Multi-parameter time sequence fusion-based intelligent identification method for calibration state of CEMS

Abstract

This application discloses a multi-parameter time-series fusion-based intelligent identification method for CEMS calibration status. The method includes: starting condition determination, identifying calibration status only for operational data, excluding shutdown status; data acquisition and preprocessing, real-time acquisition of oxygen content and various pollutant concentration data; construction of a multi-dimensional feature vector, organizing multiple parameters into a time-series feature vector according to time windows; establishment of an identification model, where the underlying rule engine uses a threshold comparison method to distinguish calibration coefficients of each factor, combining multiple calibration rules to output the site calibration status through multi-dimensional combination determination; and output and application of operating status, using the identification results to automatically remove or mark data from calibration periods during data validity audits and to automatically remove abnormal conversion peaks caused by non-operating conditions during exceedance determinations. This application, through multi-parameter time-series fusion and multi-dimensional combination determination, can achieve accurate and automatic identification of calibration status, improving data quality and regulatory efficiency.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, specifically relating to a multi-parameter time-series fusion-based intelligent identification method for CEMS calibration status. Background Technology

[0002] With increasingly stringent environmental protection requirements, continuous emission monitoring technology for stationary pollution sources has become an important component of the environmental regulatory system. As an online monitoring device installed at the emission outlet of stationary pollution sources, the continuous emission monitoring system undertakes the important responsibility of real-time monitoring of the concentration and total emissions of pollutants such as sulfur dioxide, nitrogen oxides, and particulate matter in the flue gas. The monitoring data is transmitted to the environmental protection authorities in real time, providing important evidence for environmental management, pollution discharge fees, and law enforcement supervision.

[0003] Among them, CEMS (Continuous Emission Monitoring System) exhibits special data characteristics during the calibration process. The monitoring data generated during the calibration period is invalid data and should be discarded according to the specifications. It should not be used as the basis for judging whether emissions exceed the standards. The "Rules for Marking Automatic Monitoring Equipment for Pollutant Emissions" issued by the Ministry of Ecology and Environment on July 19, 2022, clearly requires that the calibration period be marked as "automatic monitoring equipment maintenance" to ensure the validity of the data.

[0004] Existing technologies mainly rely on two methods for calibration status identification: automatic labeling and manual labeling. Automatic labeling is automatically generated by the CEMS system based on standardized calibration procedures such as daily zero-point calibration and range calibration. Manual labeling, on the other hand, requires maintenance personnel to manually select time periods and statuses on the enterprise side of the monitoring platform. However, in actual operation, automatic labeling is often affected by factors such as aging field equipment, imperfect instrument transmission protocols, and non-standard human operation, resulting in mislabeling and omissions. Manual labeling also suffers from problems such as omissions, mislabeling, and delayed labeling. In some cases, false labeling may even occur to avoid exceeding the standard data, leading to invalid data being misjudged as valid emission data or valid data being incorrectly removed. This directly interferes with the core business of CEMS, such as data validity determination, exceeding the standard determination, and total pollutant accounting.

[0005] In view of the shortcomings of existing technologies, such as non-standard automatic marking and unreliable manual marking, this invention proposes a CEMS calibration status intelligent identification method based on multi-parameter time-series fusion, which is of particular importance for achieving accurate identification of calibration status. Summary of the Invention

[0006] The purpose of this invention is to provide a method for intelligent identification of CEMS calibration status based on multi-parameter time-series fusion, which can effectively solve the problems in the background art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for intelligent identification of CEMS calibration status based on multi-parameter time-series fusion includes the following steps: Step 1, Startup Condition Determination: Based on the station's operating condition information, calibration status identification is performed only for monitoring data where the flue gas emission condition is in operation, and calibration status identification is not performed for monitoring data where the flue gas emission condition is in shutdown. Step 2, Data Acquisition and Preprocessing: Real-time acquisition of operating parameters and pollutant concentration data of stationary pollution source emission outlets using IoT monitoring technology, including oxygen content and pollutant concentration; Step 3, construct a multi-dimensional feature vector: organize the preprocessed multi-parameters according to the time window to construct a time-series feature vector, which includes the oxygen content and the concentration of various pollutants at each time point; Step 4, establish the identification model: The underlying rule engine uses a threshold comparison method to distinguish the calibration coefficients of each factor for oxygen content and the concentration and amplitude of each pollutant. Based on the identification results of the calibration coefficients of each factor, and combined with various calibration rules such as air calibration, nitrogen calibration, standard gas calibration and their combinations, the station calibration status is output through multi-dimensional combination judgment. Step 5, Operating Status Output and Application: The identified calibration status is output as a marker information and used in at least one scenario, such as automatically removing or marking calibration period data in the validity review of flue gas emission continuous monitoring system data, automatically removing abnormal conversion peaks caused by non-normal operating conditions in the judgment of exceeding standards, and not including non-normal operating data in the calculation of total pollutant amount.

[0008] Furthermore, the threshold comparison method of the underlying rule engine in step 4 includes: setting a first threshold for oxygen content, a second threshold for oxygen content, a pollutant concentration threshold, a decrease threshold, and an increase threshold; determining whether the oxygen content is lower than the first threshold for oxygen content or between the first threshold for oxygen content and the second threshold for oxygen content, and whether the concentration of each pollutant is lower than the pollutant concentration threshold or undergoes a sharp increase or decrease.

[0009] Furthermore, the calibration rules in step 4 include the following types: The air calibration pattern is characterized by oxygen content approaching the preset air oxygen content value, while other factors are close to zero. The nitrogen calibration pattern shows that all factors except dust are close to zero. The single standard gas calibration rule is characterized by calibrating only one factor at a time, and the data of the factor and other factors are mutually exclusive. The calibration pattern of mixed standard gas is characterized by simultaneous calibration of multiple factors, with the factors changing synchronously and the oxygen content approaching zero. And the continuity between different factor calibrations, that is, when the calibration of one factor is completed, the calibration of the next factor is performed.

[0010] Furthermore, the specific calibration rules in step 4, based on the identification results of each factor calibration coefficient, include the following combinations: After air calibration, connect to nitrogen calibration, then connect to standard gas calibration. After air calibration, connect to standard gas calibration; after nitrogen calibration, connect to nitrogen calibration. Nitrogen calibration is followed by air calibration, and then standard gas calibration. Nitrogen calibration is followed by standard gas calibration, and then air calibration. After calibration with standard gas, connect to nitrogen calibration, then connect to air calibration. After standard gas calibration, connect to air calibration, then connect to nitrogen calibration. Nitrogen calibration is performed after air calibration. After air calibration, connect to standard gas for calibration. Nitrogen calibration is followed by air calibration. After nitrogen calibration, connect to standard gas for further calibration. After calibration with standard gas, nitrogen calibration is performed. After calibration with standard gas, perform air calibration. Individual air calibration; Individual nitrogen calibration; Separate standard gas calibration.

[0011] Furthermore, the algorithm model rules in step 4 include: Set initial parameters, including the rate of decrease, the rate of increase, the pollutant concentration limit, the first limit of oxygen content, and the second limit of oxygen content; When real-time data is received, the system first determines whether it belongs to the monitoring data of the flue gas emission condition of the boiler shutdown state. If so, the process ends. If not, determine whether it is the first record. If it is the first record, determine whether it is calibrated or not based on whether the oxygen content is less than the first limit of oxygen content and whether the concentration of all pollutants is less than the limit of pollutant concentration. If it is not the first record, then determine whether the previous record is in calibration state, and make a judgment based on the relationship between the current oxygen content and the second limit of oxygen content and the sharp drop or rise of pollutant concentration.

[0012] Preferably, when the previous record is in calibration state, if the current oxygen content is less than or equal to the second limit of oxygen content, or if the current oxygen content is greater than the second limit of oxygen content but the pollutant concentration drops sharply to less than the decrease rate multiple and all pollutant concentrations are less than the pollutant concentration limit, then it is determined to be in calibration state; otherwise, it is determined to be in non-calibration state.

[0013] Preferably, when the previous record is in a non-calibration state, if the current oxygen content is less than or equal to the second oxygen content limit and meets one of the following conditions, it is determined to be in a calibration state: Condition 1: The pollutant concentration rises sharply to more than twice the stated increase, or the pollutant concentration is less than the stated pollutant concentration limit; Condition 2: The pollutant concentration drops sharply to less than a multiple of the stated decrease, and the pollutant concentration without any sharp rises or falls is less than the stated pollutant concentration limit; Condition 3: The pollutant concentration either rises sharply to more than a multiple of the stated increase or falls sharply to less than a multiple of the stated decrease, and the pollutant concentration without any sharp rise or fall is less than the stated pollutant concentration limit.

[0014] Furthermore, the calibration status output in step 5 includes: transmitting the identified calibration status marker information to the environmental data acquisition instrument or environmental big data platform through the data interface, and marking the data during the calibration period as invalid data, which will be automatically removed during data validity review.

[0015] A calibration status intelligent identification system for a continuous emission monitoring system for flue gas based on multi-parameter time-series fusion, comprising: The data acquisition module is used to collect operating parameters and pollutant concentration data of stationary pollution source emission outlets in real time through Internet of Things monitoring technology. The data includes oxygen content and concentrations of various pollutants. The data processing module is used to preprocess the collected data; The feature construction module is used to organize the preprocessed multi-parameters according to the time window to construct a time-series feature vector; The intelligent identification module includes a low-level rule engine and a multi-dimensional combination judgment unit. The low-level rule engine is used to distinguish the calibration coefficients of each factor by using a threshold comparison method for oxygen content and the concentration and amplitude of each pollutant. The multi-dimensional combination judgment unit is used to perform multi-dimensional combination judgment based on the identification results of the calibration coefficients of each factor, combined with various calibration rules such as air calibration, nitrogen calibration, standard gas calibration and their combinations, and outputs the station calibration status. The application output module is used to output the identified calibration status as tag information and is used in at least one scenario of continuous emission monitoring system data validity review, exceedance judgment and total pollutant calculation.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. Accurate identification of calibration status: By deeply mining the time-series patterns in the historical data of CEMS multi-parameter data through multi-parameter time-series fusion technology, and based on the data correlation characteristics between oxygen content and the concentration of various pollutants, the calibration status can be automatically and accurately identified, effectively solving the technical problems of automatic marking of mis-calibration, omission of calibration and unreliable manual marking in the existing technology.

[0017] 2. Improve data quality: Based on the data correlation patterns between multiple factors, the system automatically identifies and marks the CEMS calibration status, which can effectively avoid problems such as mislabeling, omissions, abnormal marking, false marking, and failure to mark in a timely manner. This significantly improves the accuracy of data validity review, exceedance judgment, and total pollutant accounting, ensuring the authenticity and validity of environmental protection data.

[0018] 3. Reduce operating costs: Reduce the need for manual intervention, simultaneously reduce the operating and management costs of polluting enterprises and environmental regulatory departments, improve the efficiency and accuracy of regulatory enforcement, and promote the comprehensive upgrade of the continuous emission monitoring system for stationary pollution sources towards automation and intelligence. Attached Figure Description

[0019] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating the intelligent identification method for CEMS calibration status based on multi-parameter time-series fusion proposed in this invention. Figure 2 This is a schematic diagram of the composition of the intelligent identification system for CEMS calibration status based on multi-parameter time-series fusion proposed in this invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0022] Example 1

[0023] like Figure 1 As shown, Embodiment 1 of the present invention discloses an intelligent identification method for calibration status of a continuous flue gas emission monitoring system based on multi-parameter time-series fusion, comprising the following steps: Step S1: Startup condition determination: Based on the station's operating condition information, calibration status identification is performed only for monitoring data where the flue gas emission condition is in operation, and no calibration status identification is performed for monitoring data where the flue gas emission condition is in shutdown.

[0024] Specifically, step S1 is executed by the start-up determination module deployed in the environmental data acquisition instrument or environmental big data platform. The start-up determination module first obtains real-time operating information uploaded by the incineration plant from the operating condition data interface. This operating information includes parameters such as furnace temperature, production load, and power generation. The start-up determination module has internally set operating threshold parameters, for example, the operating threshold for furnace temperature is set to 800 degrees Celsius, and the operating threshold for production load is set to 30% of the rated load. When the furnace temperature reaches 800 degrees Celsius or above and the production load reaches 30% or above the rated load, the flue gas emission condition is determined to be in operation, and the calibration status identification process is initiated. When the furnace temperature is below 800 degrees Celsius or the production load is below 30% of the rated load, the flue gas emission condition is determined to be in shutdown status, the calibration status identification process is not executed, and a non-calibration status identifier is directly output. Step S1 effectively avoids invalid calibration identification calculations under abnormal emission conditions such as shutdown, improving identification efficiency.

[0025] Step S2: Data Acquisition and Preprocessing: Real-time acquisition of operating parameters and pollutant concentration data of stationary pollution source emission outlets using IoT monitoring technology, including oxygen content and pollutant concentration.

[0026] Specifically, step S2 is completed collaboratively by the data acquisition layer and the data processing layer. The data acquisition layer, deployed in the continuous emission monitoring system at the incinerator's emission outlet, consists of multiple sensor units, each responsible for collecting different monitoring parameters. The oxygen content sensor uses a paramagnetic oxygen analyzer with a measurement range of 0 to 25%, an accuracy of ±0.1%, and a response time of less than 30 seconds. It is installed at the rear end of the flue gas sampling probe to directly measure the percentage of oxygen content in the flue gas. The pollutant concentration sensors include: a sulfur dioxide concentration sensor using ultraviolet fluorescence with a measurement range of 0 to 500 mg / m³ and an accuracy of ±2%FS; a nitrogen oxide concentration sensor using chemiluminescence with a measurement range of 0 to 500 mg / m³ and an accuracy of ±2%FS; a particulate matter concentration sensor using light scattering with a measurement range of 0 to 200 mg / m³ and an accuracy of ±10%FS; and a hydrogen chloride concentration sensor using an ion-selective electrode with a measurement range of 0 to 100 mg / m³ and an accuracy of ±5%FS. Each of the aforementioned sensors outputs analog signals ranging from 4 to 20 mA, transmitting the collected real-time monitoring data to the data acquisition module. The data acquisition module uses an industrial-grade programmable logic controller (PLC) as its core processing unit and supports the Modbus RTU communication protocol. The module is equipped with 8 analog input channels and 4 digital input channels, used to receive analog and digital signals from the sensors, respectively. The sampling frequency of the data acquisition module is set to once per minute, meaning that data acquisition and analog-to-digital conversion of all channels are completed every 60 seconds. The module integrates a digital filtering algorithm to perform preliminary moving average processing on the raw data, with a filtering window of 3 sampling points to suppress high-frequency noise interference. The data acquisition module communicates with the host computer via an RS485 interface, using the Modbus RTU protocol to transmit the acquired data to the data processing layer. The communication baud rate is 9600 bits per second, with 8 data bits, 1 stop bit, and no parity bit. Upon receiving the raw data, the data processing layer performs timestamp calibration, moving average filtering, and outlier removal. The moving average filter uses a 5-minute window, calculating the arithmetic mean of five data points (the current time and the previous four times). Outliers are identified when their deviation from the historical mean exceeds three standard deviations. When an outlier is detected, it is replaced by linear interpolation of data from adjacent time points. The preprocessed data is stored in chronological order in an in-memory database, along with complete timestamps. Step S2 provides high-quality, aligned time-series data, offering reliable input for subsequent feature construction.

[0027] Step S3: Construct a multi-dimensional feature vector: Organize the preprocessed multi-parameters according to the time window to construct a time-series feature vector, which includes the oxygen content and the concentration of various pollutants at each time step.

[0028] Specifically, step S3 is executed by the feature construction layer. The feature construction layer uses an in-memory database as its core storage component, configured with 64 gigabytes of memory to cache all monitoring data from the most recent 30 minutes. The feature construction module extracts data from the in-memory database according to a time window length of 15 minutes, meaning that each time a time-series feature vector is constructed, data from 15 time points, including the current time and the previous 14 time points, are extracted. Each time point contains multiple parameter dimensions, such as oxygen content, sulfur dioxide concentration, nitrogen oxide concentration, particulate matter concentration, and hydrogen chloride concentration—a total of 5 dimensions. Therefore, the total dimension of the time-series feature vector is 15 multiplied by 5, equal to 75 dimensions. The feature construction module organizes the data into a multi-dimensional time-series matrix in chronological order. The row index of the multi-dimensional time-series matrix represents the time point (15 rows, corresponding to 15 time points), and the column index represents the parameter dimension (5 columns, corresponding to oxygen content and the concentration of each pollutant, respectively). Each element in the multi-dimensional time-series matrix is ​​the filtered value of the corresponding parameter at the corresponding time point. The feature construction module expands the multi-dimensional time series matrix into a one-dimensional vector, arranging them in a fixed order according to time sequence and parameter dimensions. The constructed time series feature vector is then passed to the intelligent recognition layer via shared memory for the recognition model to perform calibration state judgment. Through step S3, the original multi-parameter time series can be transformed into a structured feature vector, facilitating subsequent processing by the rule engine.

[0029] Step S4: Establish an identification model: The underlying rule engine uses a threshold comparison method to distinguish the calibration coefficients of each factor for oxygen content and the concentration and amplitude of each pollutant. Based on the identification results of the calibration coefficients of each factor, and combined with various calibration rules such as air calibration, nitrogen calibration, standard gas calibration and their combinations, the station calibration status is output through multi-dimensional combination judgment.

[0030] Specifically, step S4 is completed collaboratively by the underlying rule engine of the intelligent recognition layer and the multi-dimensional combination judgment unit.

[0031] Step S41: The underlying rule engine uses a threshold comparison method to distinguish the calibration coefficients of each factor for oxygen content and the concentration and amplitude of each pollutant.

[0032] Specifically, step S41 includes: setting a first threshold for oxygen content, a second threshold for oxygen content, a pollutant concentration threshold, a decrease threshold, and an increase threshold; determining whether the oxygen content is lower than the first threshold or between the first and second thresholds, and whether the concentrations of each pollutant are lower than the pollutant concentration threshold or have experienced a sharp increase or decrease. In a specific implementation, the threshold parameters are set as follows: the first threshold for oxygen content is 5%, the second threshold is 10%, the pollutant concentration threshold is 1 mg / m³, the decrease threshold is 0.5 times, and the increase threshold is 2 times. For oxygen content data, the underlying rule engine determines whether the current oxygen content is lower than the first threshold, between the first and second thresholds, or higher than the second threshold, and outputs the calibration coefficients for oxygen content as high-confidence calibration, medium-confidence calibration, and no calibration, respectively. For each pollutant concentration, the underlying rule engine first determines whether the concentration is lower than the pollutant concentration threshold, then calculates the change in concentration between adjacent time points, and determines whether a sharp increase or decrease has occurred. The criteria for a sharp increase are: the increase in concentration from the previous moment exceeds a threshold value, i.e., the current concentration is greater than twice the concentration from the previous moment; the criteria for a sharp decrease are: the decrease in concentration from the previous moment exceeds a threshold value, i.e., the current concentration is less than 0.5 times the concentration from the previous moment. The underlying rule engine outputs calibration coefficients for each pollutant based on these rules. Step S41 quantifies the probability of each factor being in a calibration state, providing a basis for subsequent multi-dimensional combination judgments.

[0033] Step S42: Based on the identification results of the calibration coefficients of each factor, and combined with the multiple calibration rules of air calibration, nitrogen calibration, standard gas calibration and their combinations, the station calibration status is output through multi-dimensional combination judgment.

[0034] Specifically, the calibration rules in step S42 include the following types: air calibration rule, characterized by an oxygen content close to 21% and other factors close to zero; nitrogen calibration rule, characterized by all factors except dust close to zero; single standard gas calibration rule, characterized by calibrating only one factor at a time, with the data of the factor and other factors being mutually exclusive; mixed standard gas calibration rule, characterized by calibrating multiple factors simultaneously, with the factor rules changing synchronously and the oxygen content close to zero; and the continuity rule between different factor calibrations, that is, when one factor is calibrated, the next factor is calibrated.

[0035] In a preferred embodiment, the specific calibration rules based on the identification results of each factor calibration coefficient in step S42 include the following combinations: air calibration followed by nitrogen calibration followed by standard gas calibration; air calibration followed by standard gas calibration followed by nitrogen calibration; nitrogen calibration followed by air calibration followed by standard gas calibration; nitrogen calibration followed by standard gas calibration followed by air calibration; standard gas calibration followed by nitrogen calibration followed by air calibration; standard gas calibration followed by air calibration followed by nitrogen calibration; air calibration followed by nitrogen calibration; air calibration followed by standard gas calibration; nitrogen calibration followed by air calibration; nitrogen calibration followed by standard gas calibration; standard gas calibration followed by nitrogen calibration; standard gas calibration followed by air calibration; air calibration alone; nitrogen calibration alone; standard gas calibration alone. The multi-dimensional combination determination unit analyzes the change rules of each factor calibration coefficient according to the time series. When the identification result meets any calibration state transition mode, the output calibration state identifier is "calibrated"; otherwise, the output calibration state identifier is "non-calibrated".

[0036] Furthermore, the algorithm model rules in step S4 specifically include: setting initial parameters, which include the decrease rate, the increase rate, the pollutant concentration limit, the first oxygen content limit, and the second oxygen content limit. The decrease rate is denoted as d. percent The physical meaning is the proportional threshold for the decrease in pollutant concentration, which is dimensionless; the magnitude of the increase is denoted as u. percent The physical meaning is the proportional threshold for the increase in pollutant concentration, which is dimensionless; the pollutant concentration limit is denoted as nd. level The physical meaning is the upper limit of the concentration for determining whether a pollutant is close to zero, expressed in milligrams per cubic meter; the first limit of oxygen content is denoted as o. level,1 The physical meaning is the upper limit of oxygen content for determining the air calibration state, expressed as a percentage; the second limit of oxygen content is denoted as o. level,2The physical meaning of "oxygen content" is the upper limit of the oxygen content for determining the calibration status of the standard gas, expressed as a percentage. When real-time data is received, the system first determines whether it belongs to a monitoring data point where the flue gas emission condition is a shutdown state. If so, the process ends. If not, it determines whether it is the first record. If it is the first record, calibration or non-calibration is determined based on whether the oxygen content is less than the first oxygen content limit and whether all pollutant concentrations are less than the pollutant concentration limits. If it is not the first record, it determines whether the previous record was in a calibration state, and makes a determination based on the relationship between the current oxygen content and the second oxygen content limit, as well as the sharp drop or rise in pollutant concentrations. Specifically, when the previous record was in a calibration state, if the current oxygen content is less than or equal to the second oxygen content limit, or if the current oxygen content is greater than the second oxygen content limit but the pollutant concentration drops sharply to less than a multiple of the drop rate and all pollutant concentrations are less than the pollutant concentration limits, then it is determined to be in a calibration state; otherwise, it is determined to be in a non-calibration state. When the previous record was in a non-calibration state, if the current oxygen content is less than or equal to the second oxygen content limit and meets one of the following conditions, it is determined to be in a calibration state: Condition 1: The pollutant concentration rises sharply to more than a multiple of the increase, or the pollutant concentration is less than the pollutant concentration limit; Condition 2: The pollutant concentration drops sharply to less than a multiple of the decrease, and there are no pollutant concentrations with sharp increases or decreases that are less than the pollutant concentration limit; Condition 3: The pollutant concentration both rises sharply to more than a multiple of the increase and drops sharply to less than a multiple of the decrease, and there are no pollutant concentrations with sharp increases or decreases that are less than the pollutant concentration limit.

[0037] In a preferred embodiment, the decrease threshold d percent The value is 0.2, and the rise amplitude threshold u percent The value is 5, and the pollutant concentration limit is nd. level The value is 5 milligrams per cubic meter, and the first limit for oxygen content is... level,1 The value is 20.5%, which is the second limit for oxygen content. level,2 The value is set at 2%. This set of parameters is based on statistical analysis of a large amount of historical data, and can adapt to the emission characteristics of typical waste incineration plants. In actual operation, it has achieved an identification accuracy rate of 96.8%.

[0038] Step S4 enables the automatic and objective identification of calibration status based on the time-series variation patterns of multiple parameters, avoiding miscalibration and omissions caused by manual intervention.

[0039] As an innovation of this embodiment, a preferred operation for evaluating the confidence level of the calibration status can be added in step S4. Specifically, when the calibration status identifier output by the multi-dimensional combination judgment unit is "calibration", a confidence score is calculated simultaneously. The confidence score is calculated based on the degree of matching of the calibration rule type and the consistency of the calibration coefficients of each factor. For example, when the identification result completely matches a certain calibration rule and all factor calibration coefficients are in the high confidence range, the confidence score is set to 100%; when there is a partial match, the confidence score decreases linearly according to the degree of matching. The confidence score is output along with the calibration status identifier for reference by subsequent business systems. When the confidence score is lower than a preset threshold (e.g., 60%), the system automatically marks this period as "awaiting manual review", thereby further improving the reliability of data validity verification.

[0040] Step S5: Output and application of operating status.

[0041] Step S5: Output the identified calibration status as tag information and use it in at least one scenario: automatic removal or tagging of calibration period data in the validity review of flue gas emission continuous monitoring system data, automatic removal of abnormal conversion peaks caused by non-normal operating conditions in the determination of exceeding standards, and non-normal operating data not being included in the calculation of total pollutant calculation.

[0042] Specifically, step S5 is executed by the data interface service program of the application output layer. After the intelligent identification layer outputs the calibration status identifier, the data interface service program immediately encapsulates the calibration status flag information into a standardized data message. The flag information includes the calibration status identifier (value "calibrated" or "non-calibrated"), calibration start time (the moment it is first identified as being in calibration status), calibration end time (the moment it is last identified as being in calibration status), and calibration factor type (the name of the factor determined by the underlying rule engine to be calibrated). The data message is pushed to the data management module of the environmental big data platform via a RESTful application programming interface, and simultaneously to the local storage unit of the environmental data acquisition instrument via the MQTT protocol. Upon receiving the flag information, the data management module automatically marks the data for the corresponding time period as invalid data. During data validity review, the system automatically removes these marked invalid data and does not include them in the valid data statistics; during exceedance determination, the system automatically removes abnormal calculation peaks caused by abnormal operating conditions to avoid misjudgment; during pollutant total quantity calculation, the system automatically excludes data from abnormal operating periods and does not include them in the total emission calculation. Step S5 enables deep integration of calibration and identification results with core environmental protection business scenarios, effectively improving data quality and regulatory efficiency.

[0043] In another preferred embodiment, the data acquisition and preprocessing in step S2 can also be achieved as follows: the edge computing gateway connects to the field sensor devices via an RS485 interface, with the acquisition frequency set to once per minute. After preliminary filtering, the acquired raw data is uploaded to the cloud computing layer via a fourth-generation or fifth-generation mobile communication wireless network. The edge computing gateway locally stores a simplified recognition model, enabling it to independently perform basic calibration status recognition in the event of a network interruption. The recognition results are temporarily stored locally and synchronized to the cloud computing layer after network recovery. This solution is particularly suitable for multi-site centralized management scenarios, improving the system's fault tolerance and real-time performance.

[0044] Through steps S1 to S5 described above, this embodiment achieves intelligent identification of the calibration status of the continuous emission monitoring system for flue gas. In actual operation, the accuracy rate of calibration status identification reaches 96.8%, and the false alarm rate is controlled within 2.5%, effectively solving the problems of mislabeling and omissions in existing automatic and manual labeling, and reducing the operating and management costs for polluting enterprises and environmental regulatory departments.

[0045] Example 2

[0046] like Figure 2 As shown, Embodiment 2 of the present invention discloses an intelligent identification system for calibration status of a continuous flue gas emission monitoring system based on multi-parameter time-series fusion, comprising: Data acquisition module M10: Used to collect operating parameters and pollutant concentration data of stationary pollution source emission outlets in real time through Internet of Things monitoring technology. The data includes oxygen content and concentrations of various pollutants.

[0047] Specifically, the data acquisition module M10 is deployed at the incineration plant's emission outlet and consists of multiple sensor units and a data acquisition unit. The sensor units include: an oxygen content sensor, employing a paramagnetic oxygen analyzer with a measurement range of 0 to 25%, a measurement accuracy of ±0.1%, and a response time of less than 30 seconds, installed at the rear end of the flue gas sampling probe; a sulfur dioxide concentration sensor, employing ultraviolet fluorescence, with a measurement range of 0 to 500 mg / m³ and a measurement accuracy of ±2%FS; a nitrogen oxide concentration sensor, employing chemiluminescence, with a measurement range of 0 to 500 mg / m³ and a measurement accuracy of ±2%FS; a particulate matter concentration sensor, employing light scattering, with a measurement range of 0 to 200 mg / m³ and a measurement accuracy of ±10%FS; and a hydrogen chloride concentration sensor, employing ion-selective electrode, with a measurement range of 0 to 100 mg / m³ and a measurement accuracy of ±5%FS. Each sensor outputs a 4 to 20 mA analog signal to the data acquisition unit. The data acquisition unit uses an industrial-grade programmable logic controller (PLC) as its core processing unit, supports the Modbus RTU communication protocol, and is equipped with 8 analog input channels and 4 digital input channels. The data acquisition unit communicates with the host computer via an RS485 interface at a baud rate of 9600 bits per second. The data acquisition module M10 and the data processing module M20 are connected via a wired Ethernet connection at a transmission rate of 100 megabits per second.

[0048] Data processing module M20: Used to preprocess the collected data.

[0049] Specifically, the data processing module M20 is deployed on the edge computing node of the environmental protection big data platform, using a high-performance industrial server as its hardware carrier. The data processing module M20 runs a data preprocessing program, receiving the raw data stream from the data acquisition module M10 and performing the following preprocessing operations: timestamp calibration, aligning the data from each sensor according to the actual acquisition time; moving average filtering, setting the filter window size to 5 minutes, calculating the arithmetic mean of 5 data points (the current time and the previous 4 times); outlier removal, where outliers are judged by deviations from the historical mean exceeding 3 standard deviations, and linear interpolation of adjacent time points is used to replace outliers when they are detected; and data format standardization, converting the processed data into a floating-point format. The preprocessed data is transmitted to the feature construction module M30 via an internal message queue. The data processing module M20 and the feature construction module M30 exchange data via shared memory.

[0050] Feature construction module M30: Used to organize the preprocessed multi-parameters according to the time window to construct the time series feature vector.

[0051] Specifically, the feature construction module M30 uses an in-memory database as its core storage component, configured with 64 gigabytes of memory to cache all monitoring data from the most recent 30 minutes. The feature construction module M30 extracts data from the in-memory database within a 15-minute time window, retrieving data from 15 time points each time, including the current moment and the previous 14 moments. Each time point contains five parameter dimensions (oxygen content, sulfur dioxide concentration, nitrogen oxide concentration, particulate matter concentration, and hydrogen chloride concentration), resulting in a total dimensionality of 75 dimensions for the time-series feature vector. The feature construction module M30 expands the multi-dimensional time-series matrix into a one-dimensional vector and transmits it to the intelligent recognition module M40 via shared memory. The feature construction module M30 and the intelligent recognition module M40 communicate asynchronously via an internal message queue.

[0052] The intelligent recognition module M40 includes an underlying rule engine and a multi-dimensional combination judgment unit.

[0053] Specifically, the intelligent identification module M40 is deployed on the computing nodes of the environmental protection big data platform and adopts a multi-threaded parallel processing architecture. The intelligent identification module M40 includes: The underlying rule engine M41 is used to differentiate the calibration coefficients of each factor by using threshold comparison methods for oxygen content and the concentrations and amplitudes of various pollutants. Internally, the underlying rule engine M41 maintains a rule knowledge base storing threshold judgment rules for oxygen content and various pollutants. Threshold parameters include: a first threshold of 5% for oxygen content, a second threshold of 10% for oxygen content, a pollutant concentration threshold of 1 mg / m³, a decrease threshold of 0.5 times, and an increase threshold of 2 times. After receiving the time-series feature vector, the underlying rule engine M41 calculates the threshold judgment results for oxygen content and the steep increase / decrease judgment results for each pollutant concentration, and outputs the calibration coefficients for each factor.

[0054] The multi-dimensional combination judgment unit M42 is used to perform multi-dimensional combination judgment based on the identification results of various factor calibration coefficients, combined with multiple calibration rules such as air calibration, nitrogen calibration, standard gas calibration, and their combinations, and outputs the site calibration status. The multi-dimensional combination judgment unit M42 internally stores matching rules for 13 calibration status transition modes. It analyzes the changing patterns of each factor calibration coefficient according to time series analysis. When the identification result matches any calibration status transition mode, it outputs a calibration status identifier; otherwise, it outputs a non-calibration status identifier.

[0055] The intelligent recognition module M40 is set to execute once every 60 seconds, that is, to update the calibration status once per minute. The intelligent recognition module M40 is connected to the feature construction module M30 via shared memory, and the intelligent recognition module M40 is connected to the application output module M50 via a data interface.

[0056] Application output module M50: Used to output the identified calibration status as tag information and to be used in at least one of the following scenarios: data validity verification, exceedance determination and total pollutant calculation in the continuous emission monitoring system.

[0057] Specifically, the application output module M50 is equipped with a data interface service program that supports two data interface protocols: a descriptive state transfer application programming interface and a message queue telemetry transmission protocol. The application output module M50 receives calibration status marker information from the intelligent identification module M40, encapsulates this information into standardized data packets, and transmits them through the data interface to the environmental data acquisition instrument and the data management module of the environmental big data platform. Upon receiving the marker information, the data management module automatically marks the data for the corresponding time period as invalid data, and automatically removes this invalid data in subsequent data validity verification, exceedance determination, and pollutant total quantity calculation. The application output module M50 connects to the external environmental big data platform via a wired Ethernet connection and operates in full-duplex mode.

[0058] Through the collaborative operation of modules M10 to M50, this embodiment achieves intelligent identification of the calibration status of the continuous emission monitoring system for flue gas. In actual operation, the system achieves a calibration status identification accuracy rate of 96.8%, with a false alarm rate controlled within 2.5%. This effectively avoids problems such as mislabeling, missed labeling, abnormal labeling, false labeling, and untimely labeling, improving the accuracy of data validity verification, exceedance determination, and total pollutant calculation, ensuring the authenticity and validity of environmental data. The implementation of this invention can simultaneously reduce the operating and management costs of polluting enterprises and environmental regulatory departments, improve the efficiency and accuracy of regulatory enforcement, and promote the comprehensive upgrade of continuous emission monitoring systems for stationary pollution sources towards automation and intelligence.

[0059] The foregoing has shown and described the basic principles, main features, and advantages of this invention. Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this invention. Various changes and modifications can be made to this invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent identification of CEMS calibration status based on multi-parameter time-series fusion, characterized in that, Includes the following steps: Step 1, Startup Condition Determination: Based on the station's operating condition information, calibration status identification is performed only for monitoring data where the flue gas emission condition is in operation, and calibration status identification is not performed for monitoring data where the flue gas emission condition is in shutdown. Step 2, Data Acquisition and Preprocessing: Real-time acquisition of operating parameters and pollutant concentration data of stationary pollution source emission outlets using IoT monitoring technology, including oxygen content and pollutant concentration; Step 3, construct a multi-dimensional feature vector: organize the preprocessed multi-parameters according to the time window to construct a time-series feature vector, which includes the oxygen content and the concentration of various pollutants at each time point; Step 4, establish the identification model: The underlying rule engine uses a threshold comparison method to distinguish the calibration coefficients of each factor for oxygen content and the concentration and amplitude of each pollutant. Based on the identification results of the calibration coefficients of each factor, and combined with various calibration rules such as air calibration, nitrogen calibration, standard gas calibration and their combinations, the station calibration status is output through multi-dimensional combination judgment. Step 5, Operating Status Output and Application: The identified calibration status is output as a marker information and used in at least one scenario, such as automatically removing or marking calibration period data in the validity review of flue gas emission continuous monitoring system data, automatically removing abnormal conversion peaks caused by non-normal operating conditions in the judgment of exceeding standards, and not including non-normal operating data in the calculation of total pollutant amount.

2. The method according to claim 1, characterized in that, The threshold comparison method of the underlying rule engine in step 4 includes: setting a first threshold for oxygen content, a second threshold for oxygen content, a pollutant concentration threshold, a decrease threshold, and an increase threshold; determining whether the oxygen content is lower than the first threshold for oxygen content or between the first threshold for oxygen content and the second threshold for oxygen content, and whether the concentration of each pollutant is lower than the pollutant concentration threshold or undergoes a sharp increase or decrease.

3. The method according to claim 1, characterized in that, The calibration rules in step 4 include the following types: The air calibration pattern is characterized by oxygen content approaching the preset air oxygen content value, while other factors are close to zero. The nitrogen calibration pattern shows that all factors except dust are close to zero. The single standard gas calibration rule is characterized by calibrating only one factor at a time, and the data of the factor and other factors are mutually exclusive. The calibration pattern of mixed standard gas is characterized by simultaneous calibration of multiple factors, with the factors changing synchronously and the oxygen content approaching zero. And the continuity between different factor calibrations, that is, when the calibration of one factor is completed, the calibration of the next factor is performed.

4. The method according to claim 1, characterized in that, The specific calibration rules in step 4, based on the identification results of each factor calibration coefficient, include the following combinations: After air calibration, connect to nitrogen calibration, then connect to standard gas calibration. After air calibration, connect to standard gas calibration; after nitrogen calibration, connect to nitrogen calibration. Nitrogen calibration is followed by air calibration, and then standard gas calibration. Nitrogen calibration is followed by standard gas calibration, and then air calibration. After calibration with standard gas, connect to nitrogen calibration, then connect to air calibration. After standard gas calibration, connect to air calibration, then connect to nitrogen calibration. Nitrogen calibration is performed after air calibration. After air calibration, connect to standard gas for calibration. Nitrogen calibration is followed by air calibration. After nitrogen calibration, connect to standard gas for further calibration. After calibration with standard gas, nitrogen calibration is performed. After calibration with standard gas, perform air calibration. Individual air calibration; Individual nitrogen calibration; Separate standard gas calibration.

5. The method according to claim 1, characterized in that, The algorithm model rules in step 4 include: Set initial parameters, including the rate of decrease, the rate of increase, the pollutant concentration limit, the first limit of oxygen content, and the second limit of oxygen content; When real-time data is received, the system first determines whether it belongs to the monitoring data of the flue gas emission condition of the boiler shutdown state. If so, the process ends. If not, determine whether it is the first record. If it is the first record, determine whether it is calibrated or not based on whether the oxygen content is less than the first limit of oxygen content and whether the concentration of all pollutants is less than the limit of pollutant concentration. If it is not the first record, then determine whether the previous record is in calibration state, and make a judgment based on the relationship between the current oxygen content and the second limit of oxygen content and the sharp drop or rise of pollutant concentration.

6. The method according to claim 5, characterized in that, When the previous record is in calibration state, if the current oxygen content is less than or equal to the second limit of oxygen content, or if the current oxygen content is greater than the second limit of oxygen content but the pollutant concentration drops sharply to less than the decrease rate multiple and all pollutant concentrations are less than the pollutant concentration limit, then it is determined to be in calibration state; otherwise, it is determined to be in non-calibration state.

7. The method according to claim 5, characterized in that, When the previous record was in a non-calibration state, if the current oxygen content is less than or equal to the second oxygen content limit and meets one of the following conditions, it is determined to be in a calibration state: Condition 1: The pollutant concentration rises sharply to more than twice the stated increase, or the pollutant concentration is less than the stated pollutant concentration limit; Condition 2: The pollutant concentration drops sharply to less than a multiple of the stated decrease, and the pollutant concentration without any sharp rises or falls is less than the stated pollutant concentration limit; Condition 3: The pollutant concentration either rises sharply to more than a multiple of the stated increase or falls sharply to less than a multiple of the stated decrease, and the pollutant concentration without any sharp rise or fall is less than the stated pollutant concentration limit.

8. The method according to claim 1, characterized in that, The calibration status output in step 5 includes: transmitting the identified calibration status marker information to the environmental data acquisition instrument or environmental big data platform through the data interface, and marking the data during the calibration period as invalid data, which will be automatically removed during data validity review.

9. A calibration status intelligent identification system for a continuous flue gas emission monitoring system based on multi-parameter time-series fusion, characterized in that, include: The data acquisition module is used to collect operating parameters and pollutant concentration data of stationary pollution source emission outlets in real time through Internet of Things monitoring technology. The data includes oxygen content and concentrations of various pollutants. The data processing module is used to preprocess the collected data; The feature construction module is used to organize the preprocessed multi-parameters according to the time window to construct a time-series feature vector; The intelligent identification module includes a low-level rule engine and a multi-dimensional combination judgment unit. The low-level rule engine is used to distinguish the calibration coefficients of each factor by using a threshold comparison method for oxygen content and the concentration and amplitude of each pollutant. The multi-dimensional combination judgment unit is used to perform multi-dimensional combination judgment based on the identification results of the calibration coefficients of each factor, combined with various calibration rules such as air calibration, nitrogen calibration, standard gas calibration and their combinations, and outputs the station calibration status. The application output module is used to output the identified calibration status as tag information and is used in at least one scenario of continuous emission monitoring system data validity review, exceedance judgment and total pollutant calculation.

Images