Method for monitoring carbon emissions based on multi-sensor data

Abstract

The present application relates to the technical field of data processing, and more particularly to a method for monitoring energy carbon emission based on multi-sensor data, comprising: obtaining carbon emission concentration at a current time after energy consumption and carbon emission related data of each sensor; recording carbon emission related data of any sensor as target data, recording the current time and a plurality of previous times as a backtracking period, and determining the correlation weight of the target data in the backtracking period; determining the uncertainty measure value of the target data in the backtracking period; splitting the target data based on the decision tree to obtain a plurality of subsets, and determining the discrimination degree between the subsets after splitting based on the target data; constructing a splitting evaluation function of the decision tree; constructing and training a decision tree model to realize monitoring of energy carbon emission. The present application constructs a splitting evaluation function by comprehensively considering multi-sensor data, better adapts to multi-sensor data, and makes the monitoring result more accurately reflect the actual carbon emission situation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to an energy carbon emission monitoring method based on multi-sensor data. BACKGROUND

[0002] In the field of energy, carbon emission monitoring technology is constantly developing with the widespread application of multi-sensor technology. By deploying various sensors, the system can collect multi-dimensional carbon emission related data in real time, including temperature, humidity, energy consumption rate, gas concentration, etc. In the processing and analysis of carbon emission related data, the decision tree algorithm, due to its intuitive classification characteristics and strong interpretability, has become a commonly used data processing tool. Carbon emission monitoring not only needs to accurately identify the category and degree of carbon emission, but also needs to reasonably explain and analyze the monitoring results to provide strong support for energy saving and emission reduction decision-making, therefore, it is necessary to combine multi-dimensional data for subsequent classification and understanding.

[0003] However, the node splitting of traditional decision tree is based on a single information measurement method, such as information gain, information gain ratio or Gini index, etc. This single measurement method cannot fully exploit the rich information contained in the data when facing complex, high-dimensional and interrelated data collected by multi-sensor. For example, when monitoring carbon emissions in different energy production processes, the correlation between different sensor data will dynamically change with the changes in production process and environmental conditions, and the traditional splitting method is difficult to adapt to such changes, resulting in low classification accuracy of the decision tree; at the same time, the traditional decision tree does not fully consider the uncertainty and noise of multi-sensor data when splitting nodes, in actual monitoring, sensors may be disturbed by various factors, resulting in errors or outliers in the data, and the traditional splitting method is sensitive to these noises, which can easily make the generated decision tree model learn the wrong patterns, and thus affect the accuracy of carbon emission monitoring. SUMMARY

[0004] In order to solve the problem that in the process of energy carbon emission monitoring, the node splitting of traditional decision tree is based on a single information measurement method, which cannot adapt to complex, high-dimensional and interrelated data collected by multi-sensor, resulting in low classification accuracy of the decision tree, at the same time, the traditional decision tree is affected by noise, which makes the decision tree model learn the wrong patterns, and thus affects the accuracy of carbon emission monitoring, the present application proposes an energy carbon emission monitoring method based on multi-sensor data, which comprises the following steps:

[0005] The carbon emission concentration at the current moment after obtaining the energy consumption and the carbon emission related data of each sensor; the carbon emission related data of any sensor is recorded as target data, the current moment and a plurality of previous moments are recorded as a backtracking period, the correlation weight of the target data in the backtracking period is determined according to the correlation coefficient of the target data and the carbon emission concentration at each moment in the backtracking period, and the time difference between the current moment and each moment in the backtracking period; the uncertainty measurement value of the target data in the backtracking period is determined according to the information entropy and the coefficient of variation of the target data in the backtracking period; based on the target data, a decision tree is split to obtain a plurality of subsets, and the discrimination degree between the subsets after splitting based on the target data is determined according to the mean value of the carbon emission concentration corresponding to the moment included in each subset and the mean value of the carbon emission concentration corresponding to the moment included in the target data; a splitting evaluation function of the decision tree is constructed according to the correlation weight, the uncertainty measurement value, the discrimination degree, and the preset correction coefficient of each sensor; a decision tree model is constructed based on the splitting evaluation function and is trained, the newly obtained carbon emission related data of each sensor is input into the trained decision tree model, and the monitoring of energy carbon emission is realized.

[0006] The present application can more comprehensively reflect the characteristics of the data by introducing multiple information measurement methods, combining the correlation weight and uncertainty measurement of each sensor data, thereby improving the classification accuracy of the decision tree and the accuracy of carbon emission monitoring; can effectively process high-dimensional, complex and interrelated data collected by multiple sensors, overcome the dependence of traditional decision trees on single information measurement when splitting nodes, and make the model more adaptable; by introducing the correlation weight and uncertainty measurement in the backtracking period, the influence of noise on the decision tree model can be effectively reduced, the risk of learning incorrect patterns by the model can be reduced, and the reliability of the monitoring result can be improved; the splitting evaluation function constructed comprehensively considers the correlation weight, uncertainty measurement and discrimination degree, so that the splitting process of the decision tree is more scientific and reasonable, the data set can be more effectively divided, and the generalization ability of the model can be improved; by improving the accuracy of carbon emission monitoring, more reliable data support can be provided for enterprises and governments, which helps to develop more effective carbon emission reduction policies and promote the realization of sustainable development goals.

[0007] Further, the carbon emission concentration and the carbon emission related data of each sensor are carbon emission concentration and carbon emission related data of each sensor after data cleaning and normalization processing.

[0008] Further, the normalization processing adopts Z-score standardization.

[0009] Further, the carbon emission related data includes energy consumption data, temperature data, humidity data and gas concentration data.

[0010] Further, the correlation coefficient is obtained by: the difference between the target data and the carbon emission concentration in the backtracking period containing each time before and after the Pearson correlation coefficient is recorded as the correlation coefficient of the target data and the carbon emission concentration at each time in the backtracking period.

[0011] Further, the correlation weight satisfies:

[0012] ; In the formula, is the carbon emission related data of the first sensor in the correlation weight of the backtracking period, is the current time, is the first time in the backtracking period, is the number of times in the backtracking period, is the first sensor, is the correlation coefficient of the carbon emission related data and the carbon emission concentration at the first time in the backtracking period, is a preset time decay coefficient, is a proportional normalization function, is a natural exponential function.

[0013] The calculation of the correlation weight of the present application considers the time factor, and by introducing the time decay coefficient and the time difference, the influence degree of the data at different time points on the current monitoring result can be effectively reflected, so that the data weight of the recent time is higher, and the sensitivity of the model to the latest information is enhanced; by combining the correlation coefficients of different sensors and time decay, the information of multi-source data can be better integrated, the capture ability of the model to the carbon emission characteristics is improved, and the monitoring result is more comprehensive and accurate; by weighting the historical data, the data fluctuation caused by accidental factors can be smoothed, the sensitivity of the model to abnormal values is reduced, and the stability and robustness of the model are enhanced.

[0014] Further, the uncertainty measure value satisfies:

[0015] ; In the formula, is the uncertainty measure value of the carbon emission related data of the first sensor in the backtracking period, is the information entropy of the carbon emission related data of the first sensor in the backtracking period, is the coefficient of variation of the carbon emission related data of the first sensor in the backtracking period, is a preset weight adjustment parameter, is a proportional normalization function.

[0016] The uncertainty measurement value of the application combines information entropy and a coefficient of variation, can comprehensively reflect the performance of sensor carbon emission related data in the backtracking period, comprehensively considers the distribution characteristics and volatility of the data, and thus more accurately measures the uncertainty of the data; by introducing a weight adjustment parameter, the relative influence of information entropy and the coefficient of variation can be flexibly adjusted according to the actual situation, so that the uncertainty measurement can better adapt to different data characteristic scenes, increasing the application range of the method; by applying a proportional normalization function, the influence of data at different scales on uncertainty evaluation can be eliminated, so that the measurement value has comparability between different sensors, improving the reliability of the monitoring data; by quantifying the uncertainty, potential error sources can be better identified and located in the monitoring results, which helps to improve the overall precision and credibility of the monitoring system, and provides a basis for formulating more effective carbon emission reduction measures.

[0017] Further, the discrimination satisfies:

[0018] ; in the formula, is the discrimination between subsets after splitting based on carbon emission related data of the first sensor, is the number of subsets after splitting based on carbon emission related data of the first sensor, is the mean value of carbon emission concentration corresponding to the time included in the first subset after splitting based on carbon emission related data of the first sensor, is the mean value of carbon emission concentration corresponding to the time included in the first subset after splitting based on carbon emission related data of the first sensor, is the number of times included in the first subset after splitting based on carbon emission related data of the first sensor. The discrimination of the application quantifies the difference between subsets after splitting based on sensor carbon emission related data, can effectively reflect the dispersion degree of the carbon emission concentration mean value of each subset, and helps to identify efficient feature splitting points; by calculating the difference between each subset and the overall mean value, a more intuitive understanding can be provided for the decision-making process of the model, which helps to identify which data features have greater influence on carbon emission monitoring, thereby improving the transparency of the model. Further, the splitting evaluation function satisfies: ; in the formula,

[0019] The splitting evaluation function of the application quantifies the difference between subsets after splitting based on sensor carbon emission related data, can effectively reflect the dispersion degree of the carbon emission concentration mean value of each subset, and helps to identify efficient feature splitting points; by calculating the difference between each subset and the overall mean value, a more intuitive understanding can be provided for the decision-making process of the model, which helps to identify which data features have greater influence on carbon emission monitoring, thereby improving the transparency of the model.

[0020] Further, the splitting evaluation function satisfies:

[0021] ; in the formula, is the splitting evaluation function of the decision tree, ​​​​a number of sensors participating in the split evaluation function calculation, a correlation weight of carbon emission related data of the first sensor in the backtracking period, an uncertainty measurement value of carbon emission related data of the first sensor in the backtracking period, a discrimination degree between subsets after splitting based on carbon emission related data of the first sensor, a preset correction coefficient of the first sensor.

[0022] The split evaluation function of the decision tree of the present application comprehensively considers the correlation weight, the uncertainty measurement value, the discrimination degree and the preset correction coefficient, can comprehensively evaluate the effectiveness of feature splitting from multiple dimensions, and improves the scientificity of the splitting strategy; through comprehensive evaluation of the data of each sensor, the most valuable features for carbon emission monitoring can be more accurately identified, thereby effectively splitting in the tree structure, and finally improving the classification performance and accuracy of the decision tree.

[0023] Further, the training method adopts fold cross validation.

[0024] The present application has the following beneficial effects:

[0025] (1) By comprehensively considering the correlation, uncertainty and other factors of multi-sensor data to construct a split evaluation function, the potential patterns in the data can be more accurately mined; compared with the traditional single information measurement method, the present application can better adapt to the complex, high-dimensional and mutually related data collected by multiple sensors, thereby improving the classification accuracy of the decision tree on the carbon emission state, and making the monitoring result more accurately reflect the actual carbon emission situation.

[0026] (2) When calculating the uncertainty measurement value, the information entropy and the coefficient of variation of the data are considered, effectively dealing with the uncertainty and noise problems of the data, so that the improved decision tree has stronger resistance to noise and abnormal values in the sensor data, reduces the learning of error patterns, improves the stability and reliability of the model, and can output more reliable monitoring results even in the presence of interference.

[0027] (3) When constructing the split evaluation function, the preset correction coefficient of each sensor is introduced, which combines the business logic of carbon emission monitoring, such as the relationship between different energy types and carbon emission, the influence of production process on carbon emission, etc.; this makes the construction of the decision tree more in line with the actual monitoring demand, enhances the practicality of the monitoring result, and can provide more effective support for energy saving and emission reduction decision-making.

[0028] (4) Since the decision tree model is constructed by combining multi-sensor data and trained subsequently, it is clear to explain how the decision tree model classifies and monitors the carbon emission data, which is convenient for users to understand and analyze the monitoring results and assist decision making. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 is a step flow chart of the energy carbon emission monitoring method based on multi-sensor data according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be described clearly and completely below. The described embodiments are part of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0031] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0032] Please refer to Figure 1 which shows a step flow chart of the energy carbon emission monitoring method based on multi-sensor data according to an embodiment of the present application. The method comprises the following steps:

[0033] S1: Obtain the carbon emission concentration at the current time after energy consumption and the carbon emission related data of each sensor.

[0034] It should be noted that the carbon emission concentration is collected by a non-dispersive infrared analyzer, and the data of multiple sensors (including temperature sensors, humidity sensors, energy consumption sensors, and gas concentration sensors, etc.) are obtained at the same time. For example, the collection interval is 30 min / time.

[0035] Specifically, the carbon emission concentration and the carbon emission related data of each sensor are the carbon emission concentration and the carbon emission related data of each sensor after data cleaning and normalization processing.

[0036] Specifically, the normalization processing adopts Z-score standardization.

[0037] Specifically, the carbon emission related data includes energy consumption data, temperature data, humidity data, and gas concentration data.

[0038] S2: Record the carbon emission related data of any sensor as target data, record the current time and a number of previous times as a backtracking period, and determine the correlation weight of the target data in the backtracking period.

[0039] It's important to note that the correlation between sensor carbon emission data and carbon emission concentration is not static but dynamically changes over time. In carbon emission monitoring scenarios, factors such as different stages of production activities and changes in environmental conditions all affect the strength of the correlation between carbon emission data from various sensors and carbon emission concentration. For example, during peak industrial production periods, the correlation between carbon emission data from energy consumption sensors and carbon emission concentration may be stronger; while during equipment maintenance, carbon emission data from environmental sensors such as temperature and humidity will show a more significant correlation with carbon emission concentration. Furthermore, in actual monitoring, recent production processes and environmental conditions are more closely related to the current carbon emission situation. Data from a long time ago may no longer be representative due to changes in various factors. Therefore, combining a time decay factor allows for the dynamic and accurate capture of this time-varying data correlation characteristic.

[0040] The correlation weight of the target data in the backtracking period is determined based on the correlation coefficient between the target data and carbon emission concentration at each time point within the backtracking period, as well as the time difference between the current time and each time point within the backtracking period.

[0041] Specifically, the correlation coefficient is obtained as follows:

[0042] The difference between the target data and the carbon emission concentration at each point before and after the retrospective period is denoted as the correlation coefficient between the target data and the carbon emission concentration at each point in the retrospective period.

[0043] Specifically, the correlation weights satisfy:

[0044] ;

[0045] In the formula, For the first The correlation weight of carbon emission-related data from each sensor in the backtracking period. For the current moment, For the first time in the backtracking period At that moment, The number of moments within the backtracking period. For the first Carbon emission data from each sensor and carbon emission concentration within the retrospective period The correlation coefficient at each time point The preset time decay coefficient, For proportional normalization function, It is a natural exponential function.

[0046] The implementers can determine the time decay coefficient based on the total duration of the specific backtracking period. For example, if the total duration of the backtracking period is 7 days, the time decay coefficient is 0.1.

[0047] wherein, The greater, the stronger the correlation between the carbon emission related data of the sensor at that moment and the carbon emission concentration, and the greater the correlation weight, and vice versa; the data closer to the current moment has a greater value of the time decay factor , which means that the correlation coefficient between the recent sensor's carbon emission related data and the carbon emission concentration will be given a greater weight in calculating the correlation weight, and can better reflect the real association between the current sensor data and the carbon emission concentration. In the carbon emission monitoring scene, production processes, environmental conditions and other factors may change rapidly over time, and recent data can better reflect the current actual situation. By highlighting the importance of recent data through the time decay factor, it helps to make the calculated correlation weight more in line with the actual situation at the moment.

[0048] Example: For convenience of calculation, assume that the backtracking period is set to the last 4 time points (i.e. 1.5 hours of data in the past), the number of sensors is 2 (energy consumption sensor and temperature sensor), and the collected data is as shown in Table 1, the correlation coefficient As shown in Table 2 (Pearson correlation coefficient is prior art, which will not be described here), the calculation is as follows:

[0049] Table 1

[0050]

[0051] Table 2

[0052]

[0053] The weighted sum of the carbon emission related data of the energy consumption sensor is:

[0054] : ,

[0055] : ,

[0056] : ,

[0057] : ,

[0058] The sum is equal to: ;

[0059] The weighted sum of the carbon emission related data of the temperature sensor is:

[0060] : ,

[0061] : ,

[0062] : ,

[0063] : ,

[0064] Then the sum equals: ;

[0065] The correlation weight (after proportional normalization) is then: , .

[0066] S3: Determine the uncertainty measure of the target data during the backtracking period.

[0067] It should be noted that in actual multi-sensor carbon emission monitoring, the sources of data uncertainty are diverse. On the one hand, limitations in sensor accuracy and environmental interference can lead to significant data dispersion; on the other hand, the disordered distribution of data can also affect its reliability. Single measures of information entropy or dispersion cannot fully reflect these complexities. While information entropy reflects the degree of disorder in data, i.e., the chaotic state of data distribution, it cannot directly reflect the magnitude of data fluctuations; the coefficient of variation focuses on describing the dispersion of data relative to the mean, but lacks consideration of the overall pattern of data distribution. Therefore, combining both to calculate an uncertainty measure allows for a more comprehensive and accurate quantification of data uncertainty.

[0068] Based on the information entropy and coefficient of variation of the target data during the backtracking period, determine the uncertainty measure of the target data during the backtracking period.

[0069] Specifically, the uncertainty metric satisfies:

[0070] ;

[0071] In the formula, For the first The uncertainty measure of carbon emission-related data from individual sensors during the backtracking period. For the first The information entropy of carbon emission-related data from each sensor within the backtracking period. For the first The coefficient of variation of carbon emission-related data from each sensor during the backtracking period. The preset weight adjustment parameters, This is the proportional normalization function.

[0072] The calculation of information entropy and coefficient of variation is the prior art, which will not be described here.

[0073] The weight adjustment parameter can be set by the implementer according to the specific implementation, for example, 0.6.

[0074] Wherein, The coefficient of variation is used to measure the degree of disorder of the first sensor data, when increases, it means that the distribution of the carbon emission related data of the first sensor is more dispersed and disordered, in carbon emission monitoring, if the value range of the sensor data is very wide, and the occurrence probability of each value is relatively uniform, then the information entropy will be larger, which indicates that the sensor may have more noise or the regularity of the data is poor, then the uncertainty measure value of the sensor data is larger, on the contrary, when decreases, it means that the data is more concentrated and ordered, which means that the sensor data is relatively more reliable, then the uncertainty measure value of the sensor data is smaller; The coefficient of variation reflects the dispersion degree of the carbon emission related data of the first sensor relative to the mean value, if increases, it means that the dispersion degree of the carbon emission related data of the sensor relative to the mean value becomes larger, which indicates that the stability of the sensor data is poor, then the uncertainty measure value of the sensor data is larger, on the contrary, when decreases, it means that the dispersion degree of the carbon emission related data of the sensor relative to the mean value becomes smaller, the data is more stable, then the uncertainty measure value of the sensor data is smaller.

[0075] The data of the energy consumption sensor , according to the above data sequence, the coefficient of variation is 0.3258, the information entropy is 1.95, and the sum is equal to: ;

[0076] The data of the temperature sensor , according to the above data sequence, the coefficient of variation is 0.0117, the information entropy is 1.25, and the sum is equal to: ;

[0077] Then the uncertainty measure value (after proportional normalization) is: , .

[0078] S4: Based on the target data, a decision tree is used for splitting to obtain a plurality of subsets, and the discrimination degree between the subsets after splitting based on the target data is determined.

[0079] It should be noted that one of the criteria for evaluating the splitting effect of a decision tree is the degree of difference between the resulting subsets. High discriminative power between subsets indicates that the sensor data is effective in classifying different carbon emission characteristics, helping the decision tree to more accurately classify and predict carbon emission status. Traditional decision tree splitting methods rarely consider this subset discriminative power, while this invention, by quantifying discriminative power, can more comprehensively evaluate the effectiveness of the splitting. When the subsets formed after splitting a sensor data set show significant differences in the mean carbon emission data, it indicates that the sensor data has significant value in distinguishing different carbon emission levels or categories.

[0080] The discriminant strength between subsets split from the target data is determined based on the mean carbon emission concentration at the time corresponding to each subset and the mean carbon emission concentration at the time corresponding to the target data.

[0081] Specifically, the distinguishability satisfies:

[0082] ;

[0083] In the formula, For the first The discriminative power between subsets of carbon emission-related data from individual sensors after splitting. For the first The number of subsets of carbon emission-related data from each sensor after splitting. For the first The carbon emission-related data from the first sensor is split into the second... The subset contains the mean carbon emission concentration corresponding to the time interval. For the first The carbon emission data from each sensor includes the average carbon emission concentration at the corresponding time point. For the first The carbon emission-related data from the first sensor is split into the second... The number of moments contained in each subset.

[0084] in, The greater the difference, the more crucial the subset plays in distinguishing different carbon emission levels or categories. In this case, the greater the discriminative power between subsets after splitting the carbon emission-related data based on the sensor, and vice versa. The larger the value, the higher the weight of the subset with more data in the discrimination calculation. This means that the subset can more effectively reflect the difference between the subset and the whole, and thus affect the discrimination of dividing sensor data into different subsets.

[0085] Example: when splitting with the data of the energy consumption sensor, the split point is: energy consumption < 120, then subset 1 (< 120): the energy consumption value is , the corresponding carbon concentration is , the number is 2, then the mean value is ; subset 2 (≥ 120): the energy consumption value is , the corresponding carbon concentration is , the number is 2, then the mean value is ; then ;

[0086] When splitting with the data of the temperature sensor, the split point is: temperature < 25, then subset 1 (< 25): the temperature value is , the corresponding carbon concentration is , the number is 2, then the mean value is ; subset 2 (≥ 25): the temperature value is , the corresponding carbon concentration is , the number is 2, then the mean value is ; then ;

[0087] It should be noted that since the above data is simplified data under specific conditions, the subsets are completely the same after splitting, so the obtained discrimination is also the same, and there will be differences in actual application.

[0088] S5: Construct a split evaluation function of the decision tree.

[0089] It should be noted that the node split evaluation of the traditional decision tree often only relies on a single index, such as information gain, information gain ratio or Gini index, etc., which is difficult to fully mine the complex information relationship in multi-sensor data. And in the field of carbon emission monitoring, not only the characteristics of the data itself need to be considered, but also the business logic needs to be combined to achieve more accurate and effective split decisions. The correlation weight reflects the close degree of the sensor data and the carbon emission data, the uncertainty measure value embodies the reliability of the data, the discrimination measures the difference degree of the subsets after splitting based on the sensor data, and the correction coefficient of the business logic integrates the actual carbon emission monitoring business experience, and the split evaluation function is constructed by comprehensively combining these factors, which can provide a more comprehensive and accurate basis for node splitting, and meet the complex and variable actual needs of carbon emission monitoring.

[0090] According to the correlation weight, the uncertainty measure value, the discrimination, and the preset correction coefficient of each sensor, a split evaluation function of the decision tree is constructed.

[0091] Specifically, the split evaluation function satisfies:

[0092] ;

[0093] In the formula, For the split evaluation function of the decision tree, The number of sensors involved in the calculation of the split evaluation function. For the first The correlation weight of carbon emission-related data from each sensor in the backtracking period. For the first The uncertainty measure of carbon emission-related data from individual sensors during the backtracking period. For the first The discriminative power between subsets of carbon emission-related data from individual sensors after splitting. It is the first The preset correction coefficients for each sensor.

[0094] Implementers can set the preset correction coefficient of the sensor according to the specific implementation situation. For example, the energy consumption sensor is 0.8 and the temperature sensor is 0.4. Taking the energy consumption sensor as an example, since energy consumption and carbon emissions are closely related, such as large chemical enterprises, the changes in energy consumption during the production process have a rapid and significant impact on carbon emissions. Large-scale energy consumption is often the direct cause of increased carbon emissions. Therefore, the preset correction coefficient of the energy consumption sensor is 0.8.

[0095] in, Reflects the first The correlation between carbon emission-related data from individual sensors and carbon emission concentration. The larger the value, the more important the sensor data is for carbon emission monitoring, the greater its weight in the split assessment function, and the more significant its impact on the results of the split assessment function. It reflects the first The reliability of carbon emission-related data from individual sensors The smaller, The larger the value, the more reliable the sensor data is, and the greater its contribution to the results of the splitting evaluation function. Reflects the following according to the The degree of difference between subsets of carbon emission-related data from individual sensors after splitting. The larger the value, the better the splitting based on the sensor data can distinguish different carbon emission situations, and the greater its contribution to the splitting assessment function result. This means that after splitting the sensor data, the mean carbon emission data in each subset is not different, so the splitting evaluation result of this node is also 0. Based on business logic, the first The importance assessment of each sensor, when The larger, the more important the sensor data is in the carbon emission monitoring business, and the greater the influence on the result of the split evaluation function.

[0096] For example, the evaluation result of the energy consumption sensor is: The evaluation result of the temperature sensor is: ;

[0097] Compare the two evaluation results Therefore, the decision tree selects the energy consumption sensor and its corresponding split point (energy consumption < 120) to split the current node of the decision tree, and repeats the above operation for each node.

[0098] S6: Build and train a decision tree model to monitor carbon emissions.

[0099] Based on the split evaluation function, a decision tree model is built and trained, and the newly obtained carbon emission related data of each sensor is input into the trained decision tree model to monitor carbon emissions.

[0100] From the root node, input the carbon emission related data of the sensor after data cleaning and normalization into the decision tree, for each node, calculate the results of the split evaluation function corresponding to all possible features (i.e. carbon emission data of each sensor) and split points, select the feature and split point with the largest result of the split evaluation function for splitting, generate new child nodes, and continue to repeat the above splitting operation for the newly generated child nodes until the node data meets the stopping condition, such as the node data belonging to the same category, or the node data being less than a certain threshold (exemplarily, 10), the gain after splitting being less than a certain threshold (exemplarily, 0.1), etc.

[0101] Again use Fold cross validation (exemplarily, ) to train the decision tree model, each time the data set is divided into non-intersecting subsets, one subset is used as the test set and the remaining subsets are used as the training set, the decision tree model is trained and its accuracy on the test set is calculated, by adjusting the decision tree parameters (e.g. maximum depth, minimum sample size, etc.), the decision tree model with the highest accuracy in cross validation is selected as the final model (i.e. the trained decision tree model).

[0102] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for monitoring carbon emissions based on multi-sensor data, characterized in that, include: Acquire the carbon emission concentration at the current moment after energy consumption and the carbon emission data related to each sensor; The carbon emission data from any sensor is recorded as the target data, and the current time and several previous times are recorded as the backtracking period. Based on the correlation coefficient between the target data and the carbon emission concentration at each time within the backtracking period, and the time difference between the current time and each time within the backtracking period, the correlation weight of the target data in the backtracking period is determined. Based on the information entropy and coefficient of variation of the target data during the backtracking period, determine the uncertainty measure of the target data during the backtracking period; Based on the target data, a decision tree is used to split the dataset into several subsets. The discriminative power between the subsets after splitting based on the target data is determined by the mean carbon emission concentration corresponding to the time in each subset and the mean carbon emission concentration corresponding to the time in the target data. Based on the correlation weight, the uncertainty metric, the discrimination, and the preset correction coefficients of each sensor, a split evaluation function for the decision tree is constructed. A decision tree model is constructed based on the split evaluation function and trained. The carbon emission-related data of each newly acquired sensor are input into the trained decision tree model to realize the monitoring of carbon emissions.

2. The carbon emission monitoring method based on multi-sensor data according to claim 1, characterized in that, The carbon emission concentration and carbon emission correlation data of each sensor are the carbon emission concentration and carbon emission correlation data of each sensor after data cleaning and normalization.

3. The carbon emission monitoring method based on multi-sensor data according to claim 2, characterized in that, The normalization process uses Z-score standardization.

4. The carbon emission monitoring method based on multi-sensor data according to claim 2, characterized in that, The carbon emission-related data includes energy consumption data, temperature data, humidity data, and gas concentration data.

5. The energy carbon emission monitoring method based on multi-sensor data according to claim 1, characterized in that, The correlation coefficient is obtained as follows: The difference between the target data and the carbon emission concentration at each point before and after the retrospective period is denoted as the correlation coefficient between the target data and the carbon emission concentration at each point in the retrospective period.

6. The energy carbon emission monitoring method based on multi-sensor data according to claim 1, characterized in that, The correlation weights satisfy: ； In the formula, For the first The correlation weight of carbon emission-related data from each sensor in the backtracking period. For the current moment, For the first time in the backtracking period At that moment, The number of moments within the backtracking period. For the first Carbon emission data from each sensor and carbon emission concentration within the retrospective period The correlation coefficient at each time point The preset time decay coefficient, For proportional normalization function, It is a natural exponential function.

7. The carbon emission monitoring method based on multi-sensor data according to claim 1, characterized in that, The uncertainty metric satisfies: ； In the formula, For the first The uncertainty measure of carbon emission-related data from individual sensors during the backtracking period. For the first The information entropy of carbon emission-related data from each sensor within the backtracking period. For the first The coefficient of variation of carbon emission-related data from each sensor during the backtracking period. The preset weight adjustment parameters, This is the proportional normalization function.

8. The carbon emission monitoring method based on multi-sensor data according to claim 1, characterized in that, The distinguishability satisfies: ； In the formula, For the first The discriminative power between subsets of carbon emission-related data from individual sensors after splitting. For the first The number of subsets of carbon emission-related data from each sensor after splitting. For the first The carbon emission-related data from the first sensor is split into the second... The subset contains the mean carbon emission concentration corresponding to the time interval. For the first The carbon emission data from each sensor includes the average carbon emission concentration at the corresponding time point. For the first The carbon emission-related data from the first sensor is split into the second... The number of moments contained in each subset.

9. The energy carbon emission monitoring method based on multi-sensor data according to claim 1, characterized in that, The splitting evaluation function satisfies: ； In the formula, For the split evaluation function of the decision tree, The number of sensors involved in the calculation of the split evaluation function. For the first The correlation weight of carbon emission-related data from each sensor in the backtracking period. For the first The uncertainty measure of carbon emission-related data from individual sensors during the backtracking period. For the first The discriminative power between subsets of carbon emission-related data from individual sensors after splitting. It is the first The preset correction coefficients for each sensor.

10. The energy carbon emission monitoring method based on multi-sensor data according to claim 1, characterized in that, The training method adopts Cross-validation.

Images