A carbon energy dynamic monitoring method and system based on multi-source data fusion

Abstract

The present application belongs to the technical field of data processing, and particularly relates to a method and system for dynamic monitoring of energy carbon based on multi-source data fusion, which comprises the following steps: collecting high-frequency energy data and low-frequency emission data of a target area, performing frequency-raising processing on the low-frequency emission data by using an interpolation algorithm, and performing standardization processing on the two types of data to obtain energy sequences and emission sequences with unified time resolution; calculating the causal fluctuation confidence of the energy sequences at each time, and constructing a cumulative distance matrix in combination with the causal fluctuation confidence; searching for an optimal alignment path of the energy sequences and the emission sequences by using a dynamic time warping algorithm, and calculating a local slope index at each path point; dynamically adjusting the observation noise covariance of Kalman filtering according to the local slope index, performing smoothing correction on the aligned emission data, and outputting the energy carbon monitoring result. The present application solves the problem of distortion in calculation of carbon emission intensity caused by phase misplacement of multi-source data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a method and system for dynamic monitoring of energy and carbon based on multi-source data fusion. Background Technology

[0002] Driven by dual carbon objectives, the energy, chemical, and power industries are undergoing a transformation from single energy consumption monitoring to dual energy and carbon control management. Existing energy and carbon monitoring systems typically rely on sensors installed at various process stages to collect data, converting energy consumption into carbon emissions using fixed emission factors. This calculation method based on static emission factors assumes a strict linear relationship between energy consumption and carbon emissions, and has certain applicability under steady-state conditions.

[0003] However, in actual industrial production scenarios, the spatiotemporal asynchrony of multi-source data is becoming increasingly prominent. On the one hand, power data is typically acquired at high frequencies on the order of seconds or milliseconds, while gas flow or carbon concentration detection, limited by sensor response speed and detection principles, is often acquired at low frequencies on the order of minutes, resulting in inconsistent sampling frequencies. On the other hand, from the time the equipment is turned on to the time the carbon concentration change is detected at the exhaust outlet, there is a physical transmission delay affected by pipe length, gas flow rate, pipeline layout, and environmental conditions. This delay can range from tens of seconds to several minutes. This frequency difference and physical response lag directly lead to phase misalignment of the data, especially under dynamic operating conditions with frequent load fluctuations, where the timing misalignment is even more pronounced.

[0004] Using simple time alignment or linear interpolation techniques directly ignores the nonlinear time distortion caused by physical lag and frequency differences, leading to incorrect coupling between electricity data and carbon emission data at the same moment, resulting in spurious peaks or mismatches. This phase mismatch causes severe distortion in the calculated instantaneous carbon emission intensity, thereby misleading companies' carbon asset management and energy conservation and emission reduction decisions. Summary of the Invention

[0005] To address the technical problem of carbon emission intensity calculation distortion caused by phase misalignment of multi-source data, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a dynamic monitoring method for energy and carbon based on multi-source data fusion, comprising: collecting high-frequency energy data and low-frequency emission data of a target area, up-processing the low-frequency emission data using an interpolation algorithm, and standardizing the high-frequency energy data and the up-processed low-frequency emission data to obtain an energy sequence and emission sequence with a unified time resolution. The causal fluctuation confidence score of the energy sequence at each time step is calculated. The causal fluctuation confidence score characterizes the effectiveness of energy fluctuations in transmitting carbon emissions. The distance cost between the energy sequence and the emission sequence is calculated by combining the causal fluctuation confidence score, and a cumulative distance matrix is ​​constructed. The optimal alignment path between the energy sequence and the emission sequence is obtained by searching the cumulative distance matrix using the dynamic time warping algorithm, and the local slope index of the optimal alignment path at each path point is calculated. The observation noise covariance of the Kalman filter is dynamically adjusted based on the local slope index. The aligned emission data is then smoothed and corrected using the parameter-adjusted Kalman filter, and the energy and carbon monitoring results are output.

[0007] This invention identifies and corrects energy and carbon data misalignment caused by slow sensor response and physical transmission delay by constructing causal fluctuation confidence and dynamically adjusting Kalman filter parameters. It avoids instantaneous carbon emission intensity calculation errors caused by time misalignment, ensures the consistency of production-emission logic, and thus restores the true carbon emission intensity.

[0008] Preferably, the standardization processing of the high-frequency energy data and the up-frequency low-frequency emission data includes: Calculate the arithmetic mean and standard deviation of the high-frequency energy data and the up-frequency low-frequency emission data respectively; subtract the corresponding arithmetic mean from each data point and divide by the corresponding standard deviation to map each data point to the same distribution space, thereby obtaining the processed energy sequence and emission sequence.

[0009] This invention eliminates the differences in physical dimensions through standardization, providing a unified and comparable data foundation for subsequent distance-based fusion calculations.

[0010] Preferably, the causal fluctuation confidence level satisfies the following relationship: ; In the formula, For time points Confidence level of causal fluctuations For the energy sequence in the 1st The value at each moment. For the emission sequence in the 1st The value at each moment. The preset basic physical transmission delay constant, The length of the sliding window. This is the arithmetic mean of the energy data within the current window. This is the arithmetic mean of the emission data after offset within the current window. The sensor's reference noise floor constant. This is the sensitivity adjustment coefficient.

[0011] This invention introduces a causal fluctuation confidence level. When energy data fluctuates drastically but carbon emission data does not fluctuate accordingly, it identifies the lack of causal response between the two. This increases the distance cost in the subsequent alignment process, avoids the algorithm from making erroneous forced matching in pursuit of numerical closeness, and improves the physical and logical accuracy of the alignment.

[0012] Preferably, the distance cost between the energy sequence and the emission sequence satisfies the following relationship: ; In the formula, For the first in the energy sequence Point and emission sequence The cost of distance between points For the energy sequence in the 1st The value at each moment. For the emission sequence in the 1st The value at each moment. For the energy sequence at time Confidence level of causal fluctuations For the emission sequence at time Confidence level of causal fluctuations This is the penalty weighting coefficient.

[0013] This invention further ensures the accuracy of data fusion by incorporating the differences in causal fluctuation confidence in distance calculation, thereby forcing the fluctuation states of matching points to remain consistent.

[0014] Preferably, the observation noise covariance of the Kalman filter satisfies the following relationship: ; In the formula, For Kalman filtering in the first Dynamic observation noise covariance at each iteration The initial calibration noise covariance of the sensor, This is the local slope index of the optimal alignment path at the corresponding position. It is a natural constant. The distortion sensitivity coefficient, It is a nonlinear response exponent.

[0015] Preferably, the process of determining the local slope index includes: Obtain the current path point and the previous path point on the optimal alignment path; calculate the increment of the current path point and the previous path point in the energy sequence time axis direction, and the increment in the emission sequence time axis direction; calculate the maximum value of the ratio of the two increments as the local slope index. If the path points are diagonally stepped, the local slope index is set to 1.

[0016] Preferably, the basic physical transmission delay constant is obtained in the following way: Obtain the physical length of the exhaust gas pipeline and the average flow velocity of the gas inside the pipeline within the monitoring area; divide the physical length of the exhaust gas pipeline by the average flow velocity of the gas to obtain the basic physical transmission delay time; convert the basic physical transmission delay time into the corresponding number of data points according to the data sampling frequency, and use it as the basic physical transmission delay constant.

[0017] Preferably, the smoothing correction of the aligned emission data using the parameter-adjusted Kalman filter includes: When the local slope index indicates that the data time series is distorted, the weight of the current observation is reduced by increasing the observation noise covariance; the emission data is updated by relying on the state prediction value of Kalman filtering to obtain a smoothed carbon emission curve.

[0018] Preferably, the method further includes collecting environmental status data of the target area, including temperature, humidity, and vibration acceleration values ​​of key production equipment, wherein the vibration acceleration values ​​are used to assist in judging the operating status of the equipment.

[0019] Secondly, the present invention provides an energy and carbon dynamic monitoring system based on multi-source data fusion, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned energy and carbon dynamic monitoring method based on multi-source data fusion is implemented.

[0020] By adopting the above technical solution, a computer program for the energy and carbon dynamic monitoring method based on multi-source data fusion is generated and stored in a memory for loading and execution by a processor. Terminal devices are then manufactured based on the memory and processor for convenient use.

[0021] The beneficial effects of this invention are as follows: This invention incorporates physical causal logic into the dynamic time warping algorithm by constructing a causal fluctuation confidence index, thus overcoming the shortcomings of traditional algorithms that are driven solely by numerical values ​​and ignore physical causality. When energy data fluctuates drastically while carbon emission data fails to respond synchronously, the system automatically identifies the lack of causal relationship between the two and increases the distance cost to avoid erroneous matching. Simultaneously, it creatively maps the geometric slope of the warped path to the noise parameters of the Kalman filter, achieving adaptive coupling between alignment quality and filter weights. This dual mechanism of causal alignment and adaptive cleaning achieves high-precision dynamic monitoring through multi-source data fusion without relying on expensive, high-precision sensors, eliminating data phase errors and restoring the true carbon intensity. Attached Figure Description

[0022] Figure 1This is a flowchart of an energy and carbon dynamic monitoring method based on multi-source data fusion according to an embodiment of the present invention; Figure 2 This is a comparison chart of the consistency distribution of energy and carbon data association according to an embodiment of the present invention; Figure 3 This is a multi-source data dynamic time warping path mapping diagram according to an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] This invention discloses a method for dynamic monitoring of energy and carbon based on multi-source data fusion, referring to... Figure 1 This includes steps S1-S4: S1. Collect high-frequency energy data, low-frequency emission data, and environmental status data of the target area. Use interpolation algorithms to up-process the low-frequency emission data and standardize the high-frequency energy data and the up-processed low-frequency emission data to obtain energy and emission sequences with uniform time resolution.

[0026] Specifically, data from the target monitoring area is collected in real time through an industrial IoT gateway.

[0027] First, high-frequency energy data is acquired. Real-time current, voltage, and active power are read using smart meters or Hall effect current sensors installed in the main circuit of the production line. Because electrical signals respond extremely quickly, the high-frequency energy data is used as the system's reference time axis. The acquired raw energy sequence is denoted as... The sampling frequency is For example, setting the sampling frequency The frequency is 1Hz, meaning data is collected once per second.

[0028] Secondly, low-frequency emission data were acquired. Carbon dioxide concentration (ppm) was read using a non-dispersive infrared sensor installed on the chimney or exhaust outlet; flue gas velocity (m / s) was read using a differential pressure flow meter. The collected raw emission sequence was denoted as... The sampling frequency is Because gas sensors have a slow response, the sampling frequency is set... The frequency is 0.1Hz, meaning data is collected every 10 seconds.

[0029] Secondly, environmental status data is acquired by using environmental sensors deployed in the workshop to obtain temperature, humidity, and vibration acceleration values ​​of key production equipment. The vibration acceleration values ​​are used to help determine the operating status of the equipment.

[0030] because Direct calculation can lead to dimensionality mismatch. A cubic spline interpolation algorithm is used for low-frequency emission data. Frequency upsampling is performed by inserting nine fitting points between every two original emission data points to improve the data point density of low-frequency emission data and energy data. They are consistent, both at 1Hz.

[0031] Furthermore, the energy data and interpolated emission data are dimensionless. Since power and carbon concentration have different physical dimensions (e.g., power is 500kW while concentration is 400ppm), the numerical difference can be several orders of magnitude, making direct Euclidean distance calculation impossible. Therefore, standardization is used to eliminate these dimensional differences. Specifically, the arithmetic mean and standard deviation of the energy data and interpolated emission data are calculated separately using the following formulas: ; In the formula, The data values ​​before processing. The processed data value. This represents the arithmetic mean of the corresponding data. This represents the standard deviation of the corresponding data.

[0032] For example, if the arithmetic mean of energy data over a certain period of time... The standard deviation is 500. If the value is 50, and the current collected value is 550, then the processed value is... The same applies to processing emission data.

[0033] The processed energy sequence is denoted as The processed emission sequence is denoted as At this point, both the energy and emission sequences are dimensionless values, following a distribution with a mean of 0 and a variance of 1.

[0034] In this way, the frequency inconsistency problem was solved by interpolation and upsampling, and the difference in dimensions was eliminated by standardization, providing a unified and comparable data foundation for subsequent distance-based fusion calculations.

[0035] S2. Calculate the causal fluctuation confidence of the energy series at each time point. The causal fluctuation confidence represents the effectiveness of energy fluctuations in transferring carbon emissions. Combine the causal fluctuation confidence to calculate the distance cost between the energy series and the emission series, and construct the cumulative distance matrix.

[0036] It's important to note that traditional dynamic time warping algorithms only consider the magnitude difference of the numerical values ​​when calculating sequence distance. However, in industrial scenarios, there's a risk of spurious matches: energy data may fluctuate drastically over a period, while carbon emission data may not respond synchronously due to slow sensor response. Traditional algorithms might force a match based solely on numerical similarity, leading to alignment distortion. Therefore, this step introduces a causal fluctuation confidence level. .

[0037] Specifically, causal fluctuation confidence level The calculation satisfies the following relationship: ; In the formula, For time points Confidence level of causal fluctuation; For the energy sequence in the 1st The value at each moment; For the emission sequence in the 1st The value at each moment; The preset basic physical transmission delay constant; The length of the sliding window; This is the arithmetic mean of the energy data within the current window; This is the arithmetic mean of the emission data after offset within the current window; This is the sensor's reference noise floor constant, used to prevent the denominator from being zero; This is the sensitivity adjustment coefficient.

[0038] It should be further explained that the fundamental physical transmission delay constant The data is obtained as follows: the physical length of the exhaust gas pipeline and the average flow velocity of the gas inside the pipeline are obtained within the monitoring area; the physical length of the exhaust gas pipeline is divided by the average flow velocity of the gas to obtain the basic physical transmission delay time; the basic physical transmission delay time is converted into the corresponding number of data points according to the data sampling frequency, which serves as the basic physical transmission delay constant. For example, the exhaust gas duct is 30 meters long, the gas flow rate is 1 m / s, and the transmission delay time is 30 seconds, corresponding to 30 data points at a sampling frequency of 1 Hz. It is 30.

[0039] For example, the length of the sliding window Take 10 points, sensor reference noise floor constant Set the sensitivity adjustment coefficient to 0.01. Take 2.

[0040] The following calculation example demonstrates the logic behind calculating the confidence level for causal fluctuation: Assuming in the window Within this period, energy data experienced drastic fluctuations, with the local fluctuation amplitude, calculated as 4 in the numerator of the relational expression, indicating a lag. The carbon emission data after a certain time, assuming that carbon emissions do not fluctuate, the calculated local fluctuation range of the carbon emission data is 0.1. Substituting into the exponential function: Finally obtained The value is 1.

[0041] Confidence level of causal fluctuation A value of 1 indicates an extremely high degree of causal mismatch, meaning that current energy fluctuations have not been effectively transmitted to carbon emissions, and there is a lack of causal relationship between the energy and emission sequences.

[0042] Conversely, if the energy and emission sequences fluctuate synchronously, with similar numerator and denominator values, a small ratio, and a large exponential term, the confidence level of causal fluctuation is high. The value will approach 0, indicating a close causal relationship.

[0043] Furthermore, when constructing the cumulative distance matrix, the distance cost between sequence points is calculated. It satisfies the following relationship: ; In the formula, For the first in the energy sequence Point and emission sequence The cost of distance between points; For the energy sequence in the 1st The value at each moment; For the emission sequence in the 1st The value at each moment; For the energy sequence at time Confidence level of causal fluctuation; For the emission sequence at time Confidence level of causal fluctuation; The penalty weighting factor is set to 5 for example.

[0044] Continuing with the example above, if at time Confidence level of causal fluctuation A value of 1 indicates that at that moment, energy data fluctuates highly but does not match emissions data; while at that moment... Confidence level of causal fluctuation The value is 0.1, indicating that the emission data is stable at this moment.

[0045] The penalty item is: This means that even if the energy sequence and the emission sequence are numerically Euclideanly far apart... Even if the distance is very small, the final distance cost will be amplified by 5.5 times. When searching for the minimum path, the dynamic time warping algorithm will automatically avoid matching points with excessively high distance costs, thereby avoiding forcibly aligning noisy data with no causal relationship.

[0046] By introducing causal fluctuation confidence as a penalty term, the alignment process is ensured to not only conform to numerical similarity, but also to the causal logic of physical occurrence, effectively eliminating the interference of environmental noise on the alignment results.

[0047] S3. Use the dynamic time warping algorithm to search for the optimal alignment path between the energy sequence and the emission sequence in the cumulative distance matrix, and calculate the local slope index of the optimal alignment path at each path point.

[0048] Specifically, based on the cumulative distance matrix, a dynamic time warping algorithm is used to find a path from the starting point (1,1) to the ending point (n,n) of the matrix, such that the sum of the distance costs of all points on the path is minimized. This path is the optimal alignment path, denoted as a set of points. .

[0049] After obtaining the optimal alignment path, calculate the local slope index of the optimal alignment path at each point. The calculation formula is: ; In the formula, This represents the increment between the current path point and the previous path point along the energy sequence time axis. This represents the increment between the current path point and the previous path point along the emission sequence time axis. To avoid division by zero, which would result in vertical or horizontal shifts when the denominator is 0, in engineering practice, the denominator can be set to a minimum value or the length of the step direction can be directly taken.

[0050] To clearly illustrate the physical meaning of the local slope index, two typical cases are listed: In one embodiment of the present invention, when the energy sequence and emission sequence are properly aligned, the path point moves from (10, 10) to (11, 11), at which point the energy time axis increment... =1, emission time axis increment =1, Local slope index A value of 1 indicates that the data is synchronized well and there is no distortion.

[0051] In another embodiment of the invention, when there is a significant lag between the emission sequence and the energy sequence, the path point moves from (10,10) to (11,15). That is, in order to wait for a match, the emission data axis moves 5 steps while the energy data axis only moves 1 step. =1, It is 5. Local slope index A value of 5 indicates that the timeline was stretched by 5 times in order to align the data, resulting in severe time-series distortion.

[0052] By calculating the local slope index, the complex alignment path features are transformed into a single numerical value that can be quantitatively represented, accurately identifying the specific time periods when data experiences severe lag, packet loss, or abnormal stagnation, providing a direct basis for adjusting subsequent filter parameters.

[0053] S4. The observation noise covariance of the Kalman filter is dynamically adjusted according to the local slope index. The aligned emission data is then smoothed and corrected using the parameter-adjusted Kalman filter, and the energy and carbon monitoring results are output.

[0054] Specifically, while the data processed by the dynamic time warping algorithm is aligned in time, partial alignment is achieved by stretching the time axis. For example, in cases where emission sequences lag significantly behind energy sequences, the reliability of data at these points is low. This step utilizes the local slope index... Dynamically Adjusting the Observation Noise Covariance of the Kalman Filter .

[0055] Observation noise covariance The calculation satisfies the following relationship: ; In the formula, For Kalman filtering in the first Dynamic observation noise covariance during the next iteration; The initial calibration noise covariance of the sensor; This is the local slope index of the optimal alignment path at the corresponding position; It is a natural constant; The distortion sensitivity coefficient; It is a nonlinear response exponent.

[0056] For example, the initial calibration noise covariance of the sensor is set. The distortion sensitivity coefficient is 0.02. The nonlinear response exponent is 1.5. The value is 2.

[0057] Calculation example verification: In one embodiment of the invention, when the energy sequence and the emission sequence are properly aligned, i.e., the local slope index... When the value is 1: the internal structure of the logarithmic term is: , ,but The covariance remains at its initial value, the filter works normally, and the observations are trusted.

[0058] In another embodiment of the invention, when the emission sequence lags significantly relative to the energy sequence, i.e., a local slope index... When the value is 5: the internal form of the logarithmic term is: , ,but The observed noise covariance is approximately 4.7 times the initial value.

[0059] In Kalman filtering, the observation noise covariance An increase in the Kalman gain implies that the system considers the current observations inaccurate. Therefore, the Kalman gain will automatically decrease, and the state update formula will change. In this process, the weight of the difference between observed values ​​is reduced, and the system output will retain more of the trend of the predicted values.

[0060] When the local slope index This indicates that when data time series are distorted, the increase in observation noise covariance... The weight of the current observation is reduced; the emission data is updated based on the state prediction values ​​obtained by Kalman filtering to obtain a smoothed carbon emission curve.

[0061] Furthermore, during periods of good data quality, i.e. when the local slope index is 1, the detailed fluctuations of the data are preserved; during periods of poor data quality, i.e. when the local slope index is large and the time axis is forcibly aligned, the data is smoothed by increasing the noise parameter, which effectively suppresses the spikes and artifacts caused by the algorithm's forced alignment, and outputs a smooth, realistic carbon emission curve that conforms to physical laws.

[0062] Reference Figure 2 The left side of the figure shows the scatter plot of unprocessed energy and carbon monitoring data. Due to inconsistent data acquisition frequencies and transmission lags, the data points are loosely distributed, exhibiting obvious nonlinear hysteresis loop characteristics. Specifically, carbon emissions lag behind when energy decreases and fail to decrease when energy increases, indicating a severe time phase misalignment between energy consumption and carbon emission data. The right side of the figure shows the scatter plot after processing with the improved dynamic time warping and adaptive Kalman filtering of this invention. The data points are tightly converged and distributed around the linear regression trend line, exhibiting extremely high linear correlation. Figure 2 As can be seen, this invention eliminates phase error and restores the true physical mapping relationship between energy and carbon emissions.

[0063] Reference Figure 3In the figure, the solid line represents the optimal time warping path calculated by the algorithm of this invention, and the diagonal line represents the traditional linear alignment path. The marked points visually indicate the key areas where the algorithm makes adjustments. In this area, the optimal time warping path deviates significantly from the diagonal line, indicating that the algorithm automatically identifies the lag or lead of the data and achieves the best match by stretching or compressing the time axis, effectively overcoming the shortcomings of traditional linear alignment methods in dealing with nonlinear time distortion.

[0064] This invention also discloses an energy and carbon dynamic monitoring system based on multi-source data fusion, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the energy and carbon dynamic monitoring method based on multi-source data fusion of this invention.

[0065] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces. The configuration and function of these components are known in the art and will not be described in detail here.

[0066] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise expressly and specifically defined.

Claims

1. A method for dynamic monitoring of energy and carbon based on multi-source data fusion, characterized in that, include: High-frequency energy data and low-frequency emission data of the target area are collected. The low-frequency emission data is up-processed using an interpolation algorithm. The high-frequency energy data and the up-processed low-frequency emission data are then standardized to obtain energy and emission sequences with uniform time resolution. The causal fluctuation confidence score of the energy sequence at each time step is calculated. The causal fluctuation confidence score characterizes the effectiveness of energy fluctuations in transmitting carbon emissions. The distance cost between the energy sequence and the emission sequence is calculated by combining the causal fluctuation confidence score, and a cumulative distance matrix is ​​constructed. The optimal alignment path between the energy sequence and the emission sequence is obtained by searching the cumulative distance matrix using the dynamic time warping algorithm, and the local slope index of the optimal alignment path at each path point is calculated. The observation noise covariance of the Kalman filter is dynamically adjusted based on the local slope index. The aligned emission data is then smoothed and corrected using the parameter-adjusted Kalman filter, and the energy and carbon monitoring results are output.

2. The energy and carbon dynamic monitoring method based on multi-source data fusion according to claim 1, characterized in that, The standardization process for high-frequency energy data and up-frequency low-frequency emission data includes: Calculate the arithmetic mean and standard deviation of the high-frequency energy data and the up-frequency low-frequency emission data respectively; subtract the corresponding arithmetic mean from each data point and divide by the corresponding standard deviation to map each data point to the same distribution space, thereby obtaining the processed energy sequence and emission sequence.

3. The energy and carbon dynamic monitoring method based on multi-source data fusion according to claim 1, characterized in that, The causal fluctuation confidence level satisfies the following relationship: ； In the formula, For time points Confidence level of causal fluctuations For the energy sequence in the 1st The value at each moment. For the emission sequence in the 1st The value at each moment. The preset basic physical transmission delay constant, The length of the sliding window. This is the arithmetic mean of the energy data within the current window. This is the arithmetic mean of the emission data after offset within the current window. The sensor's reference noise floor constant. This is the sensitivity adjustment coefficient.

4. The energy and carbon dynamic monitoring method based on multi-source data fusion according to claim 3, characterized in that, The distance cost between the energy sequence and the emission sequence satisfies the following relationship: ； In the formula, For the first in the energy sequence Point and emission sequence The cost of distance between points For the energy sequence in the 1st The value at each moment. For the emission sequence in the 1st The value at each moment. For the energy sequence at time Confidence level of causal fluctuations For the emission sequence at time Confidence level of causal fluctuations This is the penalty weighting coefficient.

5. The energy and carbon dynamic monitoring method based on multi-source data fusion according to claim 1, characterized in that, The observation noise covariance of the Kalman filter satisfies the following relationship: ； In the formula, For Kalman filtering in the first Dynamic observation noise covariance at each iteration The initial calibration noise covariance of the sensor, This is the local slope index of the optimal alignment path at the corresponding position. It is a natural constant. The distortion sensitivity coefficient, It is a nonlinear response exponent.

6. The energy and carbon dynamic monitoring method based on multi-source data fusion according to claim 5, characterized in that, The process of determining the local slope index includes: Obtain the current path point and the previous path point on the optimal alignment path; calculate the increment of the current path point and the previous path point in the energy sequence time axis direction, and the increment in the emission sequence time axis direction; calculate the maximum value of the ratio of the two increments as the local slope index. If the path points are diagonally stepped, the local slope index is set to 1.

7. The energy and carbon dynamic monitoring method based on multi-source data fusion according to claim 3, characterized in that, The basic physical transmission delay constant is obtained as follows: Obtain the physical length of the exhaust gas pipeline and the average flow velocity of the gas inside the pipeline within the monitoring area; divide the physical length of the exhaust gas pipeline by the average flow velocity of the gas to obtain the basic physical transmission delay time; convert the basic physical transmission delay time into the corresponding number of data points according to the data sampling frequency, and use it as the basic physical transmission delay constant.

8. The energy and carbon dynamic monitoring method based on multi-source data fusion according to claim 1, characterized in that, The smoothing correction of the aligned emission data using the parameter-adjusted Kalman filter includes: When the local slope index indicates that the data time series is distorted, the weight of the current observation is reduced by increasing the observation noise covariance; the emission data is updated by relying on the state prediction value of Kalman filtering to obtain a smoothed carbon emission curve.

9. The energy and carbon dynamic monitoring method based on multi-source data fusion according to claim 1, characterized in that, It also includes collecting environmental status data of the target area, including temperature, humidity and vibration acceleration values ​​of key production equipment, which are used to help determine the operating status of the equipment.

10. A dynamic energy and carbon monitoring system based on multi-source data fusion, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement an energy and carbon dynamic monitoring method based on multi-source data fusion according to any one of claims 1-9.

Images