Personal carbon footprint real-time visual feedback system based on multi-source data fusion

Abstract

The invention relates to the technical field of carbon emission, and discloses a multi-source data fused personal carbon footprint real-time visual feedback system, which comprises a data acquisition module used for acquiring multi-source heterogeneous data from a plurality of independent data sources and representing user behaviors in real time; the processing module is connected with the data acquisition module and is used for carrying out real-time estimation on a composite state vector based on a preset state space model according to the multi-source heterogeneous data, and the composite state vector at least comprises a current activity state of a user and personalized carbon emission factors representing emission characteristics of the user; and calculating the personal carbon footprint based on the estimation result of the composite state vector. According to the method, the unified dynamic state space model is constructed, and the activity state, the physical parameters and the personalized emission factors of the user are integrated into the composite state vector for joint real-time estimation, so that the real-time performance, the continuity and the accuracy of personal carbon footprint accounting are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon emission, in particular to a multi-source data fusion personal carbon footprint real-time visualization feedback system. BACKGROUND

[0002] With the increasing global concern about climate change and the deepening of the concept of sustainable development, the demand for quantifying, tracking and managing personal carbon footprint is becoming increasingly urgent. For this reason, a variety of applications and digital tools have appeared in the prior art to help individual users calculate and understand their carbon emissions. These tools collect user behavior data and combine corresponding carbon emission factors to calculate in order to improve public environmental awareness and guide low-carbon lifestyles.

[0003] However, the existing technology still has its inherent limitations in the implementation of personal carbon footprint accounting and feedback. The current mainstream technical solution usually relies on manual input of data by users, or performs lagging and batch processing on isolated and discrete data sources such as monthly electricity bills and credit card consumption records. This calculation mode is scattered and discontinuous, treating each user activity as an independent and unrelated data point for static summation, making it difficult to capture the continuity and contextual relevance of user behavior in the time and space dimensions. Therefore, this approach not only cannot provide real-time carbon footprint feedback, but also limits the accuracy and credibility of the accounting results due to the neglect of behavior scenarios.

[0004] In addition, the existing technology generally uses standardized and universal emission factors from public databases when calculating carbon emissions. The inherent defect of this "one-size-fits-all" approach is that it cannot reflect significant differences between individuals. In fact, different users have different behavior patterns (such as driving habits), energy efficiency of tools used (such as vehicle models, home appliance energy efficiency levels), and energy structure of the geographical location, all of which have a decisive impact on their actual carbon emissions. Using uniform static factors for calculation inevitably leads to deviations between the final carbon footprint results and the user's actual situation, thereby weakening the personalized reference value and persuasiveness of feedback information to users.

[0005] Further, in terms of providing feedback to the user, the prior art solutions mostly stay in the aspect of retrospective information display. They usually present the total amount of carbon emissions and classified statistics of the user's past behavior to the user in the form of daily, weekly or monthly reports. This feedback mode is essentially passive, which only tells the user "what has happened", but fails to effectively guide the user "how to do better in the future". Due to the lag of feedback, when the user receives the information, the key behavioral decision-making of generating emissions has already been completed, and the system misses the best opportunity to exert positive influence at the key decision-making node. Therefore, how to transform the feedback information from a simple historical summary into an intelligent suggestion with foresight, which can guide the user to make low-carbon choices at the key moment, has become a technical problem to be solved in the field. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application provides a multi-source data fusion personal carbon footprint real-time visualization feedback system, which solves the problem of inaccurate calculation results and lack of personalized and forward-looking guidance ability of feedback caused by the single data processing method and static model of the existing personal carbon footprint accounting method.

[0007] To achieve the above purpose, the present application is implemented by the following technical scheme: a multi-source data fusion personal carbon footprint real-time visualization feedback system, comprising: a data acquisition module for acquiring real-time multi-source heterogeneous data representing user behavior from multiple independent data sources; a processing module connected with the data acquisition module, for real-time estimation of a composite state vector based on a preset state space model according to the multi-source heterogeneous data, wherein the composite state vector at least contains a user current activity state and a personalized carbon emission factor representing user emission characteristics; and calculating a personal carbon footprint based on the estimation result of the composite state vector; a feedback generation module connected with the processing module, for generating and outputting feedback information containing the personal carbon footprint.

[0008] Preferably, the processing module is specifically configured to perform recursive Bayesian filtering on the composite state vector by a set of weighted particles using the sequential Monte Carlo method, and calculate the expected value of the composite state vector at time t as its real-time estimation result by the following formula : ; wherein: represents the estimated value of the composite state vector at time t; represents the total number of particles in the particle set; an index of a single particle; denotes a state hypothesis of the i-th particle in the particle group at time t; denotes a normalized importance weight of the i-th particle at time t.

[0009] Preferably, one or more logically complete activity data segments are identified and extracted from the estimation result of the composite state vector; and based on the activity data segments, the personalized carbon emission factor is updated using Bayesian inference method, and the updating process follows the following proportional relationship: ; wherein: denotes the personalized carbon emission factor vector of user i to be updated; denotes the identified activity data segment; denotes the emission factor prior distribution parameter of the group g to which the user i belongs; denotes the posterior probability of the personalized carbon emission factor given the activity data segment and the group prior parameter; denotes the likelihood of generating the activity data segment given the personalized carbon emission factor; denotes the prior probability of the personalized carbon emission factor.

[0010] Preferably, the Bayesian inference method is based on a hierarchical Bayesian model, which at least includes a global parameter layer for representing general prior knowledge and an individual parameter layer for representing the individual user to be updated.

[0011] Preferably, based on the estimation result of the composite state vector, a key decision node of user behavior is identified in real time; At the decision node, based on the state transition equation in the state space model, a predictive factual trajectory reflecting user habits and at least one virtual counterfactual trajectory with low-carbon alternative behavior are simulated in parallel; The expected carbon emission difference between the factual trajectory and the counterfactual trajectory is calculated by the following formula , and based on the a decision signal for triggering active intervention is generated: ； wherein: represents the expected carbon emission difference, representing the potential carbon emission reduction; represents the predictive fact trajectory; represents the virtual counter-factual trajectory; represents a predefined carbon emission mapping function for calculating the total expected carbon emission according to the input trajectory.

[0012] Preferably, the feedback generation module is further configured to, upon receiving the decision signal, generate and output a scenarioized recommendation information containing the low-carbon alternative behavior as part of the feedback information.

[0013] Preferably, the composite state vector further contains a continuous physical state vector associated with the current activity state of the user, the continuous physical state vector at least including the geographical position coordinates and / or the moving speed of the user.

[0014] Preferably, the feedback generation module is configured to visualize the personal carbon footprint, and according to the current activity state of the user estimated by the composite state vector, attribute the carbon emission to different activity categories and then output.

[0015] A multi-source data fusion personal carbon footprint real-time visualized feedback method, comprising the following steps: S1, real-time acquisition of multi-source heterogeneous data representing user behavior from multiple independent data sources; S2, according to the multi-source heterogeneous data, real-time estimation of a composite state vector based on a preset state space model, wherein the composite state vector at least contains a current activity state of the user and a personalized carbon emission factor representing the emission characteristics of the user; S3, calculating a personal carbon footprint based on the estimation result of the composite state vector; S4, generating and outputting feedback information containing the personal carbon footprint.

[0016] The present application provides a multi-source data fusion personal carbon footprint real-time visualized feedback system. It has the following advantages: 1、The present application significantly improves the real-time, continuity and accuracy of personal carbon footprint accounting by constructing a unified dynamic state space model and jointly estimating the user's activity state, physical parameters and personalized emission factors in a composite state vector. Compared to traditional methods that rely on discrete and lagging data for segmented addition calculation, the present application uses the sequential Monte Carlo method to real-time integrate multi-source heterogeneous data streams and dynamically infer the user's behavior patterns in the complete spatio-temporal context. This method overcomes the disadvantages of data fragmentation and lack of context information, making the carbon footprint accounting no longer a static and isolated event accumulation, but a continuous evolution dynamic process closely coupled with user behavior, thereby providing more accurate and timely environmental impact awareness for users.

[0017] 2、The present application introduces a hierarchical Bayesian model-based emission factor adaptive correction mechanism, giving the system the ability to learn and evolve individually, thereby achieving accurate characterization of each user's emission characteristics. Traditional methods usually use universal and static public emission factors, which cannot reflect the differences between individuals. The present application estimates personalized emission factors as part of the state and continuously updates the posterior of this factor using complete activity data segments extracted from the user's real behavior as evidence. This enables the system's emission model to automatically and data-drivenly "learn" and adapt to the user's specific habits over time, ultimately providing a truly personalized carbon footprint assessment result.

[0018] The present application improves the system from a passive information recording and display tool to an intelligent advisor that can provide forward-looking and scenario-based decision guidance by integrating a proactive intervention opportunity identification function based on counterfactual simulation. Existing technologies are mostly limited to post-mortem analysis of historical behavior, while the present application can actively discover and evaluate the value of low-carbon behavior at key decision nodes of user behavior by simulating "fact trajectories" and "counterfactual trajectories" with low-carbon alternatives injected in parallel and quantifying their potential carbon emission reduction differences. This ability to provide quantitative recommendations before or during decision-making effectively transforms emission reduction information from retrospective reporting to forward-looking guidance, greatly enhancing the effectiveness of positive behavior intervention for users. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 The system architecture diagram of the present application; Figure 2 The method flowchart of the present application. DETAILED DESCRIPTION

[0020] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0021] Embodiments: Please refer to the accompanying Figure 1 The embodiment of the present application provides a multi-source data fusion personal carbon footprint real-time visualization feedback system, comprising: A data acquisition module is configured to acquire multi-source heterogeneous data representing user behavior from multiple independent data sources in real time. In the embodiment, the data acquisition module serves as the perception front end and data basis of the system, and its core responsibility is to systematically and continuously capture multi-source heterogeneous data from the physical and digital environment in which the user is located, and convert these raw and unstructured information streams into structured observation evidence that can be understood and utilized by subsequent processing modules.

[0022] In the specific implementation of the present application, the data acquisition module is designed as a highly modular and extensible set of software components. The module establishes stable and reliable communication links with multiple independent data sources through a series of pre-set application program interfaces (APIs), background services, and interactions with hardware drivers. These data sources provide data streams reflecting their daily activities to the system after explicit authorization by the user.

[0023] Preferably, the data sources cover at least the following dimensions to ensure comprehensive perception of user behavior: First, in terms of physical displacement perception, the data acquisition module integrates access capabilities for positioning services. The location-related data it acquires not limited to the latitude and longitude coordinates directly provided by Global Positioning System (GPS) satellites, but also compatible with auxiliary positioning technologies such as cellular network base station triangulation and Wi-Fi access point fingerprint positioning. This multi-mode positioning fusion strategy aims to ensure that the system can still obtain continuous and accurate location information even in indoor environments where GPS signals are weak or unavailable. These location information is the fundamental basis for subsequent processing modules to infer user travel modes (such as walking, driving, taking public transportation) and stay locations (such as home and office).

[0024] Second, in terms of consumption behavior perception, the data acquisition module is configured to securely interface with mainstream electronic payment gateways. By analyzing the observation data generated by payment records , the system is able to acquire key elements of the consumption behavior, such as the timestamp of the transaction, the transaction amount, and the crucial Merchant Category Code (MCC). The MCC provides a powerful semantic label to the system, allowing the processing module to make a preliminary characterization of the consumption behavior. For example, a payment record at a gas station is a strong indicator of an energy replenishment behavior, while a payment record at a supermarket implies a consumption activity related to food or daily supplies.

[0025] Furthermore, to acquire deeper contextual situational information, the data acquisition module can also access the operating system level of the mobile device to acquire privacy-processed application usage logs . The user's interaction behavior with a specific application often directly reflects their intention. For example, after the user opens a navigation application, the system can predict that they will soon start a trip activity; using a food delivery application is directly related to a food consumption activity. This information provides valuable prior knowledge for the processing module when making state transition predictions.

[0026] In addition, to directly quantify energy consumption in specific scenarios such as at home or at work, the data acquisition module also supports integration with the smart home Internet of Things ecosystem. Through communication with devices such as smart meters and smart sockets , the module can acquire real-time or cumulative power consumption data accurate to specific devices. This data allows the system to more finely decompose and account for carbon emissions generated by the general "home" activity.

[0027] It can be understood that the data generated by the above data sources has significant differences in format, timestamp accuracy, update frequency, and semantic level, i.e., it is highly heterogeneous. Therefore, a core technical task of the data acquisition module is to preprocess and normalize these heterogeneous data. The module contains a synchronization and buffering mechanism to handle the asynchronous arrival of data and align all acquired information to a unified system time axis.

[0028] More importantly, the data acquisition module is responsible for encapsulating each type of synchronized and parsed data into a standardized Observation Vector at each discrete time step . The vector is the only interface for the data acquisition module to deliver information to the processing module, and its structure is as follows: ; The Observation Vector here constitutes a The collection of all available evidence about the user's status at any given moment. It integrates previously unrelated and fragmented information from the physical and digital worlds into a unified data representation.

[0029] Finally, the continuous observation vector sequence output by the data acquisition module in this embodiment is , which is the basis for all high-level cognitive and inference functions in subsequent steps. Specifically, the observation vector sequence is the direct basis for the weight update of the sequential Monte Carlo filtering algorithm in step S2. The processing module calculates the state hypothesis With real observation Likelihood between , it is possible to select the state estimate that best matches the actual situation from among numerous possibilities. Therefore, the effective operation of the data acquisition module and the richness and standardization of the data it provides directly determine the upper limit and ability of the entire system to perform accurate state estimation, adaptive learning, and forward-looking feedback.

[0030] a processing module, connected to the data acquisition module, configured to estimate a composite state vector in real time based on a preset state space model according to multi-source heterogeneous data, wherein the composite state vector includes at least the user's current activity state and a personalized carbon emission factor representing the user's emission characteristics; and calculate the individual carbon footprint based on the estimated composite state vector; In this embodiment, the processing module constitutes the computing hub and decision-making core of the system. Its function is to receive the normalized, multi-source, heterogeneous observation data streams provided by the data acquisition module and perform in-depth, multi-layered analysis and processing on them, ultimately outputting a precise understanding of user status, adaptive learning results for system parameters, and forward-looking insights into future behavior.

[0031] In a preferred implementation, the processing module is not a single computing unit, but rather an integrated processing engine comprised of multiple collaborative, functionally coupled sub-functions. Its internal operating logic strictly adheres to the technical process designed by the present invention, transforming raw observation data into high-value, decision-making structured information.

[0032] When the processing module receives the observation vector sequence passed by the data acquisition module, it first performs the instantiation and initialization of the state space model. The fundamental purpose of this step is to build a mathematical framework that can fully describe the problem domain. A key technical feature is that the processing module defines a composite state vector (CompositeStateVector) , which does not simply describe the user's location or single activity, but creatively integrates implicit information from multiple dimensions into a unified estimation target: ; The composite state vector here , is a discrete state variable representing the user's macro activity, is a continuous physical parameter associated with the activity, and is a core innovation of the present application, a dynamic emission factor vector representing the user's personal emission characteristics, which evolves and learns over time. Embedding in the state vector allows its estimation to be jointly conducted with the estimation of the user's activity in the same probabilistic framework, thus solving the problem of static parameters and lack of personalization in traditional methods.

[0033] To drive the evolution of the state vector, the processing module constructs the system dynamics based on the state transition equation and the observation equation . Among them, the nonlinear function reflects the inherent laws and inertia of user behavior, and establishes a complex mapping relationship between the invisible internal state and the available external data.

[0034] Next, the core task of the processing module is to perform real-time joint state estimation. Given the complexity and variability of user behavior and the high uncertainty of observation data, the state space model exhibits typical nonlinear and non-Gaussian characteristics. Therefore, the present embodiment preferably uses the Sequential Monte Carlo (SMC) method, specifically a particle filter, to recursively infer the user's composite state.

[0035] During the inference process, the processing module maintains a set of a large number of (e.g., N) weighted random samples, i.e., "particles" . At each time step, the module performs prediction and update in a loop. In the prediction phase, it uses the state transition equation to propagate each particle from the past time to the current time. In the update phase, when the new observation vector arrives, the module evaluates the degree of agreement between each propagated particle and the true observation according to the observation equation , i.e., calculates the likelihood , and updates the importance weight of each particle accordingly.

[0036] After completing the weight update, the processing module obtains the expected estimation value of the composite state vector at time by performing a weighted average of the states of all particles: ; wherein,​ is a comprehensive estimate of the user's current most likely activity state, physical parameters, and personalized emission factors, and is the data cornerstone of all subsequent advanced functionalities of the system. represents the hypothesis of the state of the pair of particles, while is the quantification of the confidence of this hypothesis.

[0037] Further, in order to endow the system with the ability of learning and evolution, the processing module also implements an adaptive correction function of personalized carbon emission factors. This function aims to solve the inherent defect that the emission factor library is too universal and lacks individual accuracy. The processing module continuously analyzes the state sequence output by the above state estimation process, and uses the built-in pattern recognition algorithm to automatically identify and extract logically complete activity data segments from it. These segments, such as a complete driving trip, provide high-quality training samples rich in contextual information for parameter learning.

[0038] When such activity segments are obtained, the processing module starts a parameter update process based on a hierarchical Bayesian model. The mathematical principle of this process follows Bayes' theorem, and its update relationship is as follows: ; In this formula, is the personalized emission factor vector of the user to be updated, is the activity data segment as evidence, and is the prior distribution parameter inherited from a higher level (group level). The processing module obtains the posterior probability about by calculating the product of the likelihood and the prior probability . This data "calibrated" posterior distribution reflects the system's more accurate "belief" about the user's individual emission characteristics. Its expected value will then be used to update the part of the composite state vector, forming a closed loop of mutual promotion and continuous optimization between state estimation and parameter learning.

[0039] Finally, in order to realize the leap of the invention from passive recording to active guidance, the processing module also integrates an active intervention opportunity identification sub-function based on counterfactual simulation. The processing module monitors the state estimation sequence in real time to identify key decision nodes of the user's behavior.

[0040] At these nodes, the processing module uses the learned state transition model to conduct a parallel "thought experiment". It simulates the "fact trajectory" that the user is most likely to follow according to habits On the other hand, it simulates a parallel "counterfactual trajectory" by injecting a virtual low-carbon alternative (e.g., changing the mode of transportation from car to subway) .

[0041] Subsequently, the processing module invokes a predefined carbon emission mapping function to quantitatively assess the potential emissions of both trajectories and calculate their difference : ; This value intuitively represents the expected amount of emission reduction that can be achieved by adopting the low-carbon suggestion. When this value exceeds a dynamically adjusted threshold, the processing module determines that there is a high-value intervention opportunity and generates a decision signal to inform the subsequent feedback generation module.

[0042] In summary, the processing module in this embodiment is a highly integrated computing engine that not only achieves precise real-time tracking of user behavior, but also provides solid technical support for dynamic, personalized, and instructive carbon footprint feedback through internalized learning mechanisms and forward-looking simulation capabilities.

[0043] The feedback generation module is connected to the processing module and is used to generate and output feedback information containing personal carbon footprint.

[0044] In this embodiment, the feedback generation module serves as the final output terminal and human-computer interaction interface of the system. Its core function is to convert the highly abstract and structured data output by the processing module into feedback information that is intuitive, understandable, and instructive to the user. This module is a bridge connecting complex background computation and front-end user perception, and is a key link in realizing the closed-loop value of the system.

[0045] In a preferred implementation, the feedback generation module is designed as a dynamic information aggregation and presentation engine. Instead of simply displaying raw calculation results, it intelligently selects and constructs the most appropriate feedback strategy based on information type and current user context to achieve effective information delivery.

[0046] First, in terms of retrospective analysis and attribution presentation, the feedback generation module continuously receives real-time optimal estimates of the user's composite state vector from the processing module . A key technical step is that the module can dynamically combine two core pieces of information contained in this vector: the user's current most likely activity state and the adaptively corrected, highly personalized carbon emission factor vector .

[0047] By multiplying the specific activity state with its corresponding personalized emission factor , the feedback generation module is able to calculate the user's instantaneous carbon footprint at the current moment. These instantaneous values are accumulated over time, which constitutes the user's total carbon footprint at different time scales such as day, week, month, etc.

[0048] More importantly, this module utilizes the accurate identification of the user's activity state to achieve precise attribution of carbon emissions. Instead of providing only a general total emission number, it is able to decompose the total emissions into different predefined activity categories, such as {transportation, home energy consumption, dining consumption, shopping and entertainment}, etc. Furthermore, through visual components on the user interface, such as pie charts, bar charts or Sankey diagrams, this attribution relationship is clearly and intuitively presented to the user, allowing the user to easily understand the main sources of their personal carbon footprint, providing data support for subsequent behavior adjustment.

[0049] Secondly, a core innovation of the present invention is that the feedback generation module has the ability to generate forward-looking, scenario-based intervention information. This function aims to transform the role of the system from a passive post-recorder to an active, intelligent advisor that provides assistance at key decision points.

[0050] Specifically, the feedback generation module is configured to continuously monitor decision signals from the processing module. The generation of this decision signal is based on the result of the processing module performing counterfactual simulation and identifying high-value intervention opportunities. Once receiving this signal, the feedback generation module will activate its active intervention mode.

[0051] In this mode, the module will no longer present historical data, but generate a highly scenario-based recommendation information facing the future. The core content of this information comes from the expected carbon emission difference calculated by the processing module in the counterfactual simulation, whose calculation formula is as follows: ; Where, is the predicted user habitual trajectory, is the virtual trajectory with low-carbon replacement behavior injected, and is a predefined carbon emission mapping function.

[0052] The feedback generation module will integrate this quantified, potential carbon emission reduction benefit along with specific low-carbon replacement solutions into a concise and clear notification or suggestion. For example, at the decision node when the user is about to start a car commute, the system may push the following information: "If you choose to take the subway for this trip, it is expected to help you reduce grams of carbon emissions and a similar estimated travel time. Do you want a subway route planned for you? This type of advice, provided before or during decision-making and containing clearly quantified benefits, greatly increases the likelihood that users will adopt low-carbon behaviors.

[0053] Finally, to form a complete feedback loop capable of continuous learning and optimization, the feedback generation module is responsible for capturing user interactions with the generated information. Whether it's how long a user spends browsing a retrospective data report or responding to a prospective intervention suggestion (e.g., clicking "Accept" or "Ignore"), these interactions themselves are considered valuable new information that reflects user preferences and intentions.

[0054] In this embodiment, the feedback generation module encodes the captured interaction behavior data and feeds it back to the upstream module of the system as new observation evidence or control input. For example, the user's acceptance behavior can be regarded as a positive control input. , used to strengthen the probabilistic weighting of corresponding low-carbon behaviors in the state transition model within the processing module. In this way, the system learns from every interaction with the user, making its future state estimates more accurate, its parameter models more personalized, and its intervention recommendations more tailored to the user's actual needs and willingness to accept.

[0055] In summary, the feedback generation module in this embodiment is not only a window for information display, but also a human-computer interaction hub that integrates multiple functions such as data fusion, visual attribution, intelligent intervention and learning closed loop. It effectively converts complex background calculation results into insights and guidance that are of practical value to users. It is the ultimate executor and value embodiment of the overall technical concept of the present invention.

[0056] Reference Attachment Figure 2 Another embodiment of the present invention provides a method for real-time visualization feedback of personal carbon footprint by fusion of multi-source data, comprising the following steps: S1. Real-time acquisition of multi-source heterogeneous data representing user behavior from multiple independent data sources; S2. Based on the multi-source heterogeneous data and a preset state space model, a composite state vector is estimated in real time, where the composite state vector includes at least the user's current activity state and a personalized carbon emission factor representing the user's emission characteristics; S3. Calculate the personal carbon footprint based on the estimation result of the composite state vector; S4. Generate and output feedback information including personal carbon footprint.

[0057] The method of this embodiment can be used to execute the above system embodiment. Its principles and technical effects are similar and will not be described in detail here.

[0058] While embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and variations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time visual feedback system for personal carbon footprint based on multi-source data fusion, characterized by: include: The data acquisition module is used to acquire multi-source heterogeneous data representing user behavior from multiple independent data sources in real time; a processing module, connected to the data acquisition module, configured to estimate a composite state vector in real time based on the multi-source heterogeneous data and a preset state space model, wherein the composite state vector includes at least the user's current activity state and a personalized carbon emission factor representing the user's emission characteristics; and calculate the individual carbon footprint based on the estimated composite state vector; The feedback generation module is connected to the processing module and is used to generate and output feedback information including the personal carbon footprint.

2. The multi-source data fusion personal carbon footprint real-time visualization feedback system according to claim 1 is characterized in that: The processing module specifically adopts the sequential Monte Carlo method to perform recursive Bayesian filtering on the composite state vector through a set of weighted particles, and calculates the expected value of the composite state vector at time t as its real-time estimation result using the following formula: : ； in: represents the estimated value of the composite state vector at time t; represents the total number of particles in the particle group; is the index of a single particle; represents the state hypothesis of the i-th particle in the particle group at time t; represents the normalized importance weight of the i-th particle at time t.

3. The multi-source data fusion personal carbon footprint real-time visualization feedback system according to claim 1 is characterized in that: identifying and extracting one or more logically complete user activity data segments from the estimation result of the composite state vector; Based on the activity data fragment, the personalized carbon emission factor is updated a posteriori using the Bayesian inference method, and the updating process follows the following proportional relationship: ； in: represents the personalized carbon emission factor vector of user i to be updated; represents the identified activity data segment; Represents the prior distribution parameter of the emission factor of the group g to which the user i belongs; represents the posterior probability of the personalized carbon emission factor given the activity data segment and the population prior parameters; represents the likelihood of generating the activity data segment given the personalized carbon emission factor; represents the prior probability of the personalized carbon emission factor.

4. The multi-source data fusion personal carbon footprint real-time visualization feedback system according to claim 3 is characterized by: The Bayesian inference method is based on a hierarchical Bayesian model, which at least includes a global parameter layer for representing general prior knowledge and an individual parameter layer for representing individual users to be updated.

5. The multi-source data fusion personal carbon footprint real-time visualization feedback system according to claim 1 is characterized in that: Based on the estimation result of the composite state vector, identifying key decision nodes of user behavior in real time; At the decision node, based on the state transition equation in the state space model, a predictive factual trajectory reflecting the user's habits and at least one virtual counterfactual trajectory incorporating low-carbon substitution behavior are simulated in parallel; The expected carbon emission difference between the factual trajectory and the counterfactual trajectory is calculated by the following formula: , and based on the Generate decision signals for triggering proactive intervention: ； in: represents the expected carbon emission difference, indicating the potential carbon emission reduction; represents said predictive fact trajectory; represents the virtual counterfactual trajectory; Represents a predefined carbon emission mapping function, which is used to calculate the total expected carbon emissions based on the input trajectory.

6. The multi-source data fusion personal carbon footprint real-time visualization feedback system according to claim 5 is characterized by: The feedback generation module is further configured to, upon receiving the decision signal, generate and output contextual recommendation information including the low-carbon alternative behavior as part of the feedback information.

7. The multi-source data fusion personal carbon footprint real-time visualization feedback system according to claim 1 is characterized in that: The composite state vector further includes a continuous physical state vector associated with the current activity state of the user, wherein the continuous physical state vector includes at least the geographic location coordinates and / or movement speed of the user.

8. The multi-source data fusion personal carbon footprint real-time visualization feedback system according to claim 1 is characterized by: The feedback generation module is used to visualize the personal carbon footprint and output the carbon emissions after attributing them to different activity categories based on the user's current activity state estimated by the composite state vector.

9. A method for real-time visualization feedback of personal carbon footprint based on multi-source data fusion, according to any one of claims 1 to 8, characterized in that: The following steps are involved: S1. Real-time acquisition of multi-source heterogeneous data representing user behavior from multiple independent data sources; S2. Based on the multi-source heterogeneous data, a composite state vector is estimated in real time based on a preset state space model, wherein the composite state vector at least includes the user's current activity state and a personalized carbon emission factor representing the user's emission characteristics; S3. Calculating a personal carbon footprint based on the estimated result of the composite state vector; S4. Generate and output feedback information including the personal carbon footprint.

Images