A zero-carbon smart classroom carbon emission real-time monitoring platform and method based on edge computing and multi-source perception fusion

Abstract

The application discloses a zero-carbon smart classroom carbon emission real-time monitoring platform and method based on edge computing and multi-source sensing fusion, belongs to the field of smart classroom carbon management and edge computing technology, and through time synchronization, space alignment and exception elimination of the edge gateway on multi-source data, a trusted data frame is generated; based on the trusted data frame, the current activity scene of the classroom is identified, and a carbon emission accounting model is called according to the activity scene, so that the carbon emission and carbon emission intensity are calculated in real time; the accounting result, the trusted data frame abstract and the timestamp are subjected to hash operation and are chained to store the evidence, and the unalterable monitoring evidence record is generated. The application realizes the space-time alignment and trusted fusion of multi-source data on the edge side, improves the real-time performance, accuracy and verifiability of the classroom carbon emission monitoring, and reduces the privacy risk caused by the original personnel data uploading.

Description

Technical Field

[0001] This invention relates to the field of carbon management and edge computing technology for smart classrooms, and in particular to a real-time carbon emission monitoring platform and method for zero-carbon smart classrooms. Background Technology

[0002] With the advancement of green campuses, zero-carbon buildings, and smart classrooms, classroom carbon emission monitoring has become an important component of campus carbon management. Traditional classroom carbon emission monitoring typically relies on a single smart meter or manual meter reading, which can only obtain the total electricity consumption and cannot distinguish between different power sources such as mains electricity, solar power, and wind power, nor can it correlate carbon emissions with classroom activities, the number of people, or environmental changes. This extensive monitoring method has significant shortcomings.

[0003] First, direct fusion of data from multiple sensors is difficult. The smart meters, photovoltaic inverters, wind power meters, carbon dioxide sensors, people counting sensors, temperature and humidity sensors, and light sensors deployed in the classroom all have different sampling frequencies, communication protocols, clock references, and installation locations. For example, a smart meter might upload active power data every second, a carbon dioxide sensor might upload concentration values ​​every 10 seconds, and a people counting sensor might output the number of people every 5 seconds based on a video stream. Without time synchronization and spatial alignment at the edge, direct fusion will lead to data misalignment, resulting in distorted carbon emission accounting.

[0004] Secondly, sensor drift and abnormal data can severely impact monitoring results. Long-term sensor operation can lead to issues such as zero-point drift, decreased sensitivity, communication interruptions, electromagnetic interference, and fluctuating readings. For example, an uncalibrated carbon dioxide sensor may output abnormally high values, a people counting sensor may miscount under obstructed or backlit conditions, and smart meters may generate duplicate messages or packet loss during communication congestion. Without eliminating or correcting these abnormal data points, the correlation between carbon emission intensity and population data will lose its reliability.

[0005] Secondly, traditional carbon accounting methods lack differentiation based on activity scenarios. Classrooms may be in different states at different times, such as being empty, in class, studying, taking exams, or holding meetings. The personnel density, equipment operating status, air conditioning and lighting load, and carbon dioxide concentration changes vary significantly under different scenarios. If a uniform calculation is performed using a fixed period and a fixed allocation factor, it cannot reflect the true carbon emission intensity and is difficult to support refined management.

[0006] Finally, the monitoring results lack credible evidence storage and traceability verification. Data from traditional monitoring platforms is stored on local servers or cloud databases, which are at risk of being modified, deleted, or tampered with, making it difficult to serve as a reliable basis for campus carbon management, green building evaluation, or carbon asset development.

[0007] Therefore, there is a need for a real-time carbon emission monitoring platform and method for zero-carbon smart classrooms that can perform spatiotemporal alignment of multi-source data, anomaly removal, activity scene recognition, real-time carbon accounting, and trusted evidence storage at the edge. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a real-time carbon emission monitoring platform and method for zero-carbon smart classrooms based on edge computing and multi-source sensing fusion, in order to solve the problems of monitoring distortion, insufficient real-time performance, lack of scene differentiation, and difficulty in evidence storage and verification caused by multi-source sensor data drift.

[0009] The objective of this invention is achieved as follows:

[0010] A method for real-time carbon emission monitoring of a zero-carbon smart classroom based on edge computing and multi-source sensing fusion includes:

[0011] S1: Deploy multi-source sensors in the classroom to collect power data, carbon dioxide concentration data, personnel data, and environmental data; the power data includes mains power consumption data, photovoltaic power generation data, and / or wind power data.

[0012] S2: The edge gateway receives multi-source sensor data, performs time synchronization and spatial alignment on the multi-source sensor data, and generates a unified spatiotemporal data frame; it then performs anomaly removal on the unified spatiotemporal data frame to generate a reliable data frame.

[0013] S3: Based on the trusted data frame, identify the current activity scenario in the classroom through the activity scenario recognition model; and call the corresponding carbon emission accounting model according to the current activity scenario to calculate the real-time carbon emissions and / or carbon emission intensity.

[0014] S4: Perform a hash operation on the real-time carbon emissions and / or carbon emission intensity, trusted data frame digest, edge gateway identifier and timestamp to generate monitoring and evidence storage data, and upload the monitoring and evidence storage data to the blockchain or evidence storage platform to add a timestamp;

[0015] S5: Outputs real-time monitoring results, carbon emission intensity classification, and / or abnormal alarms.

[0016] Preferably, the multi-source sensors in step S1 include at least one of the following: smart meter, photovoltaic inverter, wind power meter, carbon dioxide sensor, people counting sensor, temperature and humidity sensor, light sensor, and access control sensor. The power data includes mains active power, photovoltaic power generation, wind power, voltage, current, and power quality data; the personnel data includes at least one of the following: number of people, number of people entering and exiting, and seat occupancy status.

[0017] Preferably, the time synchronization and spatial alignment in step S2 specifically include: synchronizing the clocks of multi-source sensors and edge gateways via NTP or PTP protocols; performing sliding window resampling and interpolation processing on data with different sampling frequencies to generate a data sequence with a unified timestamp; mapping the coordinates of each sensor to a classroom spatial grid or seating area to establish a mapping relationship between sensor identifiers, spatial locations, and classroom areas; and combining the time-aligned data with the spatially mapped data to generate a unified spatiotemporal data frame.

[0018] Preferably, the anomaly removal in step S2 specifically includes: performing anomaly detection for each data channel in the unified spatiotemporal data frame using at least one of the 3σ criterion, absolute median difference (MAD), Kalman filtering, or isolated forest algorithm; performing secondary verification of the anomaly detection results in conjunction with physical constraints, the physical constraints including at least one of non-negative power, carbon dioxide concentration range, non-negative number of people, and photovoltaic power not exceeding the rated power; removing or correcting the data points determined to be anomalies to generate a reliable data frame.

[0019] Preferably, the activity scene recognition model in step S3 specifically includes: acquiring the number of people, carbon dioxide concentration change rate, mains power, photovoltaic power, time information and timetable information in a reliable data frame; identifying the current activity scene in the classroom through a rule judgment model or machine learning classification model; the activity scene includes at least one of the following: vacant scene, class scene, self-study scene, exam scene, and meeting scene.

[0020] Preferably, the carbon emission accounting model in step S3 specifically includes: calculating the carbon emissions from mains electricity based on mains electricity consumption and grid emission factors; calculating the green electricity deduction based on photovoltaic power generation, wind power generation, and green electricity emission factors; calculating the net carbon emissions based on the mains electricity carbon emissions, other direct carbon emissions, and the green electricity deduction; and calculating the carbon emission intensity based on the net carbon emissions and the number of people, area, and time corresponding to the current activity scenario.

[0021] Preferably, the method further includes: performing time-series matching between the photovoltaic power generation curve and / or wind power generation curve and the classroom load curve to calculate the green electricity direct supply ratio; determining the green electricity deduction amount based on the green electricity direct supply ratio; and separately marking the green electricity that is not directly supplied according to the grid-connected electricity volume or green certificate method.

[0022] Preferably, the generation of monitoring and evidence storage data in step S4 specifically includes: performing hash operations on the trusted data frame, real-time carbon emissions, carbon emission intensity, activity scene label, edge gateway identifier, and timestamp to generate a data hash; combining multiple data hashes layer by layer according to a Merkle tree structure to generate a Merkle tree root hash; uploading the Merkle tree root hash to a blockchain or evidence storage platform and affixing a blockchain timestamp to generate an immutable monitoring and evidence storage record.

[0023] Preferably, it also includes: anonymizing, aggregating, or localizing personnel data at the edge gateway; uploading only the anonymized statistical results, carbon emission accounting results, and evidence hashes to the cloud, without uploading the original personnel images or identity information.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0025] (1) Multi-source data time synchronization and spatial alignment are achieved through edge gateways to solve the problems of different sampling frequencies and inconsistent spatial granularity;

[0026] (2) By eliminating anomalies and verifying physical constraints, the problem of monitoring distortion caused by sensor drift, communication interruption and abnormal readings is solved;

[0027] (3) Through activity scenario recognition, carbon emission intensity can be calculated differently according to scenarios such as class, self-study, exam, and vacancy;

[0028] (4) Reduce the privacy risks associated with uploading original personnel data by anonymizing the data at the edge;

[0029] (5) Improve the traceability and verifiability of monitoring results by storing evidence through hash and blockchain timestamps. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0031] Figure 1 This is a schematic diagram of the overall process of the method of the present invention.

[0032] Figure 2 This is a schematic diagram of the system framework of the present invention.

[0033] Figure 3 This is a schematic diagram of the spatiotemporal alignment and anomaly removal process of the edge gateway in this invention.

[0034] Figure 4 This is a schematic diagram of the activity scene identification and carbon emission accounting process of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] like Figure 1-4 The method for real-time carbon emission monitoring of a zero-carbon smart classroom based on edge computing and multi-source sensing fusion, as shown, specifically includes:

[0037] S1: Deploy multi-source sensors and collect data.

[0038] The following sensors are deployed in a smart classroom:

[0039] Smart meters: installed at the classroom power distribution box, they collect active power, voltage, current and energy of the mains power, with a sampling period of 1 second;

[0040] Photovoltaic inverter: installed at the photovoltaic grid connection point, it collects photovoltaic power generation and electricity generation data, with a sampling period of 5 seconds;

[0041] Wind power meter: installed at the wind power grid connection point, collects wind power and power generation, with a sampling period of 5 seconds;

[0042] Carbon dioxide sensor: installed in the middle of the classroom and at the return air vent, to collect CO2 concentration, with a sampling period of 10 seconds;

[0043] People counting sensor: installed at the classroom door and above the podium, to collect the number of people in the classroom and the number of people entering and leaving, with a sampling period of 5 seconds;

[0044] Temperature and humidity sensor: installed in the middle of the classroom, it collects temperature and humidity data, with a sampling period of 30 seconds;

[0045] Light sensor: installed near the window and in the podium area to collect light intensity, with a sampling period of 30 seconds;

[0046] Access control sensor: Installed at the classroom door to collect data on people entering and exiting.

[0047] The edge gateway receives the above data every 5 seconds. The edge gateway can be an industrial gateway or edge computing box with multi-protocol access capabilities, supporting protocols such as Modbus, MQTT, HTTP, and BACnet, and has local computing, local storage, and offline resume capabilities.

[0048] S2: The edge gateway performs time synchronization, spatial alignment, and anomaly removal.

[0049] Time synchronization: Clock synchronization between sensors and edge gateways is performed via the NTP protocol, with synchronization accuracy controlled within 50 milliseconds. For data with different sampling frequencies, sliding window resampling and linear interpolation are used to generate data sequences with uniform timestamps. For example, the average value of smart meter data sampled for 1 second is taken over a 5-second window, and CO2 data sampled for 10 seconds is linearly interpolated over a 5-second window to generate data frames with uniform time granularity.

[0050] Spatial Alignment: The classroom is divided into several spatial grids, such as the podium area, front row area, middle row area, back row area, window area, and equipment area. The coordinates of each sensor are mapped to the corresponding grid, establishing a mapping relationship between sensor identification, spatial location, and classroom area. For example, CO2 sensor A, installed in the middle of the classroom, is mapped to the middle row area; CO2 sensor B, installed at the return air vent, is mapped to the equipment area; the number of people sensor C, installed at the door, is mapped to the entrance area; and the light sensor D, installed near the window, is mapped to the window area.

[0051] A unified spatiotemporal data frame can be represented as:

[0052]

[0053] Where Pgrid represents mains power, Ppv represents photovoltaic power, Pwind represents wind power, and C represents the total power output. CO2 Where N is the carbon dioxide concentration, T is the number of people, H is the temperature, L is the light intensity, S is the spatial network identifier, and t is the unified timestamp.

[0054] Anomaly removal: Anomaly detection is performed on each data channel using the median absolute difference (MAD).

[0055]

[0056] like:

[0057]

[0058] Then determine x i Anomalies are identified, where k is a preset threshold, for example, 3. Further verification is performed using physical constraints, such as non-negative power, CO2 concentration between 400ppm and 5000ppm, non-negative number of people, photovoltaic power not exceeding rated power, and temperature within a reasonable range. For data points identified as anomalies, correction is performed using previous value hold, linear interpolation, or Kalman filter prediction. For data channels with multiple consecutive periods of anomalies, a sensor drift alarm is generated. After removing or correcting anomalies, a reliable data frame is generated.

[0059] S3: Activity scene recognition and real-time carbon accounting.

[0060] Based on trusted data frames, the number of participants, CO2 concentration change rate, mains power, photovoltaic power, time information, and class schedule information are extracted to identify the current activity scenario.

[0061] If the number of users is 0 and the mains power is below the standby threshold, it is determined to be an idle scenario;

[0062] If the number of participants is greater than 0 and the rate of increase in CO2 concentration exceeds the first threshold, it is determined to be a classroom or meeting scenario.

[0063] If the number of participants is stable and the mains power is stable, it is determined to be a self-study scenario.

[0064] If the course schedule is marked as an exam and the number of participants is greater than 0, it is determined to be an exam scenario.

[0065] Activity scene recognition models can employ either rule-based judgment models or machine learning classification models. An example of a rule-based judgment model is shown below:

[0066]

[0067] The carbon emission accounting model is as follows:

[0068] Carbon emissions from mains electricity:

[0069]

[0070] Green electricity credit amount:

[0071]

[0072] Net carbon emissions:

[0073]

[0074] Carbon emission intensity:

[0075]

[0076] Wherein, EFgrid is the grid emission factor, EFgreen is the green electricity emission factor, Eother is other direct carbon emissions, N is the number of people, A is the classroom area, and is the accounting period.

[0077] Furthermore, by time-matching the photovoltaic power generation curve and / or wind power generation curve with the classroom load curve, the proportion of direct green electricity supply is calculated:

[0078]

[0079] Where Pgreen(t) is the photovoltaic and / or wind power output curve, and Pload(t) is the classroom load curve. The green electricity deduction amount is determined according to the proportion of direct green electricity supply. Green electricity not directly supplied is marked separately according to the on-grid electricity volume or green certificate method to avoid double counting.

[0080] S4: Generate monitoring and evidence storage data.

[0081] The trusted data frame, real-time carbon emissions, carbon emission intensity, activity scenario label, edge gateway identifier, and timestamp are hashed.

[0082]

[0083] Multiple data hashes are combined layer by layer according to the Merkle tree structure to generate the Merkle tree root hash H. root The data is uploaded to a blockchain or evidence storage platform and stamped with a blockchain timestamp to generate an immutable monitoring and evidence storage record. The Merkle tree structure can be built according to time periods, such as generating a Merkle tree root hash every 5 minutes and a daily root hash every 24 hours, thus balancing evidence storage frequency and storage efficiency.

[0084] S5: Output monitoring results.

[0085] It outputs real-time carbon emissions, carbon emission intensity, activity scenarios, green electricity direct supply ratio, carbon emission intensity classification, and anomaly alarms. Anomaly alarms include sensor drift alarms, data interruption alarms, carbon emission intensity exceeding threshold alarms, and green electricity consumption anomaly alarms. Carbon emission intensity classification can be divided into three levels—low, medium, and high—based on preset thresholds, for example:

[0086] Low: CI < first threshold;

[0087] In the middle: First threshold ≤ CI < Second threshold;

[0088] High: CI ≥ Second threshold.

[0089] Example 2: Platform Example

[0090] Please see Figure 2 This embodiment provides a real-time carbon emission monitoring platform for a zero-carbon smart classroom based on edge computing and multi-source sensing fusion, including:

[0091] Multi-source sensing module: used to collect power data, carbon dioxide concentration data, personnel data, and environmental data;

[0092] Edge gateway module: used to perform time synchronization, spatial alignment and anomaly removal on multi-source sensor data, and generate reliable data frames;

[0093] Scene recognition and carbon accounting module: used to identify the current activity scene in the classroom and call the corresponding carbon emission accounting model to calculate the real-time carbon emissions and / or carbon emission intensity;

[0094] Evidence storage and output module: Used to generate monitoring evidence data and add blockchain timestamps, output real-time monitoring results and / or abnormal alarms.

[0095] Example 3: Application Case

[0096] A university's smart classroom is equipped with smart meters, photovoltaic inverters, CO2 sensors, attendance sensors, and edge gateways. The edge gateways collect data every 5 seconds to perform time synchronization, spatial mapping, and anomaly removal.

[0097] During a specific class period, the classroom had 45 students, a mains power output of 3.2 kW, a photovoltaic power output of 1.1 kW, and a CO2 concentration increase rate of 12 ppm / min. The edge gateway identified this as a class session, calculated the net carbon emissions and carbon intensity, and hashed the results onto the blockchain for verification. The system output a "moderate" carbon intensity rating and suggested increasing the proportion of direct green electricity supply or optimizing air conditioning operation strategies.

[0098] This example illustrates that the present invention can achieve reliable fusion of multi-source data and real-time carbon accounting at the edge, improving monitoring accuracy and verifiability.

[0099] The above description of the embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A zero-carbon smart classroom carbon emission real-time monitoring method based on edge computing and multi-source perception fusion, characterized in that, include: S1: Deploy multi-source sensors in the classroom to collect power data, carbon dioxide concentration data, personnel data, and environmental data; the power data includes mains power consumption data, photovoltaic power generation data, and / or wind power data. S2: The edge gateway receives multi-source sensor data, performs time synchronization and spatial alignment on the multi-source sensor data, and generates a unified spatiotemporal data frame; it then performs anomaly removal on the unified spatiotemporal data frame to generate a reliable data frame. S3: Based on the trusted data frame, identify the current activity scene in the classroom using the activity scene recognition model; Based on the current activity scenario, the corresponding carbon emission accounting model is invoked to calculate the real-time carbon emissions and / or carbon emission intensity. S4: Perform a hash operation on the real-time carbon emissions and / or carbon emission intensity, trusted data frame digest, edge gateway identifier and timestamp to generate monitoring and evidence storage data, and upload the monitoring and evidence storage data to the blockchain or evidence storage platform to add a timestamp; S5: Outputs real-time monitoring results, carbon emission intensity classification, and / or abnormal alarms.

2. The method of claim 1, wherein, The multi-source sensors mentioned in step S1 include at least one of the following: smart meters, photovoltaic inverters, wind power meters, carbon dioxide sensors, people counting sensors, temperature and humidity sensors, light sensors, and access control sensors; The power data includes active power of mains power, photovoltaic power generation, wind power generation, voltage, current, and power quality data; the personnel data includes at least one of the following: number of people, number of people entering and exiting, and seat occupancy status.

3. The method of claim 1, wherein, The time synchronization and spatial alignment described in step S2 specifically include: Clock synchronization of multi-source sensors and edge gateways via NTP or PTP protocols; Data with different sampling frequencies are resampled and interpolated using a sliding window to generate a data sequence with a unified timestamp. Map the coordinates of each sensor to the classroom space grid or seating area to establish a mapping relationship between sensor identification, spatial location and classroom area; The time-aligned data is combined with the spatially mapped data to generate a unified spatiotemporal data frame.

4. The method of claim 1, wherein, The anomaly removal described in step S2 specifically includes: For each data channel in the unified spatiotemporal data frame, at least one of the following algorithms is used for anomaly detection: 3σ criterion, absolute median difference (MAD), Kalman filtering, or isolated forest algorithm. The abnormal detection results are verified by combining physical constraints, which include at least one of the following: non-negative power, carbon dioxide concentration range, non-negative number of people, and photovoltaic power not exceeding the rated power. Data points identified as abnormal are removed or corrected to generate reliable data frames.

5. The method of claim 1, wherein, The activity scene recognition model mentioned in step S3 specifically includes: Obtain the number of people, carbon dioxide concentration change rate, mains power, photovoltaic power, time information, and timetable information from trusted data frames; The current activity scenario in the classroom can be identified through rule-based judgment models or machine learning classification models. The activity scenarios include at least one of the following: vacant scenarios, class scenarios, self-study scenarios, exam scenarios, and meeting scenarios.

6. The method of claim 1, wherein, The carbon emission accounting model mentioned in step S3 specifically includes: Calculate the carbon emissions from mains electricity based on mains electricity consumption and grid emission factors; The green electricity deduction is calculated based on photovoltaic power generation, wind power generation, and green electricity emission factors; Calculate the net carbon emissions based on the mains electricity carbon emissions, other direct carbon emissions, and the green electricity credits. The carbon emission intensity is calculated based on the net carbon emissions and the number of people, area, and time corresponding to the current activity scenario.

7. The method of claim 6, wherein, Also includes: The proportion of direct green electricity supply is calculated by matching the photovoltaic power generation curve and / or wind power generation curve with the classroom load curve in a time sequence. The amount of green electricity deduction is determined based on the green electricity direct supply ratio, and green electricity that is not directly supplied is marked separately according to the amount of electricity fed into the grid or the green certificate method.

8. The method according to claim 1, characterized in that, The generation of monitoring and evidence storage data in step S4 specifically includes: A hash operation is performed on the trusted data frame, real-time carbon emissions, carbon emission intensity, activity scene label, edge gateway identifier, and timestamp to generate a data hash; Multiple data hashes are combined layer by layer according to the Merkle tree structure to generate the Merkle tree root hash; The Merkle tree root hash is uploaded to a blockchain or evidence storage platform and stamped with a blockchain timestamp to generate an immutable monitoring and evidence storage record.

9. The method according to claim 1, characterized in that, Also includes: Personnel data can be anonymized, aggregated, or localized at the edge gateway. Only the anonymized statistical results, carbon emission accounting results, and evidence hashes are uploaded to the cloud; the original personnel images or identity information are not uploaded.

10. A real-time carbon emission monitoring platform for a zero-carbon smart classroom based on edge computing and multi-source sensing fusion, used to execute the method according to any one of claims 1-9, characterized in that, include: The multi-source sensing module is used to collect power data, carbon dioxide concentration data, personnel data, and environmental data. The edge gateway module is used to perform time synchronization, spatial alignment, and anomaly removal on multi-source sensor data to generate reliable data frames; The scene recognition and carbon accounting module is used to identify the current activity scene in the classroom and call the corresponding carbon emission accounting model to calculate the real-time carbon emissions and / or carbon emission intensity. The evidence storage and output module is used to generate monitoring evidence data and add blockchain timestamps, and output real-time monitoring results and / or abnormal alarms.

Images