A carbon emission monitoring method based on big data

Abstract

The present application discloses a carbon emission monitoring method based on big data, the method comprising: using a monitoring resource division mechanism to divide a carbon emission monitoring area into several sub-areas; based on each sub-area, determining the two monitoring points with the smallest probability index in the same area as the key decision points, and using a linkage decision-making mechanism to obtain the target carbon emissions of the sub-area in the next cycle; based on the target carbon emissions of each sub-area in the next cycle, generating and executing the carbon emission monitoring resource allocation strategy of the carbon emission monitoring area in the next cycle; according to a preset time interval, periodically executing the above content to achieve periodic dynamic update of the monitoring resources of the carbon emission monitoring area. Thus, based on the reasonable division of sub-areas and the differentiated prediction of carbon emissions, the monitoring resources of each sub-area are dynamically and periodically updated to improve the differentiated monitoring efficiency of the carbon emission monitoring area.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication networks, and in particular to a carbon emission monitoring method based on big data. Background Art

[0002] At present, regional carbon emission monitoring is the process of monitoring and evaluating carbon emissions in a specific geographical area. This monitoring usually aims to understand the carbon emission level in the area so as to take measures to reduce carbon emissions or formulate policies to meet carbon reduction targets. This includes carbon emission monitoring of combustion sources such as factories, power plants, and heating boilers. By installing monitoring equipment, such as gas analyzers, chimney emission monitors, etc., the gas emissions of these emission sources can be monitored in real time, and the use of mobile edge computing can process and store data on edge devices close to the data source to reduce data transmission and latency, provide faster response and more efficient network resource utilization. In the existing technology, most of the computing resources are evenly distributed on each edge device, which leads to low computing efficiency when performing calculations. Some areas with high total carbon emissions are prone to insufficient computing resources. The above problems exist in the existing technology.

[0003] The Chinese invention patent application number 202311465777.0 discloses a mobile edge computing system and method for carbon emission monitoring. The system divides the area where carbon emission detection is required into several sub-areas, extracts the total historical carbon emissions in the sub-areas, and determines the representative edge monitoring device of each sub-area and its periodic influence coefficient and temperature influence coefficient. The system is substituted into the sub-area emission total estimation strategy to estimate the total carbon emissions in the sub-area, thereby allocating computing resources, further improving the allocation efficiency of computing resources, and avoiding the occurrence of insufficient computing resources in areas with high total carbon emissions.

[0004] However, the method of dividing sub-regions into equal areas cannot accurately reflect the carbon emission characteristics of different regions, which may lead to chaotic carbon emission patterns in the sub-regions, which is not conducive to the subsequent selection of representative edge devices and cannot reflect the differences between different edge devices. It may lead to large errors and limitations in carbon emission predictions. Since the carbon emission law characteristics of each edge device are not fully considered, the prediction results lack pertinence and accuracy. In addition, the prediction of sub-region carbon emissions based on the proportion of a single representative edge device has great limitations. Summary of the invention

[0005] This application provides a carbon emission monitoring method based on big data, dynamically and periodically updates the monitoring resources of each sub-region based on the reasonable division of sub-regions and differentiated prediction of carbon emissions, and improves the differentiated monitoring efficiency of carbon emission monitoring areas.

[0006] This application provides a carbon emission monitoring method based on big data, including:

[0007] S101, using a preset monitoring resource division mechanism, divide the monitoring points in the carbon emission monitoring area into a plurality of sub-areas, each sub-area including at least one monitoring point, and each monitoring point is used to periodically count the carbon emissions monitored at each preset time interval;

[0008] S102, based on each sub-region, the two monitoring points with the smallest probability index in the same region are determined as key decision points, and the target carbon emissions of the sub-region in the next cycle are obtained by using a pre-set linkage decision-making mechanism; the probability index in the same region is obtained by a monitoring resource division mechanism;

[0009] S103, based on the target carbon emissions of each sub-region in the next period, generating and executing a carbon emission monitoring resource allocation strategy for the carbon emission monitoring region in the next period;

[0010] S104, periodically executing steps S102 to S103 at preset time intervals to achieve periodic dynamic update of the monitoring resources in the carbon emission monitoring area.

[0011] Preferably, the pre-set monitoring resource division mechanism includes:

[0012] A1. Based on each monitoring point, obtain the carbon emissions recorded at each time node in the preset historical time period, generate several data points, the data points include the time node and its corresponding carbon emissions, fit the several data points, and generate the carbon emissions change curve corresponding to the monitoring point;

[0013] A2. Generate monitoring point pairs formed by combining all monitoring points in pairs, and obtain the key time node sequences formed by the two carbon emission change curves based on each monitoring point pair;

[0014] A3. Based on the key time node sequences of the two carbon emission change curves, a first similarity algorithm is used to calculate the first similarity characteristic value of the monitoring point pair;

[0015] A4, based on the direction values ​​corresponding to several truncation curves formed by the carbon emission change curve of each monitoring point in the monitoring point pair under the key time node sequence, a second similarity characteristic value of the monitoring point pair is calculated using a preset second similarity algorithm;

[0016] A5. Perform a weighted summation of the first similarity eigenvalue and the second similarity eigenvalue to obtain a co-region probability index of the monitoring point pair;

[0017] A6. Based on the same-area probability index of all pairs of monitoring points, the monitoring points whose same-area probability index is greater than a preset probability threshold are divided into a sub-area, and finally several sub-areas are generated.

[0018] Preferably, the key time node sequence formed by the two carbon emission change curves of the monitoring point pair is expressed as The key time node sequence includes at least two key time nodes, and the key time nodes are set as: the time points corresponding to the time starting point, time end point, and intersection of the two carbon emission change curves. n is the total number of key time nodes in the key time node sequence. As the starting point of time, is the end point of time, and the number of intersections is n-2.

[0019] Preferably, the carbon emission monitoring resource allocation strategy specifically includes:

[0020] D1. Sum the target carbon emissions of all sub-regions in the carbon emission monitoring area in the next cycle to obtain the target total carbon emissions of the carbon emission monitoring area in the next cycle;

[0021] D2. Calculate the ratio of the target carbon emissions of each sub-region to the total target carbon emissions to obtain the monitoring resource allocation ratio of each sub-region;

[0022] D3. Based on the monitoring resource allocation ratio of each sub-area in the next cycle, corresponding monitoring resources are allocated to each sub-area in advance.

[0023] Preferably, in S102, the pre-set linkage decision-making mechanism specifically includes:

[0024] S201, based on each decision key point in the sub-region, generate a carbon multivariate feature vector of the next cycle, input it into the corresponding pre-trained carbon emission prediction model, and output the decision result of the decision key point in the next cycle, the decision result includes the decision carbon emission and the decision ratio;

[0025] Among them, the carbon multivariate feature vector is expressed as [a, b, c], a represents the time attribute characteristics of the corresponding period, b represents the environmental parameter characteristics of the corresponding period, and c represents the volatility of the environmental parameter characteristics of the corresponding period compared with the previous period;

[0026] S202, based on the decision results of all decision key points in the sub-region in the next cycle, calculate the target carbon emissions of the sub-region in the next cycle.

[0027] Preferably, each of the monitoring points corresponds to a pre-trained carbon emission prediction model, and the carbon emission prediction model is obtained in the following manner:

[0028] B1. Collect a large number of carbon multivariate feature vectors corresponding to the historical periods of the monitoring point, label them to form a training data set, and the label content is set as: the actual carbon emissions monitored in the corresponding historical period, and the contribution ratio of the carbon emissions of the monitoring point to its sub-region in the corresponding historical period;

[0029] B2. Use the labeled training data set to train the pre-set neural network structure, select part of it as a validation set for verification, optimize the model parameters, and obtain the final carbon emissions prediction model.

[0030] Preferably, in S202, the target carbon emission of the sub-region in the next cycle is obtained according to the following calculation formula:

[0031] ,

[0032] Among them, Q is the target carbon emission of the sub-region in the next cycle, is the carbon emission of the decision-making key point m in the next cycle, is the decision ratio of the decision key point m in the next cycle, is the carbon emission of decision key point n in the next cycle, is the decision ratio of the decision key point n in the next cycle, and are the decision confidence factors of decision key points m and n respectively.

[0033] Preferably, the determination method of the decision confidence factor of each decision key point in the sub-region includes:

[0034] S301, based on each sub-region, calculate the average similarity between each decision key point and all other monitoring points in the sub-region except the decision key point:

[0035] ,

[0036] in, is the average similarity between the decision key point m and all other monitoring points in the subregion, is the average similarity between the decision key point n and all other monitoring points in the subregion, K is the total number of monitoring points in the subregion except the decision key point, is the probability index of the same area between the decision key point m and the kth monitoring point, is the probability index of the same area between the decision key point n and the kth monitoring point;

[0037] S302, calculating the decision confidence factor of the key decision point according to the following formula:

[0038] ,

[0039] in, and are the decision confidence factors of decision key points m and n respectively, is the probability index of the same area between decision key points.

[0040] Preferably, after S102, the method further includes: generating a prediction deviation tolerance range for each sub-region in the next cycle based on the target carbon emissions of each sub-region in the next cycle, specifically including:

[0041] S401, calculating the difference between the target carbon emission of the sub-region in the next cycle and the actual carbon emission of the previous cycle as the carbon emission cycle change value;

[0042] S402, generating a carbon multivariate feature vector of the sub-region in the next cycle, adding the carbon emission cycle change value as a new dimension to the carbon multivariate feature vector, and obtaining a four-dimensional feature vector of the sub-region in the next cycle;

[0043] S403, inputting the four-dimensional feature vector of the sub-region in the next cycle into the pre-trained deviation tolerance prediction model corresponding to the sub-region, and outputting the prediction deviation tolerance range of the sub-region in the next cycle;

[0044] Each sub-region corresponds to a pre-trained deviation tolerance prediction model. The deviation tolerance prediction model is obtained by:

[0045] C1. Collect a large number of historical periods with reasonable resource allocation in the corresponding sub-region and their corresponding four-dimensional feature vectors, and label them. The label content is set as: the difference between the actual carbon emissions in the corresponding historical period and the predicted target carbon emissions;

[0046] C2. Use the labeled training data set to train the pre-set neural network structure, select part of it as the validation set for verification, optimize the model parameters, and obtain the final deviation-tolerant prediction model.

[0047] Preferably, the S104 further includes:

[0048] S501, after the next cycle monitoring is completed, the difference between the actual carbon emissions of each sub-region in the cycle and its corresponding target carbon emissions is counted to obtain a prediction deviation value;

[0049] S502, based on each sub-region, compare the prediction deviation value of the cycle with its corresponding prediction deviation tolerance range. If the prediction deviation value is less than the prediction deviation tolerance range, continue monitoring the next new cycle; otherwise, mark the corresponding sub-region as abnormal.

[0050] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0051] Through the classification of monitoring points and sub-region division based on historical carbon emission patterns, the carbon emission characteristics of different regions can be reflected more accurately. Highly similar sub-regions make the carbon emission prediction more targeted and accurate, reducing the prediction error caused by the large differences in carbon emission characteristics of monitoring points within the sub-region; reasonable sub-region division allows computing resources to be differentially allocated according to the carbon emission characteristics and prediction needs of each sub-region, avoiding the waste of resources in areas with chaotic carbon emission characteristics, and improving resource utilization efficiency and monitoring efficiency; based on the reasonable division of sub-regions and differentiated prediction of carbon emissions, the monitoring resources of each sub-region are dynamically and periodically updated, so that the monitoring resources can match the actual carbon emissions, thereby improving the differentiated monitoring efficiency of carbon emission monitoring areas.

[0052] Based on all the key decision points in each sub-region, the introduction of carbon multivariate feature vectors and carbon emissions prediction models can more accurately predict the carbon emissions of each key decision point. The prediction model trained based on a large amount of historical period data can mine the historical monitoring rules of the key decision points, and based on the input carbon multivariate feature vector, identify the actual carbon emissions and contribution ratios corresponding to the highly similar period, and output the decision result that best matches the current carbon multivariate feature vector through model learning. The carbon multivariate feature vector introduces the time attribute characteristics of the period, the environmental parameter characteristics and the volatility of its periodic influence, and represents the diversified characteristics that affect carbon emissions within the period, which improves the robustness of the prediction and enhances the generalization ability and accuracy of model training. Based on the decision results of all key decision points determined in the sub-region and the influence of the local prediction value of the key decision point on the global prediction, a decision confidence factor is introduced, and the carbon emissions predicted from multiple limited perspectives in the sub-region are combined to avoid the limitations of a single perspective and improve the pertinence and accuracy of the total carbon emissions prediction in the sub-region.

[0053] By generating a four-dimensional feature vector for each sub-region and inputting it into the deviation tolerance prediction model, it is possible to more comprehensively consider the multiple factors affecting carbon emissions, and accurately evaluate the maximum error range allowed by the prediction model for each sub-region when making reasonable monitoring resource allocation based on historical cycle data, thereby identifying sub-regions with abnormalities within the cycle and taking remedial measures; the prediction deviation tolerance range provides a scientific basis for monitoring resource allocation. When the prediction deviation value exceeds the tolerance range, the decision confidence factor of the key decision point is adjusted in time to optimize the subsequent monitoring resource allocation and avoid waste or shortage of resources; the prediction model and monitoring strategy can be dynamically adjusted according to the actual monitoring results, which enhances the adaptability and flexibility of monitoring. When faced with complex and changeable carbon emission monitoring areas, it can better adapt to environmental changes and changes in monitoring needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Schematic diagram of the process of a carbon emission monitoring method based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0056] It should be noted that the terms “vertical”, “horizontal”, “up”, “down”, “left”, “right” and similar expressions used in this document are only for illustrative purposes and do not represent the only implementation method.

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0058] Embodiment 1: Figure 1 It is a flow chart of a carbon emission monitoring method based on big data according to an embodiment of the present invention.

[0059] like Figure 1 As shown, a carbon emission monitoring method based on big data includes the following steps:

[0060] S101, using a preset monitoring resource division mechanism, the carbon emission monitoring area is divided into several sub-areas, each sub-area includes at least one monitoring point, and each monitoring point is used to periodically count the carbon emissions monitored at preset time intervals, and the preset time interval is set to 24 hours.

[0061] Among them, the carbon emission monitoring area includes several monitoring points. The monitoring points are set as edge monitoring devices for real-time monitoring of carbon emissions. In order to more reasonably and differentiate the allocation of computing resources for statistical carbon emissions in the carbon emission monitoring area, it is necessary to accurately predict the carbon emissions of each sub-area in future time intervals. Therefore, a monitoring resource division mechanism is needed. Based on the historical carbon emission rules of each monitoring point, the resource allocation of the monitoring point is classified, so as to improve the accuracy of subsequent carbon emission prediction of each sub-area based on highly similar sub-area divisions, and then reasonably allocate the carbon emission computing resources allocated to each sub-area.

[0062] Specifically, the pre-set monitoring resource allocation mechanism includes:

[0063] A1. Based on each monitoring point, the carbon emissions recorded at each time node in a preset historical time period are obtained, and several data points are generated. The data points include the time nodes and their corresponding carbon emissions. Several data points are fitted to generate a carbon emission change curve corresponding to the monitoring point.

[0064] Among them, the setting of the preset historical time period is determined according to the actual situation. For example, the preset historical time period can be set to the past quarter; each time node is used to represent the time point when the monitoring point counts carbon emissions, and the difference between the two time nodes is the length of the preset time interval.

[0065] A2. Generate monitoring point pairs formed by combining all monitoring points in pairs, and obtain the key time node sequence formed by the two carbon emission change curves based on each monitoring point pair. .

[0066] Among them, the key time node sequence includes at least two key time nodes, and the key time nodes are set as: the time starting point, time end point, and time point corresponding to the intersection of the two carbon emission change curves (if the two carbon emission change curves have an intersection), expressed as , n is the total number of key time nodes in the key time node sequence, As the starting point of time, is the end point of time, and the number of intersections is n-2.

[0067] A3. Based on the key time node sequences of the two carbon emission change curves, a preset first similarity algorithm is used to calculate the first similarity characteristic value of the monitoring point pair.

[0068] Specifically, the first similarity algorithm is set as:

[0069] ,

[0070] in, is the first similar eigenvalue, is the deviation area value of the carbon emission change curve of the two monitoring points between the i-th key time node and the next key time node, is the average deviation area of ​​the carbon emission change curves of the two monitoring points (the average value of the area difference between the two curves at all adjacent time nodes), n is the total number of key time nodes in the key time node sequence, is the i-th key time node in the key time node sequence, is the i+1th key time node in the key time node sequence, and They are the carbon emission change curves of monitoring point a and monitoring point b, is the adjustment coefficient of the deviation area value, which is set according to the actual situation and expert experience to control the influence of the difference in the deviation area value on the first similarity eigenvalue.

[0071] A4. Based on the direction values ​​corresponding to the n-1 truncation curves formed by the carbon emission change curve of each monitoring point in the monitoring point pair under the key time node sequence, the second similarity characteristic value of the monitoring point pair is calculated using a preset second similarity algorithm.

[0072] Specifically, the second similarity algorithm is set as:

[0073] ,

[0074] in, is the second similarity characteristic value, D is the comprehensive direction difference value of the two carbon emission change curves of the monitoring point pair, is the directional difference value of the i-th truncation curve formed by the two carbon emission change curves of the monitoring point pair, for The contribution factor value to D, n is the total number of key time nodes in the key time node sequence, is the direction value of the i-th truncation curve of the carbon emission change curve of monitoring point a (set to the slope value of the truncation curve at the i-th key time node), is the direction value of the i-th truncation curve of the carbon emission change curve of monitoring point b, is the deviation area value of the carbon emission change curve of the two monitoring points between the i-th key time node and the next key time node, The deviation area value of the carbon emission change curves of the two monitoring points between the jth key time node and the next key time node, and the range of i and j is set to [1, n-1].

[0075] A5. Perform a weighted summation of the first similarity eigenvalue and the second similarity eigenvalue to obtain a co-region probability index of the monitoring point pair.

[0076] The weight values ​​corresponding to the first similarity feature value and the second similarity feature value are set according to actual conditions, and the sum of the two can be 1.

[0077] A6. Based on the same-area probability index of all pairs of monitoring points, the monitoring points whose same-area probability index is greater than a preset probability threshold are divided into a sub-area, and finally several sub-areas are generated, each of which includes at least one monitoring point.

[0078] Among them, the preset probability threshold can be set according to actual conditions and historical experience, and is used to define the probability value of two monitoring points being divided into a sub-area. For example, the preset probability threshold is set to 0.6 to ensure that the carbon emission characteristics of the monitoring points in the sub-area are highly similar.

[0079] S102: Based on each sub-region, the two monitoring points with the smallest probability index in the same region are determined as key decision points, and the target carbon emissions of the sub-region in the next cycle are obtained by using a pre-set linkage decision-making mechanism.

[0080] Specifically, in the sub-region, the two monitoring points with the lowest probability index in the same area are selected as the key decision points for the subsequent carbon emissions prediction. They represent the two monitoring points with the largest difference in the carbon emission change curves in the sub-region, and can fully reflect the maximum difference in carbon emission characteristics in the sub-region, avoiding large error deviations in the local prediction method of a single limited decision point, thereby affecting the allocation of monitoring resources in the entire region. In order to save monitoring computing resources and avoid predicting the carbon emissions of each monitoring point in the next cycle, local monitoring points are selected to predict the carbon emissions of the global region. Therefore, the selection of key decision points based on local to global predictions is extremely critical.

[0081] S103: Based on the target carbon emissions of each sub-region in the next period, a carbon emission monitoring resource allocation strategy for the carbon emission monitoring region in the next period is generated and executed.

[0082] Specifically, the carbon emission monitoring resource allocation strategy includes:

[0083] D1. Sum the target carbon emissions of all sub-regions in the carbon emission monitoring area in the next cycle to obtain the target total carbon emissions of the carbon emission monitoring area in the next cycle;

[0084] D2. Calculate the ratio of the target carbon emissions of each sub-region to the total target carbon emissions to obtain the monitoring resource allocation ratio of each sub-region;

[0085] D3. Based on the monitoring resource allocation ratio of each sub-region in the next cycle, corresponding monitoring resources are allocated to each sub-region in advance, the carbon emission characteristics between sub-regions are differentiated, and monitoring computing resources are allocated in a targeted manner. Based on accurate carbon emission predictions, the monitoring efficiency of the entire carbon emission monitoring area is improved.

[0086] S104, periodically executing steps S102 to S103 at preset time intervals to achieve periodic dynamic update of the monitoring resources in the carbon emission monitoring area.

[0087] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0088] By classifying monitoring points and dividing sub-regions based on historical carbon emission patterns, the carbon emission characteristics of different regions can be reflected more accurately. Highly similar sub-regions make carbon emission predictions more targeted and accurate, reducing prediction errors caused by large differences in carbon emission characteristics of monitoring points within sub-regions. Reasonable sub-region division allows computing resources to be differentially allocated according to the carbon emission characteristics and prediction needs of each sub-region, avoiding the waste of resources in areas with chaotic carbon emission characteristics and improving resource utilization efficiency and monitoring efficiency.

[0089] Based on the reasonable division of sub-regions and differentiated prediction of carbon emissions, the monitoring resources of each sub-region are dynamically and periodically updated so that the monitoring resources can match the actual carbon emissions and improve the differentiated monitoring efficiency of the carbon emission monitoring area.

[0090] Example 2: Predicting carbon emissions in a global area based on the prediction results of a single edge monitoring device may result in large errors due to the complexity and variability of the carbon emission monitoring area. Therefore, a more accurate carbon emissions prediction method is extremely critical, and the linkage decision-making mechanism in Example 1 is further limited.

[0091] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0092] In some embodiments, in step S102, the pre-set linkage decision-making mechanism specifically includes:

[0093] S201, based on each decision key point in the sub-area, generates a carbon multivariate feature vector for the next cycle, inputs it into the corresponding pre-trained carbon emission prediction model, and outputs the decision result of the decision key point in the next cycle, the decision result includes the decision carbon emissions and the decision ratio.

[0094] Specifically, the carbon multivariate feature vector is expressed as [a, b, c], where a represents the time attribute characteristics of the corresponding period, b represents the environmental parameter characteristics of the corresponding period, and c represents the volatility of the environmental parameter characteristics of the corresponding period compared with the previous period.

[0095] The time attribute feature of the corresponding period is set as follows: quantize according to the month, quantize January to December to 1 to 12 respectively as the integer part of the time attribute feature, and divide the month into the first ten days, the middle ten days, and the last ten days, corresponding to the quantization values ​​0.1, 0.5, and 0.9 respectively, as the decimal part of the time attribute feature; for example, the corresponding period is in July, and it is in the first ten days, so the time attribute feature of the corresponding period is expressed as 7.1.

[0096] The environmental parameter characteristics of the corresponding period are set as follows: the environmental parameter characteristics are any one or more specific parameters of temperature, humidity, and wind speed in the corresponding period. Each parameter can be quantified according to its actual value or weather forecast value. The quantified value is any one of the maximum value, minimum value, and average value of the environmental parameter in the corresponding period. The temperature can be quantified as a Celsius value, the humidity can be quantified as a percentage value, and the wind speed can be quantified as meters per second.

[0097] The environmental parameter characteristic volatility of the corresponding period compared with the previous period is set as: The environmental parameter characteristic volatility indicates the degree of change of the environmental parameter characteristics of the corresponding period compared with the previous period. The volatility is obtained by calculating the difference between the environmental parameter characteristics of the corresponding period and the previous period, and then dividing it by the environmental parameter characteristic value of the previous period.

[0098] Specifically, each monitoring point corresponds to a pre-trained carbon emission prediction model. The carbon emission prediction model is obtained in the following ways:

[0099] B1. Collect the carbon multivariate feature vectors corresponding to a large number of historical periods of the monitoring point, label them to form a training data set, and the annotation content is set as: the actual carbon emissions monitored in the corresponding historical period, and the contribution ratio of the monitoring point to the carbon emissions of its sub-region in the corresponding historical period (the ratio of the actual carbon emissions of the monitoring point in the corresponding historical period to the total carbon emissions of its sub-region in the same historical period).

[0100] B2. Use the labeled training data set to train the pre-set neural network structure, select part of it as a validation set for verification, continuously optimize the model parameters, and obtain the final carbon emissions prediction model.

[0101] Specifically, after obtaining the quantified carbon multivariate feature vector of the next cycle (the environmental parameter characteristics of the next cycle can be obtained according to the weather forecast software), it is input into the pre-trained carbon emission prediction model to predict the decision results of the next cycle. The detailed process of model prediction is as follows:

[0102] The quantized carbon multivariate feature vector [a, b, c] is input into the carbon emission prediction model; the carbon emission prediction model obtains all labels whose similarity (Euclidean distance) with the input carbon multivariate feature vector meets the preset similarity threshold (the preset similarity threshold is optimized to the optimal value during the training process to minimize the prediction error of the validation set) based on the input feature vector, and performs weighted summation on the two values ​​in all labels respectively, and combines them to output a decision result, i.e., the decision carbon emissions and decision ratio for the next period.

[0103] The decision results obtained can be used to formulate monitoring resource allocation strategies for carbon emission monitoring areas. According to the prediction results, monitoring resources can be allocated differentially so that resources can be more reasonably allocated to each sub-area, thereby improving monitoring efficiency.

[0104] S202, based on the decision results of all decision key points in the sub-region in the next cycle, calculate the target carbon emissions of the sub-region in the next cycle.

[0105] Specifically, the target carbon emissions of the sub-region in the next cycle are obtained according to the following calculation formula:

[0106] ,

[0107] Among them, Q is the target carbon emission of the sub-region in the next cycle, is the carbon emission of the decision-making key point m in the next cycle, is the decision ratio of the decision key point m in the next cycle, is the carbon emission of decision key point n in the next cycle, is the decision ratio of the decision key point n in the next cycle, and The decision confidence factors pre-set for decision key points m and n are used to reflect the influence of the predicted value of the decision key point on the global prediction. They are set according to the actual situation and expert experience, and the sum can be 1.

[0108] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0109] Based on all the key decision points in each sub-region, the introduction of carbon multivariate feature vectors and carbon emission prediction models can more accurately predict the carbon emissions of each key decision point. The prediction model trained based on a large amount of historical cycle data can mine the historical monitoring rules of key decision points, identify the actual carbon emissions and contribution ratios corresponding to the highly similar cycles based on the input carbon multivariate feature vectors, and output the decision results that best match the current carbon multivariate feature vectors through model learning.

[0110] The carbon multivariate feature vector introduces the time attribute characteristics of the cycle, the environmental parameter characteristics and the volatility of the cycle impact, and represents the diversified characteristics that affect carbon emissions within the cycle, which improves the robustness of the prediction and enhances the generalization ability and accuracy of model training;

[0111] Based on the decision results of all decision key points determined in the sub-region, and based on the influence of the local prediction value of the decision key point on the global prediction, a decision confidence factor is introduced. The carbon emissions predicted from multiple limited perspectives in the sub-region are combined to avoid the limitations of a single perspective and improve the pertinence and accuracy of the total carbon emissions prediction in the sub-region.

[0112] Embodiment 3: In embodiment 2, the decision confidence factor of each decision key point in the sub-region is not specifically determined. It may not be possible to differentiate the importance of the contribution of different decision key points to the prediction of the entire sub-region based solely on a pre-set fixed value.

[0113] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0114] In some embodiments, the decision confidence factor of each decision key point in the sub-region is determined by:

[0115] S301, based on each sub-region, calculate the average similarity between each decision key point and all other monitoring points in the sub-region except the decision key point, taking the decision key point m as an example:

[0116] ,

[0117] in, is the average similarity between the decision key point m and all other monitoring points in the subregion, K is the total number of monitoring points in the subregion except the decision key point, is the probability index of the same area between the decision key point m and the kth monitoring point.

[0118] Similarly, the average similarity of decision key point n is: ,

[0119] S302, calculating the decision confidence factor of the key decision point according to the following formula:

[0120] ,

[0121] in, and are the decision confidence factors of decision key points m and n respectively, is the probability index of the same area between decision key points, is the average similarity between the decision key point m and all other monitoring points in the subregion, is the average similarity between the decision key point n and all other monitoring points in the subregion.

[0122] Specifically, the average similarity is used to reflect the representativeness of decision key points m and n in the sub-region, that is, the degree of fit between them and the overall carbon emission characteristics of the sub-region. The larger the value, the stronger the representativeness. By comprehensively considering the similarity between decision key points and their average similarity with other points in the sub-region, the decision confidence factor can more accurately reflect the credibility and importance of decision key points in the sub-region carbon emission prediction. According to the size of the decision confidence factor, more attention is paid to decision key points with higher credibility, thereby improving the prediction accuracy.

[0123] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0124] By calculating the average similarity between each decision key point and other monitoring points in the sub-region, the representativeness of the decision key point for the sub-region is jointly determined based on the average similarity and the same-region probability index between the decision key points, and the decision confidence factor is set for the decision key point in each sub-region. The importance of the contribution of different decision key points to the sub-region prediction is differentiated; it more accurately reflects the credibility and importance of the decision key points in the sub-region carbon emission prediction, thereby improving the prediction accuracy.

[0125] Embodiment 4: Due to the complexity and variability of carbon emission monitoring areas, there is no mechanism to calibrate and verify the prediction results when each sub-area conducts periodic differential monitoring. If there is an obvious deviation error in the prediction index of a sub-area, error prediction may be carried out subsequently, resulting in the prediction deviation deviating more and more from reality, which is not conducive to the dynamic management and update of monitoring resources.

[0126] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0127] In some embodiments, after step S102, the method further includes: generating a prediction deviation tolerance range for each sub-region in the next cycle based on the target carbon emissions of each sub-region in the next cycle, specifically including:

[0128] S401, calculating the difference between the target carbon emission of the sub-region in the next cycle and the actual carbon emission of the previous cycle as the carbon emission cycle change value.

[0129] S402, generate the carbon multivariate feature vector of the sub-region in the next period (the method of generating the carbon multivariate feature vector can refer to Example 2, and the present invention will not elaborate on this), add the carbon emission periodic change value as a new dimension to the carbon multivariate feature vector, and obtain the four-dimensional feature vector of the sub-region in the next period.

[0130] The four-dimensional feature vector not only includes the time attribute characteristics, environmental parameter characteristics and their volatility, but also includes the periodic change characteristics of carbon emissions, thus more comprehensively reflecting the impact of periodic changes on carbon emissions.

[0131] S403, inputting the four-dimensional feature vector of the sub-region in the next cycle into the pre-trained deviation tolerance prediction model corresponding to the sub-region, and outputting the predicted deviation tolerance range of the sub-region in the next cycle.

[0132] It should be noted that each sub-region corresponds to a pre-trained deviation tolerance prediction model. The deviation tolerance prediction model is obtained in the following ways:

[0133] C1. Collect historical periods and their corresponding four-dimensional feature vectors in which a large number of monitoring resources in the corresponding sub-region are reasonably allocated (the allocated monitoring resources do not affect the monitoring efficiency of carbon emissions in the corresponding period), and label them. The label content is set to: the difference between the actual carbon emissions in the corresponding historical period and the predicted target carbon emissions.

[0134] C2. Use the labeled training data set to train the pre-set neural network structure, select part of it as the validation set for verification, continuously optimize the model parameters, and obtain the final deviation-tolerant prediction model.

[0135] In some embodiments, step S104 further includes:

[0136] S501, after the next cycle of monitoring is completed, the difference between the actual carbon emissions of each sub-region in the cycle and its corresponding target carbon emissions is counted to obtain a prediction deviation value.

[0137] S502, based on each sub-region, compare the prediction deviation value of the cycle with its corresponding prediction deviation tolerance range. If the prediction deviation value is less than the prediction deviation tolerance range, continue monitoring the next new cycle; otherwise, mark the corresponding sub-region as abnormal, adjust the decision confidence factor of the decision key point of the sub-region marked as abnormal, and then continue monitoring the next new cycle.

[0138] Specifically, the decision confidence factor of the decision key point of the sub-region marked with abnormality is adjusted, including:

[0139] Based on step S202, the decision confidence factor of the decision key point with a larger decision confidence factor is gradually increased or decreased according to the set unit value (which can be set to a smaller value, for example, 0.05) (the choice of adjustment direction of increase or decrease in the actual scenario can be selected based on which adjustment direction the deviation value is alleviated faster) until the difference between the calculated target carbon emissions and the actual carbon emissions is less than the predicted deviation tolerance range for the first time, and the decision confidence factor of the decision key point is updated.

[0140] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0141] By generating a four-dimensional feature vector for each sub-region and inputting it into the deviation tolerance prediction model, we can more comprehensively consider the multiple factors that affect carbon emissions, accurately evaluate the maximum error range allowed by the prediction model for each sub-region when making reasonable monitoring resource allocation based on historical cycle data, and thus identify sub-regions with abnormalities within the cycle and take remedial measures;

[0142] The prediction deviation tolerance range provides a scientific basis for the allocation of monitoring resources. When the prediction deviation value exceeds the tolerance range, timely adjustment of the decision confidence factor of the key decision point can optimize the subsequent monitoring resource allocation and avoid waste or shortage of resources. It can dynamically adjust the prediction model and monitoring strategy according to the actual monitoring results, enhance the adaptability and flexibility of monitoring, and better adapt to environmental changes and changes in monitoring needs when facing complex and changeable carbon emission monitoring areas.

[0143] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A carbon emission monitoring method based on big data, characterized in that: include: S101, using a preset monitoring resource division mechanism, divide the monitoring points in the carbon emission monitoring area into a plurality of sub-areas, each sub-area including at least one monitoring point, and each monitoring point is used to periodically count the carbon emissions monitored at each preset time interval; S102, based on each sub-region, the two monitoring points with the smallest probability index in the same region are determined as key decision points, and the target carbon emissions of the sub-region in the next cycle are obtained by using a pre-set linkage decision-making mechanism; the probability index in the same region is obtained by a monitoring resource division mechanism; The monitoring resource allocation mechanism includes: A1. Based on each monitoring point, obtain the carbon emissions recorded at each time node in the preset historical time period, generate several data points, the data points include the time node and its corresponding carbon emissions, fit the several data points, and generate the carbon emissions change curve corresponding to the monitoring point; A2. Generate monitoring point pairs formed by combining all monitoring points in pairs, and obtain the key time node sequence formed by the two carbon emission change curves based on each group of monitoring point pairs; A3. Calculate the first similarity feature value of the monitoring point pair using the first similarity algorithm; A4, based on the direction values ​​corresponding to several truncation curves formed by the carbon emission change curve of each monitoring point in the monitoring point pair under the key time node sequence, the direction value is set to the slope value of the truncation curve at the key time node, and the second similarity feature value of the monitoring point pair is calculated using the second similarity algorithm; A5. Perform a weighted summation of the first similarity eigenvalue and the second similarity eigenvalue to obtain the same-area probability index of the monitoring point pair; A6. Based on the same-area probability index of all pairs of monitoring points, the monitoring points whose same-area probability index is greater than a preset probability threshold are divided into a sub-area, and finally several sub-areas are generated; S103, based on the target carbon emissions of each sub-region in the next period, generating and executing a carbon emission monitoring resource allocation strategy for the carbon emission monitoring region in the next period; S104, periodically executing steps S102 to S103 at preset time intervals to achieve periodic dynamic update of monitoring resources in the carbon emission monitoring area.

2. The carbon emission monitoring method based on big data according to claim 1, characterized in that: The key time node sequence formed by the two carbon emission change curves of the monitoring point pair is expressed as: The key time node sequence includes at least two key time nodes, and the key time nodes are set as: the time points corresponding to the time starting point, time end point, and intersection of the two carbon emission change curves. n is the total number of key time nodes in the key time node sequence. As the starting point of time, is the end point of time, and the number of intersections is n-2.

3. The carbon emission monitoring method based on big data according to claim 1, characterized in that: The carbon emission monitoring resource allocation strategy specifically includes: D1. Sum the target carbon emissions of all sub-regions in the carbon emission monitoring area in the next cycle to obtain the target total carbon emissions of the carbon emission monitoring area in the next cycle; D2. Calculate the ratio of the target carbon emissions of each sub-region to the total target carbon emissions to obtain the monitoring resource allocation ratio of each sub-region; D3. Based on the monitoring resource allocation ratio of each sub-area in the next cycle, corresponding monitoring resources are allocated to each sub-area in advance.

4. The carbon emission monitoring method based on big data according to claim 1, characterized in that: In S102, the pre-set linkage decision-making mechanism specifically includes: S201, based on each decision key point in the sub-region, generate a carbon multivariate feature vector of the next cycle, input it into the corresponding pre-trained carbon emission prediction model, and output the decision result of the decision key point in the next cycle, the decision result includes the decision carbon emission and the decision ratio; Among them, the carbon multivariate feature vector is expressed as [a, b, c], a represents the time attribute characteristics of the corresponding period, b represents the environmental parameter characteristics of the corresponding period, and c represents the volatility of the environmental parameter characteristics of the corresponding period compared with the previous period; S202, based on the decision results of all decision key points in the sub-region in the next cycle, calculate the target carbon emissions of the sub-region in the next cycle.

5. The carbon emission monitoring method based on big data according to claim 4, characterized in that: Each of the monitoring points corresponds to a pre-trained carbon emission prediction model, and the carbon emission prediction model is obtained in the following manner: B1. Collect a large number of carbon multivariate feature vectors corresponding to the historical periods of the monitoring point, label them to form a training data set, and the label content is set as: the actual carbon emissions monitored in the corresponding historical period, and the contribution ratio of the carbon emissions of the monitoring point to its sub-region in the corresponding historical period; B2. Use the labeled training data set to train the pre-set neural network structure, select part of it as a validation set for verification, optimize the model parameters, and obtain the final carbon emissions prediction model.

6. The carbon emission monitoring method based on big data according to claim 4, characterized in that: In S202, the target carbon emission of the sub-region in the next cycle is obtained according to the following calculation formula: ， Among them, Q is the target carbon emission of the sub-region in the next cycle, is the carbon emission of the decision-making key point m in the next cycle, is the decision ratio of the decision key point m in the next cycle, is the carbon emission of decision key point n in the next cycle, is the decision ratio of the decision key point n in the next cycle, and are the decision confidence factors of decision key points m and n respectively.

7. The carbon emission monitoring method based on big data according to claim 6, characterized in that: The method for determining the decision confidence factor of each decision key point in the sub-area includes: S301, based on each sub-region, calculate the average similarity between each decision key point and all other monitoring points in the sub-region except the decision key point: ， in, is the average similarity between the decision key point m and all other monitoring points in the subregion, is the average similarity between the decision key point n and all other monitoring points in the subregion, K is the total number of monitoring points in the subregion except the decision key point, is the probability index of the same area between the decision key point m and the kth monitoring point, is the probability index of the same area between the decision key point n and the kth monitoring point; S302, calculating the decision confidence factor of the key decision point according to the following formula: ， in, and are the decision confidence factors of decision key points m and n respectively, is the probability index of the same area between decision key points.

8. The carbon emission monitoring method based on big data according to claim 4, characterized in that: After S102, the method further includes: generating a prediction deviation tolerance range for each sub-region in the next cycle based on the target carbon emissions of each sub-region in the next cycle, specifically including: S401, calculating the difference between the target carbon emission of the sub-region in the next cycle and the actual carbon emission of the previous cycle as the carbon emission cycle change value; S402, generating a carbon multivariate feature vector of the sub-region in the next cycle, adding the carbon emission cycle change value as a new dimension to the carbon multivariate feature vector, and obtaining a four-dimensional feature vector of the sub-region in the next cycle; S403, inputting the four-dimensional feature vector of the sub-region in the next cycle into the pre-trained deviation tolerance prediction model corresponding to the sub-region, and outputting the prediction deviation tolerance range of the sub-region in the next cycle; Each sub-region corresponds to a pre-trained deviation tolerance prediction model. The deviation tolerance prediction model is obtained by: C1. Collect a large number of historical periods with reasonable resource allocation in the corresponding sub-region and their corresponding four-dimensional feature vectors, and label them. The label content is set as: the difference between the actual carbon emissions in the corresponding historical period and the predicted target carbon emissions; C2. Use the labeled training data set to train the pre-set neural network structure, select part of it as the validation set for verification, optimize the model parameters, and obtain the final deviation-tolerant prediction model.

9. The carbon emission monitoring method based on big data according to claim 8, characterized in that: The S104 further includes: S501, after the next cycle monitoring is completed, the difference between the actual carbon emissions of each sub-region in the cycle and its corresponding target carbon emissions is counted to obtain a prediction deviation value; S502, based on each sub-region, compare the prediction deviation value of the cycle with its corresponding prediction deviation tolerance range. If the prediction deviation value is less than the prediction deviation tolerance range, continue monitoring the next new cycle; otherwise, mark the corresponding sub-region as abnormal.

Images