data collected by each sensor at the same time;

[0010] Step 2, construct a carbon emission prediction model, and use the historical electricity consumption of the power equipment and the carbon emission data collected by each sensor as a training set to train the carbon emission prediction model;

[0011] Step 3, obtain the electricity consumption of each power equipment during a certain period of time, input it into the trained carbon emission prediction model, obtain the carbon emissions predicted by each sensor during this period of time, and determine whether the pipeline has leaked.

[0012] Compared with the existing technology, the beneficial effects of the present invention:

[0013] (1) The present invention places sensors for collecting carbon emissions at the connection points between power equipment and pipelines, near the connection positions between pipelines, and various positions of pipelines. And each sensor can, according to its set position, pre-determine which power equipment's carbon emissions in the front end can be collected by this sensor, thereby forming a data topology structure and realizing a network for multi-sensor collaborative monitoring. Based on the one-way flow direction of gas in the pipeline, an adaptive transfer mechanism for sensor-collected data is formed, which greatly reduces the manual inspection effort when checking pipeline leaks in the later stage.

[0014] (2) The present invention trains a carbon emission prediction model based on the historical electricity consumption of power equipment and the data of all sensors at the same moment, fully integrating the electricity-carbon correlation data, so that the trained model can perform linkage analysis and prediction on power data and carbon emission data, thereby improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can also be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is a flowchart of the method of the present invention;

[0017] Figure 2 It is one of the simplified schematic diagrams of the carbon emission pipeline of the power equipment in the embodiment of the present invention;

[0018] Figure 3 It is the second of the simplified schematic diagrams of the carbon emission pipeline of the power equipment in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. The components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but only represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0020] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present invention, terms such as "first" and "second" are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance, or implying any such actual relationship or order between these entities or operations. In addition, terms such as "connected" and "coupled" can be directly connected between components or indirectly connected via other components.

[0021] The present invention is implemented by the following technical solutions. As Figure 1 shown, an industrial multi-stage carbon flow real-time monitoring and leakage prediction method includes the following steps:

[0022] Step 1, set a number of sensors at the connection between the power equipment and the carbon emission pipeline and inside the carbon emission pipeline; collect the historical electricity consumption of the power equipment and the carbon emission data collected by each sensor at the same time.

[0023] The electricity consumption over a historical period of time can be obtained through the electricity meter connected to the power equipment k where the parameter τ represents a historical period of time and raw indicates the raw data. The time period τ can be divided into T time periods, that is, the electricity consumption at T + 1 time points is divided.

[0024] The sensors arranged in the pipeline are used to obtain the carbon emissions output from the power equipment into the pipeline, and the carbon emissions collected by all sensors are obtained for each time point t t = 0, 1,..., T, where N represents the number of sensors. Since the number of sensors is large and the sensors are set at different positions, it can also be understood as representing the carbon emissions collected by the sensor set at position i.

[0025] Please refer to Figure 2 , which shows a simplified schematic diagram of the carbon emission pipeline of the power equipment. The circles in the pipeline represent a sensor, and the arrows in the pipeline represent the gas flow direction. For example, the carbon emissions collected by sensor 5 are equal to the sum of the carbon emissions generated by power equipment 1, power equipment 2, and power equipment 3. It can be seen that the carbon emissions collected by sensors set at different positions may be different. According to the gas flow rate and flow direction, if the gas at sensor 5 is less than the sum of the carbon emissions of power equipment 1, power equipment 2, and power equipment 3, then it is very likely that there is a leakage before the gas reaches sensor 5, or there is a deviation in the data collected by the sensor (caused by the sensor's own failure).

[0026] The purpose of this solution is to train a neural network model using historical data to predict the carbon emissions of sensors at a certain moment, and then compare them with the real-time carbon emissions to monitor the leakage problem of carbon emissions. At the same time, through the transitivity mechanism of the sensor layout positions, it is possible to check whether there are deviations in the data collected by the sensors. If there are deviations, they will be corrected in a timely manner.

[0027] To ensure that the time for obtaining carbon emissions is highly consistent with the time of electricity consumption the timestamps of the electricity meters and each sensor are normalized using the following formula:

[0028] t adjusted = t raw + Δt device + ∈ jitter ;

[0029] where, t adjusted represents the normalized timestamp, t raw represents the original timestamp of any electricity meter or any sensor, Δt device represents the clock cumulative deviation, ∈ jitter represents the transmission jitter error.

[0030] After the time normalization process, at time t, the electricity consumption of the power equipment is K is the number of power equipment, and the carbon emissions collected by each sensor are From Figure 2 it can be seen that the carbon emissions collected by one sensor may come from only 1 power equipment or may come from multiple power equipment, which is known when the position of the sensor is fixed. Therefore, all sensors are constructed into a topological graph structure, with each sensor as a node, and the attribute of each node is the power equipment that the sensor can collect. For example Figure 2 in [diagram], sensor 5 is used as a node, and its attributes include power equipment 1, power equipment 2, and power equipment 3. The edges between nodes are unidirectionally connected, following the flow direction of the gas in the pipeline. The previous sensor only points unidirectionally to the next sensor. For example Figure 3 in the pipeline where power equipment 2 is located, sensor 1 points to sensors 2, 3, and 4 respectively, sensor 2 points to sensors 3 and 4 respectively, sensor 3 points to 4, and sensor 3 will not point back to sensor 2 or 1. In this way, the carbon emissions collected by the latter node can represent the carbon emissions collected by all the previous nodes. That is to say, the carbon emissions collected by sensor 4 should be equal to those collected by sensor 3, should also be equal to those collected by sensor 2, and should also be equal to those collected by sensor 1, because the attributes of these 4 sensors are only power equipment 2. It is easy to understand that Figure 2The carbon emissions collected by sensor 5 are usually greater than those collected by sensor 4 because sensor 5 has to collect the emissions from not only power equipment 2 but also power equipment 1 and power equipment 3. If only power equipment 2 is started, the carbon emissions collected by sensor 5 will be equal to those collected by sensor 4.

[0031] Furthermore, there should be a corresponding relationship between the power consumption of each power equipment and the carbon emissions collected by the sensors. When the power consumption of a power equipment is high during a certain period, the carbon emissions will also increase accordingly. For a single power equipment, its power consumption and carbon emissions can reach a linear relationship. However, when the number of started power equipment increases, especially when different types of power equipment are in different working conditions, it is difficult to describe the relationship between the power consumption and carbon emissions of all power equipment under various working conditions. Therefore, after collecting a large amount of power consumption data of power equipment and carbon emission data collected by sensors, these data are used as a training set to train the carbon emission prediction model.

[0032] The carbon emission prediction model is used to predict the carbon emissions that each sensor should collect at a certain moment based on the power consumption of the power equipment and the attributes of each sensor. If there is a gap between the actual carbon emissions collected by the sensor and the prediction result, the pipeline is checked based on the location of the sensor to check for pipeline leaks. If there is no pipeline leak, the sensor is corrected.

[0033] Step 2: Build a carbon emission prediction model and use the historical power consumption of the power equipment and the carbon emission data collected by each sensor as a training set to train the carbon emission prediction model.

[0034] The carbon emission prediction model is implemented based on a dynamic graph adjacency matrix. First, construct a graph structure G=(V, R, W) of the sensors, where V represents the set of all sensor nodes, and any sensor node v i ∈V; R represents the set of associated edges. If the gas flowing through node v i also flows through node v j , then there is an edge <v i →v j > between node v i →v j >, and <v i →v j > is stored in R; W represents the attributes of the nodes. If node v i can collect the carbon emissions of power equipment 1, power equipment 2, and power equipment 3, then the attribute of node v i is {for power equipment 1, power equipment 2, power equipment 3}.

[0035] The historical carbon emissions of node v i in the time series are Represent the historical carbon emissions of all nodes in the time series as Represent the carbon emissions of all nodes in graph G at each time step, using a learnable mapping function f with parameter matrix θ θ Map onto the feature dimension:

[0036]

[0037] Obtain the time series features

[0038] Calculate the embedded feature vector:

[0039] E1 = Q1 × E;

[0040] where E1 is the embedded feature vector; Q1 is a learnable embedded feature tensor, which is an N×(T + 1) second-order tensor composed of the number of nodes N in graph G and the total number of time steps T + 1. Each element in this matrix is initialized as a random number within (0, 0.5).

[0041] Perform dynamic graph convolution on the embedded feature vector E1:

[0042]

[0043] where E2 is the feature vector after dynamic graph convolution; σ is the sigmoid activation function; Q2 is a learnable parameter tensor for dynamic graph convolution, which is a U×(T + 1) second-order tensor composed of the number of edges in graph G and the total number of time steps T + 1. U is the number of edges in the associated edge set R. Each element in this matrix is initialized as a random number within (0, 0.5); is the adjacency tensor after the regularized Laplacian transformation of the adjacency tensor A of the dynamic graph. A is an N×N matrix, where the elements are A i,j , i = 1, 2,..., N, j = 1, 2,..., N. If there is an edge between node v i and node v j , then A i,j = 1, otherwise A i,j = 0; is:

[0044]

[0045] where D is a diagonal matrix and I is an identity matrix.

[0046] Construct the objective function of the carbon emission prediction model based on dynamic graph convolution:

[0047]

[0048] Among them, θ represents the objective function; Ω represents the set of power consumption of all power equipment at time t; M is the number of sensors that can collect the carbon emissions of power equipment k, and j = 1, 2,..., M; represents the carbon emissions collected by sensor j; * is the convolution operation; e k is the carbon emission factor of power equipment k; Θ L is the adaptive weight matrix applied to the initial node features, Θ O is the adjacency weight matrix corresponding to Θ L ; represents the transpose of the matrix; [H j represents the node information transfer mechanism, and there is:

[0049]

[0050] Among them, [H j contains the carbon emissions collected by all sensors before the j-th sensor. As Figure 3 shown, for example, sensors 1, 2, 3, and 4 can all collect the carbon emissions of power equipment 2. At this time, M = 4 and j = 1, 2, 3, 4. Then, the [H j corresponding to sensor 4 not only contains the data of sensor 4 but also the data of sensors 1, 2, and 3. σ is the sigmoid activation function; α and β are hyperparameters; H 0 is the initial node information, that is, the data of the sensors that can collect the carbon emissions of power equipment k before time point t; I represents the identity matrix; Q3 is the adjacency weight matrix corresponding to I.

[0051] It can be seen that through the objective function of the carbon emission prediction model constructed in this solution, the latter sensor can learn the data collected by the previous sensor, and based on the power consumption of the power equipment at the corresponding time, learn the relationship between the carbon emissions collected by the sensors at different positions of the power equipment, so as to have the ability to collect the carbon emissions at different positions of multiple power equipment under different working conditions.

[0052] Step 3: Obtain the power consumption of each power equipment in a certain period of time. After inputting it into the trained carbon emission prediction model, obtain the predicted carbon emissions collected by each sensor during this period.

[0053] For example, obtain the power consumption of all power equipment from 10:00 to 15:00 on January 1, 2025 (obtained by electricity meters), divide this period into multiple moments, input the power consumption at the corresponding moments into the carbon emission prediction model, and the model outputs the prediction results of the data collected by each sensor; then obtain the actual data of the sensors at the corresponding moments. If there are errors in the prediction results and the actual results of sensor 7 at multiple moments, and there are also errors in the prediction results and the actual results of all sensors after sensor 7, and the error ranges are also similar, it is determined that there is a risk of leakage in the pipeline before sensor 7, and the risk of leakage in the pipeline between sensor 6 and sensor 7 is relatively large, and manual inspection is required, thereby reducing the intensity of manual inspection. If there are only a few sensors among all sensors, such as one or two sensors, whose prediction results and actual results have errors, and these one or two sensors are not the sensors at the very end of the pipeline, then it can be judged that there is a problem with the accuracy of these one or two sensors and they need to be calibrated. The specific calibration means can use existing technologies, and this solution does not make any limitations or provide protection for them.

[0054] As described above, the above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An industrial multi-stage carbon flow real-time monitoring and leakage prediction method, characterized in that: Including the following steps: Step 1: Set several sensors at the connection between the power equipment and the carbon emission pipeline and inside the carbon emission pipeline; collect the historical power consumption of the power equipment and the carbon emission data collected by each sensor at the same time. Step 2: Build a carbon emission prediction model, and use the historical power consumption of the power equipment and the carbon emission data collected by each sensor as a training set to train the carbon emission prediction model. Step 3: Obtain the power consumption of each power equipment during a certain period of time. After inputting it into the trained carbon emission prediction model, obtain the predicted carbon emissions collected by each sensor during this period, and judge whether the pipeline leaks.

2. The industrial multi-stage carbon flow real-time monitoring and leakage prediction method according to claim 1, wherein: The specific steps of Step 2 include the following steps: Step 2-1: Build a graph structure of the sensor. Step 2-2: Map the historical carbon emissions to the feature dimension. Step 2-3: Build the target model of the carbon emission prediction model based on dynamic graph convolution.

3. The industrial multi-stage carbon flow real-time monitoring and leakage prediction method according to claim 2, characterized in that: The specific steps of step 2-1 include the following: construct the graph structure G=(V, R, W) of the sensor, where V represents the set of all sensor nodes, and any sensor node v i ∈V; R represents the set of associated edges. If the gas flowing through node v i also flows through node v j , then there is an edge <v i →v j > between node v i and node v j , and <v i →v j > is stored in R; W represents the attributes of the node, including the electrical equipment that can collect carbon emissions at the node.

4. The industrial multi-stage carbon flow real-time monitoring and leakage prediction method according to claim 3, characterized in that: The specific steps of Step 2-2 include the following steps: The historical carbon emissions of node vi over the time series are Represent the historical carbon emissions of all nodes over time series as represent the carbon emissions of all nodes in graph G at time steps; Use the learnable mapping function f with parameter matrix θ θ to map to the feature dimension: Obtain time series features 5. The industrial multi-stage carbon flow real-time monitoring and leakage prediction method according to claim 4, characterized in that: The specific steps of Step 2-3 include the following steps: Perform dynamic graph convolution on the embedded feature vector E1: Among them, E2 is the feature vector after dynamic graph convolution; σ is the sigmoid activation function; Q2 is the learnable parameter tensor of dynamic graph convolution, and Q2 is a U×(T + 1) second-order tensor composed of the number of edges in graph G and the total number of time steps T + 1. U is the number of edges in the associated edge set R, and each element in this matrix is initialized to a random number within (0, 0.5); is the adjacency tensor after the adjacency tensor A of the dynamic graph undergoes regularized Laplacian transformation. A is an N×N matrix, where the elements are A i,j , i = 1, 2,..., N, j = 1, 2,..., N. If there is an edge between node v i and node v j , then A i,j = 1; otherwise, A i,j = 0; is: where D is a diagonal matrix and I is an identity matrix. Build the objective function of the carbon emission prediction model based on dynamic graph convolution. Among them, θ represents the objective function; Ω represents the set of power consumption of all power equipment at time t; M is the number of sensors that can collect the carbon emissions of power equipment k, and j = 1, 2,..., M; represents the carbon emissions collected by sensor j; * is the convolution operation; e k is the carbon emission factor of power equipment k; Θ L is the adaptive weight matrix applied to the initial node features, Θ O is the adjacency weight matrix corresponding to Θ L ; represents the transpose of the matrix; [H j represents the node information transfer mechanism.

6. The industrial multi-stage carbon flow real-time monitoring and leakage prediction method according to claim 1, characterized in that: The specific steps of Step 3 include the following steps: Obtain the power consumption of all power equipment during a certain period of time through the electricity meter, divide this period of time into multiple moments, input the power consumption at the corresponding moment into the trained carbon emission prediction model, and the model outputs the prediction results of the data collected by each sensor. Then obtain the real data of the sensor at the corresponding moment. If the prediction results and the real results of a certain sensor have errors at multiple moments, and the prediction results and the real results of all sensors after this sensor also have errors, it is judged that there is a leakage risk in the pipeline before this sensor and manual inspection is required. If there are errors between the prediction results and the real results of only one or two sensors among all sensors, and these one or two sensors are not the sensors at the very end of the pipeline, then it can be judged that the accuracy of these one or two sensors is problematic and needs to be corrected.

Images