A flue gas emission monitoring data verification method and device, electronic equipment and storage medium

Abstract

This invention discloses a method, apparatus, electronic device, and storage medium for verifying flue gas emission monitoring data, belonging to the field of flue gas emission monitoring data verification. The method includes: acquiring instantaneous flue gas concentration measurements, cumulative power generation, and cumulative coal consumption of the target emission source; predicting theoretical values ​​of flue gas pollutant concentrations based on cumulative power generation and cumulative coal consumption, and correcting the prediction error threshold based on the cumulative power generation; performing deviation analysis on the instantaneous flue gas concentration measurements and theoretical values, determining an anomaly when the deviation exceeds the correction threshold, and distinguishing between equipment anomalies and human interference based on the temporal distribution characteristics of the deviation, triggering corresponding alarm signals. Therefore, by implementing this invention, the problems of high false alarm rates due to the inability of static thresholds to adapt to fluctuations in unit operating conditions, and the inability to effectively distinguish whether data anomalies are caused by random equipment failures or human interference / cheating, can be solved in the existing technology.

Description

Technical Field

[0001] This invention relates to the field of flue gas emission monitoring data verification technology, specifically to a method, apparatus, electronic device, and storage medium for flue gas emission monitoring data verification. Background Technology

[0002] With the increasing national efforts to control air pollution, Continuous Emission Monitoring Systems (CEMS) have become a core technological means for monitoring the emissions of power plants and other polluting entities. The authenticity, accuracy, and completeness of CEMS monitoring data directly affect the calculation of pollution discharge fees and the impartiality of environmental administrative law enforcement. However, existing flue gas emission data verification technologies still have certain limitations. On the one hand, traditional verification methods often rely on static threshold settings or simple linear correlation analysis, ignoring the nonlinear dynamic changes in the pollutant formation mechanism of generator units under different load conditions and coal consumption levels. This leads to a decrease in the prediction accuracy of verification models when the unit operates under varying conditions or in complex environments, and the fixed allowable deviation is difficult to adapt to actual fluctuations, easily resulting in false alarms or missed alarms. On the other hand, existing technologies can usually only detect whether there are numerical anomalies in the data, lacking the ability to analyze the deep temporal distribution characteristics of abnormal data. This makes it difficult for the system to effectively distinguish whether the anomaly is caused by random failures such as equipment aging and sensor drift, or by targeted behaviors that violate physical laws, such as human interference and data tampering. This greatly limits the efficiency of regulatory authorities in identifying and obtaining evidence of human fraud. Summary of the Invention

[0003] This invention provides a method, apparatus, electronic device, and storage medium for verifying flue gas emission monitoring data. These solutions address the problems in the prior art where static thresholds are difficult to adapt to fluctuations in unit operating conditions, leading to high false alarm rates and the inability to effectively distinguish whether data anomalies are caused by random equipment malfunctions or human interference / cheating.

[0004] An embodiment of the present invention provides a method for verifying flue gas emission monitoring data, comprising: Acquire instantaneous flue gas concentration measurements, cumulative power generation, and cumulative coal consumption during the operation of the target emission source; Based on the cumulative power generation and the cumulative coal consumption, a feature term quantization transformation operation is performed to generate a multi-dimensional feature vector; the multi-dimensional feature vector is then input into a preset long short-term memory network model so that the long short-term memory network model generates a theoretical prediction value of flue gas pollutant concentration based on the multi-dimensional feature vector. The operating condition correction coefficient for the target emission source is determined based on the cumulative power generation; a weighted correction operation is performed on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold. A consistency residual analysis is performed between the instantaneous flue gas concentration measurement value and the theoretical prediction value of the flue gas pollutant concentration to generate a real-time deviation index. When the real-time deviation index exceeds the dynamic allowable deviation threshold, an abnormal situation is determined to exist in the target emission source, and an abnormal source tracing and classification process is performed based on the temporal distribution characteristics of the real-time deviation index within a preset time window; wherein, if the temporal distribution characteristics conform to a preset random discrete distribution pattern, the abnormal situation is determined to be an equipment abnormality and a fault alarm signal is triggered; if the temporal distribution characteristics conform to a preset directional offset pattern or logical conflict pattern, the abnormal situation is determined to be human interference and a cheating alarm signal is triggered.

[0005] Furthermore, the step of performing feature term quantization transformation operation based on the cumulative power generation and the cumulative coal consumption to generate a multi-dimensional feature vector includes: The cumulative power generation and the cumulative coal consumption are time-aligned according to a preset time interval to generate their respective time-aligned initial data sequences. Perform missing value interpolation and completion processing on the initial time-aligned data sequence to obtain the corresponding complete time-synchronized sequence data; Perform a differential operation on the data corresponding to the cumulative power generation in the complete time-synchronized sequence data to generate an instantaneous power generation value; perform a differential operation on the data corresponding to the cumulative coal consumption in the complete time-synchronized sequence data to generate an instantaneous coal consumption rate value; Calculate the ratio of the instantaneous coal consumption rate to the instantaneous power generation value to obtain the unit coal consumption for power generation. The instantaneous power generation value, the instantaneous coal consumption rate value, and the unit coal consumption value are combined and normalized to generate a multi-dimensional feature vector.

[0006] Furthermore, the Long Short-Term Memory network model is trained in the following manner: Obtain a training dataset for predicting flue gas pollutant concentrations; wherein the training dataset for predicting flue gas pollutant concentrations includes several historical time-series samples, each of which includes a historical multi-dimensional feature vector as input and a historical measured value of flue gas concentration as a label. According to the preset time step and batch size, all the historical time series samples are divided into several batches of training sequence samples; The training sequence samples from each batch are sequentially input into the long short-term memory network model to be trained for iterative training until the preset training termination condition is met. In each iteration, the long short-term memory network model performs time-dependent feature extraction on the historical multi-dimensional feature vectors in the current batch of training sequence samples, and outputs the corresponding predicted value of flue gas pollutant concentration. The prediction error between the predicted value of the flue gas pollutant concentration and the measured value of the historical flue gas concentration is calculated by a preset loss function, and a loss function value is generated; a preset optimizer is used to backpropagate and update the network weight parameters of the long short-term memory network model according to the loss function value.

[0007] Furthermore, determining the operating condition correction factor for the target emission source based on the cumulative power generation includes: Perform time-series differential processing on the cumulative power generation to construct a real-time load data sequence; Calculate the statistical dispersion of the real-time load data sequence within a preset sliding time window to generate a load fluctuation intensity index characterizing the stability of the current operating condition; Retrieve a preset working condition association mapping relationship; wherein the working condition association mapping relationship represents the corresponding logic between load fluctuation intensity and correction coefficient value; Based on the load fluctuation intensity index, the operating condition correction coefficient for the target emission source is determined using the operating condition correlation mapping relationship.

[0008] Furthermore, the step of performing a weighted correction operation on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold includes: The product of the preset historical prediction error standard deviation and the working condition correction coefficient is calculated to obtain the corrected standard deviation under the current working condition. Based on the preset confidence coefficient, the corrected standard deviation is calculated by interval expansion, and the calculation result is determined as the dynamic allowable deviation threshold.

[0009] Furthermore, the step of performing a consistency residual analysis on the instantaneous flue gas concentration measurement and the theoretical predicted value of the flue gas pollutant concentration to generate a real-time deviation index includes: The instantaneous flue gas concentration measurement value and the theoretical prediction value of flue gas pollutant concentration are subjected to synchronous differential calculation to obtain a real-time residual value containing positive and negative directional characteristics; The absolute value of the real-time residual is calculated to generate a real-time deviation index that characterizes the magnitude of the prediction deviation.

[0010] Furthermore, the step of determining that the target emission source has an anomaly when the real-time deviation index exceeds the dynamic allowable deviation threshold, and performing anomaly source tracing and classification processing based on the temporal distribution characteristics of the real-time deviation index within a preset time window, includes: The real-time deviation index is compared numerically with the dynamic allowable deviation threshold; If the real-time deviation index is greater than the dynamic allowable deviation threshold, it is determined that there is an abnormality in the target emission source and the current moment is locked as the abnormal starting point. Extract the time-series deviation data of the abnormal starting point within the preset time window, and calculate the continuous duration and waveform change rate of the time-series deviation data; If the continuous duration is less than the preset duration threshold, the waveform change rate exhibits high-frequency oscillation, and the waveform change rate does not exceed the preset abrupt change gradient threshold, the time sequence distribution characteristics are confirmed to conform to the random discrete distribution pattern, the abnormal situation is determined to be an equipment abnormality, and a fault alarm signal is triggered. If the duration of continuous maintenance is greater than or equal to the duration threshold and the waveform change rate does not exceed the abrupt gradient threshold, the timing distribution characteristics are confirmed to conform to the preset directional offset pattern, the abnormal situation is determined to be human interference and a cheating alarm signal is triggered. If the waveform change rate exceeds the preset abrupt change gradient threshold, the timing distribution characteristics are confirmed to conform to the preset logical conflict mode, the abnormal situation is determined to be human interference, and a cheating alarm signal is triggered.

[0011] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0012] One embodiment of the present invention provides a flue gas emission monitoring data verification device, including: a data acquisition module, an intelligent prediction module, a dynamic threshold module, and an anomaly verification module; The data acquisition module is used to acquire instantaneous flue gas concentration measurements, cumulative power generation, and cumulative coal consumption during the operation of the target emission source. The intelligent prediction module is used to perform feature item quantization transformation operation based on the cumulative power generation and the cumulative coal consumption to generate a multi-dimensional feature vector; and input the multi-dimensional feature vector into a preset long short-term memory network model so that the long short-term memory network model generates a theoretical prediction value of flue gas pollutant concentration based on the multi-dimensional feature vector. The dynamic threshold module is used to determine the operating condition correction coefficient of the target emission source based on the cumulative power generation; and to perform a weighted correction operation on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold. The anomaly verification module is used to perform consistency residual analysis on the instantaneous flue gas concentration measurement value and the theoretical predicted value of flue gas pollutant concentration to generate a real-time deviation index. When the real-time deviation index exceeds the dynamic allowable deviation threshold, it is determined that there is an anomaly in the target emission source, and anomaly source tracing and classification processing is performed based on the temporal distribution characteristics of the real-time deviation index within a preset time window. Specifically, if the temporal distribution characteristics conform to a preset random discrete distribution pattern, the anomaly is determined to be an equipment anomaly and a fault alarm signal is triggered. If the temporal distribution characteristics conform to a preset directional offset pattern or logical conflict pattern, the anomaly is determined to be human interference and a cheating alarm signal is triggered.

[0013] Based on the above method embodiments, the present invention provides corresponding electronic device embodiments.

[0014] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the flue gas emission monitoring data verification method according to any one of the above-described method embodiments.

[0015] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments.

[0016] One embodiment of the present invention provides a storage medium storing a computer program thereon, wherein, when the computer program is running, it controls the device where the storage medium is located to execute any of the flue gas emission monitoring data verification methods described in the above-described method embodiments.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method, apparatus, electronic device, and storage medium for verifying flue gas emission monitoring data. The method acquires instantaneous flue gas concentration measurements, cumulative power generation, and cumulative coal consumption during the operation of a target emission source. Based on the cumulative power generation and cumulative coal consumption, a feature term quantization transformation is performed to generate a multi-dimensional feature vector, which is then input into a preset long short-term memory network model to obtain a theoretical prediction value for flue gas pollutant concentration. An operating condition correction coefficient is determined based on the cumulative power generation, and a weighted correction is applied to the standard deviation of historical prediction errors to generate a dynamic allowable deviation threshold. A real-time deviation index is generated by performing a consistency residual analysis between the instantaneous flue gas concentration measurements and the theoretical prediction values ​​of flue gas pollutant concentrations. When the real-time deviation index exceeds the dynamic allowable deviation threshold, an anomaly is determined in the target emission source. Anomaly source tracing and classification are performed based on the temporal distribution characteristics of the real-time deviation index within a preset time window. If the distribution conforms to a random discrete distribution pattern, it is determined to be an equipment anomaly and a fault alarm signal is triggered. If the distribution conforms to a directional offset pattern or a logical conflict pattern, it is determined to be human interference and a cheating alarm signal is triggered.

[0018] This application generates a dynamic allowable deviation threshold by performing a weighted correction operation on the historical error standard deviation based on the operating condition correction coefficient determined by the cumulative power generation. This effectively solves the problem in existing technologies where static thresholds cannot adapt to fluctuations in unit load conditions, leading to a high false alarm rate. Simultaneously, by analyzing the temporal distribution characteristics of the real-time deviation index and identifying random discrete or directional offset patterns, it achieves accurate source tracing and classification of random equipment failures and human interference / cheating behavior, overcoming the shortcomings of existing technologies that can only identify numerical anomalies but cannot distinguish the causes and underlying nature of the anomalies. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a method for verifying flue gas emission monitoring data according to an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of a flue gas emission monitoring data verification device provided in an embodiment of the present invention. Detailed Implementation

[0021] 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.

[0022] like Figure 1As shown, to address the problems in existing technologies where static thresholds are difficult to adapt to fluctuations in unit operating conditions, leading to high false alarm rates during verification, and the inability to effectively distinguish whether data anomalies are caused by random equipment failures or human interference / cheating, an embodiment of the present invention provides a method for verifying flue gas emission monitoring data, comprising at least the following steps: Step S1: Obtain the instantaneous flue gas concentration measurement, cumulative power generation, and cumulative coal consumption during the operation of the target emission source; Specifically, the target emission sources mainly refer to the thermal power generating units of coal-fired power plants. During the continuous operation of the target emission sources, the verification device establishes a communication connection with the continuous emission monitoring system (CEMS) deployed on-site, directly reading the pollutant concentration readings collected in real time by the flue gas sampling probes and analyzers. The sulfur dioxide concentration, nitrogen oxide concentration, and particulate matter concentration read are determined as the instantaneous flue gas concentration measurements. At the same time, the verification device connects to the distributed control system (DCS) or plant-level monitoring information system (SIS) of the generating unit, synchronously collecting load-side and fuel-side data of the unit at a preset high-frequency sampling frequency. Among them, the cumulative power generation refers to the cumulative active power output recorded by the power metering device at the generator unit outlet from the start of the statistical period; the cumulative coal consumption refers to the cumulative weight of coal fed into the furnace recorded by the electronic belt scale of the boiler coal feeder system from the start of the statistical period. The above-mentioned instantaneous flue gas concentration measurements, cumulative power generation, and cumulative coal consumption are all raw time-series data streams with timestamps. By executing step S1, a complete data foundation reflecting the real-time operating status of the unit can be constructed from three dimensions: emission end indicators, power load indicators, and fuel consumption indicators. This provides a real input basis that includes physical causal relationships for subsequent theoretical value prediction based on multi-source data fusion.

[0023] Step S2: Based on the cumulative power generation and the cumulative coal consumption, perform feature term quantization transformation to generate a multi-dimensional feature vector; input the multi-dimensional feature vector into a preset long short-term memory network model so that the long short-term memory network model can generate a theoretical prediction value of flue gas pollutant concentration based on the multi-dimensional feature vector. In a preferred embodiment, the step of performing feature term quantization transformation operation based on the cumulative power generation and the cumulative coal consumption to generate a multi-dimensional feature vector includes: The cumulative power generation and the cumulative coal consumption are time-aligned according to a preset time interval to generate their respective time-aligned initial data sequences. Perform missing value interpolation and completion processing on the initial time-aligned data sequence to obtain the corresponding complete time-synchronized sequence data; Perform a differential operation on the data corresponding to the cumulative power generation in the complete time-synchronized sequence data to generate an instantaneous power generation value; perform a differential operation on the data corresponding to the cumulative coal consumption in the complete time-synchronized sequence data to generate an instantaneous coal consumption rate value; Calculate the ratio of the instantaneous coal consumption rate to the instantaneous power generation value to obtain the unit coal consumption for power generation. The instantaneous power generation value, the instantaneous coal consumption rate value, and the unit coal consumption value are combined and normalized to generate a multi-dimensional feature vector.

[0024] In a preferred embodiment, the long short-term memory network model is trained in the following manner: Obtain a training dataset for predicting flue gas pollutant concentrations; wherein the training dataset for predicting flue gas pollutant concentrations includes several historical time-series samples, each of which includes a historical multi-dimensional feature vector as input and a historical measured value of flue gas concentration as a label. According to the preset time step and batch size, all the historical time series samples are divided into several batches of training sequence samples; The training sequence samples from each batch are sequentially input into the long short-term memory network model to be trained for iterative training until the preset training termination condition is met. In each iteration, the long short-term memory network model performs time-dependent feature extraction on the historical multi-dimensional feature vectors in the current batch of training sequence samples, and outputs the corresponding predicted value of flue gas pollutant concentration. The prediction error between the predicted value of the flue gas pollutant concentration and the measured value of the historical flue gas concentration is calculated by a preset loss function, and a loss function value is generated; a preset optimizer is used to backpropagate and update the network weight parameters of the long short-term memory network model according to the loss function value.

[0025] Specifically, since the cumulative power generation and cumulative coal consumption collected from the distributed control system are usually cumulative values ​​that increase monotonically over time, it is difficult to directly use the cumulative values ​​to characterize the current generator load status and combustion intensity. Therefore, it is necessary to transform them into instantaneous rate indicators with physical meaning through characteristic term quantization transformation.

[0026] First, the data undergoes time axis alignment and completion processing. Since sampling frequencies may differ between different sensors or subsystems, the flue gas emission monitoring data verification method constructs a standard time axis according to a preset time interval (e.g., one data point per minute). The method maps the acquired cumulative power generation and cumulative coal consumption onto the standard time axis, generating a time-aligned initial data sequence. For missing values ​​in the initial time-aligned data sequence caused by communication packet loss or equipment failure, the method uses a linear interpolation algorithm to calculate the values ​​at the missing positions, thereby obtaining a continuous and complete time-synchronized complete sequence of data.

[0027] Next, differential operations are performed on the complete time-synchronized sequence data to extract instantaneous physical quantities. The standard time interval is set to... In the At each sampling time, the cumulative power generation was recorded as follows: The cumulative coal consumption record is as follows The method for verifying flue gas emission monitoring data utilizes the first-order backward difference formula to calculate the... Instantaneous power generation value at each sampling time and instantaneous coal consumption rate values The calculation formula is as follows: in, Indicates the first The cumulative power generation at each sampling time. Indicates the first The cumulative coal consumption at each sampling time point. Through the above differential calculation, the macroscopic cumulative amount is transformed into the microscopic instantaneous power output and instantaneous fuel consumption rate, which can more sensitively capture the dynamic changes in the unit's operating conditions.

[0028] Subsequently, in order to quantify the impact of the generator set's combustion efficiency and coal quality characteristics on emissions, the flue gas emission monitoring data verification method further calculated the ratio of the instantaneous coal consumption rate to the instantaneous power generation, obtaining the unit coal consumption per unit of power generation. The calculation formula is as follows: Unit coal consumption for power generation It can indirectly reflect the quality of the coal being burned and the thermal conversion efficiency of the unit, and is an important characteristic dimension for predicting pollutant concentration.

[0029] After completing the physical quantity calculations, the method for verifying flue gas emission monitoring data will include the instantaneous power generation value. Instantaneous coal consumption rate value and unit coal consumption for power generation The original feature vector is formed by combining these components. Because the dimensions and orders of magnitude of different physical quantities differ significantly (e.g., power is typically in the megawatt range, while coal consumption rate is in the ton range), to accelerate the convergence of subsequent models, the flue gas emission monitoring data verification method uses a maximum-minimum normalization method to process each dimension of the original feature vector separately, generating a multi-dimensional feature vector with values ​​ranging from [0, 1]. .

[0030] After generating multi-dimensional feature vectors, the flue gas emission monitoring data verification method inputs them into a pre-trained Long Short-Term Memory (LSTM) network model. The LSTM model utilizes its internal forget gate, input gate, and output gate structure to process the time-series sequence composed of multi-dimensional feature vectors from multiple consecutive time points, extracting the lag effect characteristics of generator load changes and fuel consumption changes on pollutant generation, and finally outputting the... Theoretical predicted values ​​of flue gas pollutant concentrations at each sampling time. .

[0031] Regarding the training process of the Long Short-Term Memory (LSTM) network model: To ensure the long short-term memory network model possesses accurate nonlinear mapping capabilities, the flue gas emission monitoring data verification method pre-acquires a training dataset for predicting flue gas pollutant concentrations. This training dataset contains several historical time-series samples, each consisting of a historical multi-dimensional feature vector (input) and the corresponding historical measured flue gas concentration value (label).

[0032] During the training phase, the verification method for flue gas emission monitoring data was implemented according to a preset time step. and batch size All historical time-series samples are divided into several batches of training sequence samples. In each iteration of training, the Long Short-Term Memory (LSTM) network model receives the current batch of training sequence samples and calculates the predicted values ​​of flue gas pollutant concentrations through forward propagation. The flue gas emission monitoring data verification method utilizes a pre-defined mean squared error loss function. The prediction bias is calculated using the following formula: in, This represents the number of samples in the current batch. The first one representing the model output The predicted value for each sample, Representing the The historical measured values ​​of flue gas concentration for each sample.

[0033] Using a pre-defined optimizer (such as the Adam optimizer), the flue gas emission monitoring data verification method is based on the loss function value. Calculate the gradient of the network weight parameters and perform backpropagation update operation to continuously adjust the internal weights and biases of the long short-term memory network model until the loss function value converges or reaches the preset number of iterations, thereby obtaining the trained long short-term memory network model.

[0034] By executing step S2, the coarse-grained cumulative statistical data can be refined into multi-dimensional features that reflect the real-time combustion status of the unit. Furthermore, a deep learning model can be used to uncover the complex nonlinear time-series correlation between operating parameters and emission concentrations, providing a high-precision theoretical benchmark value for subsequent anomaly verification.

[0035] Step S3: Determine the operating condition correction coefficient of the target emission source based on the cumulative power generation; perform a weighted correction operation on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold. In a preferred embodiment, determining the operating condition correction factor for the target emission source based on the cumulative power generation includes: Perform time-series differential processing on the cumulative power generation to construct a real-time load data sequence; Calculate the statistical dispersion of the real-time load data sequence within a preset sliding time window to generate a load fluctuation intensity index characterizing the stability of the current operating condition; Retrieve a preset working condition association mapping relationship; wherein the working condition association mapping relationship represents the corresponding logic between load fluctuation intensity and correction coefficient value; Based on the load fluctuation intensity index, the operating condition correction coefficient for the target emission source is determined using the operating condition correlation mapping relationship.

[0036] In a preferred embodiment, the step of performing a weighted correction operation on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold includes: The product of the preset historical prediction error standard deviation and the working condition correction coefficient is calculated to obtain the corrected standard deviation under the current working condition. Based on the preset confidence coefficient, the corrected standard deviation is calculated by interval expansion, and the calculation result is determined as the dynamic allowable deviation threshold.

[0037] Specifically, traditional verification methods typically use a fixed error range as the judgment standard. However, during variable load operation (such as load increases and decreases, and peak shaving), the combustion state of generator units is unstable, and the fluctuation of flue gas emission concentration is naturally greater than during steady-state operation. Using the same standard can easily lead to false alarms. Therefore, the flue gas emission monitoring data verification method introduces an operating condition correction mechanism to dynamically adjust the width of the judgment scale.

[0038] First, to quantify the severity of current operating condition changes, the flue gas emission monitoring data verification method performs time-series differential processing on the cumulative power generation to construct a real-time load data sequence. Since the cumulative power generation is an integral quantity that increases monotonically with time, by performing first-order differential operations on the cumulative power generation, the instantaneous power output of the target emission source at each moment can be reconstructed, i.e., the real-time load.

[0039] Next, the flue gas emission monitoring data verification method calculates the statistical dispersion of the real-time load data sequence within a preset sliding time window, generating a load fluctuation intensity index characterizing the stability of the current operating conditions. The length of the sliding time window is set to... At the current moment Verification method for flue gas emission monitoring data, extracting data from time [time]. At that time Real-time load data segments were obtained, and the load fluctuation intensity index was calculated using the standard deviation formula. The calculation formula is as follows: in, Indicates the first time within the sliding time window Real-time load values ​​at each point in time. This represents the arithmetic mean of real-time load values ​​within the sliding time window. The calculated load fluctuation intensity index... The larger the value, the more frequent or greater the load fluctuations of the generator set during the current time period, and the more unstable the operating conditions.

[0040] Subsequently, the flue gas emission monitoring data verification method retrieves a pre-set operating condition correlation mapping relationship. This mapping relationship pre-stores the correspondence between different load fluctuation intensity ranges and correction coefficient values. Generally, this correspondence exhibits a positive correlation: the higher the load fluctuation intensity index, the larger the corresponding operating condition correction coefficient, to tolerate a reasonable increase in prediction error under varying operating conditions; conversely, the lower the load fluctuation intensity index, the closer the operating condition correction coefficient is to 1, maintaining strict verification standards. The flue gas emission monitoring data verification method then uses the calculated load fluctuation intensity index... Substitute the values ​​into the operating condition correlation mapping relationship for searching or matching to determine the operating condition correction factor of the target emission source at the current moment. .

[0041] After obtaining the operating condition correction factor, the flue gas emission monitoring data verification method performs a weighted correction calculation on the preset historical prediction error standard deviation based on the operating condition correction factor, generating a dynamic allowable deviation threshold. (Preset historical prediction error standard deviation) It is the statistical standard deviation of the residual distribution between the predicted and actual values ​​of a long short-term memory network model under stable baseline conditions during the training or validation phase. It represents the basic predictive ability of the model under ideal conditions.

[0042] The verification method for flue gas emission monitoring data first calculates the preset standard deviation of historical prediction errors. With operating condition correction factor The product of these two factors yields the corrected standard deviation under the current operating conditions. : Subsequently, in order to construct a statistically significant confidence interval, the flue gas emission monitoring data verification method calculates the interval expansion of the corrected standard deviation based on a preset confidence coefficient (e.g., a value of 2 or 3, corresponding to approximately 95% or 99% confidence levels, respectively), and determines the calculation result as the dynamic allowable deviation threshold. The calculation formula is as follows: in, This is the preset confidence level coefficient. Through the above step S3, the verification system can adaptively expand or contract the red line range for anomaly judgment based on the stability or severity of the actual operating conditions of the unit. While ensuring sensitivity to human cheating, it can effectively shield false alarms caused by non-substantial data fluctuations due to normal load changes in the unit.

[0043] Step S4: Perform a consistency residual analysis on the instantaneous flue gas concentration measurement value and the theoretical predicted value of flue gas pollutant concentration to generate a real-time deviation index; when the real-time deviation index exceeds the dynamic allowable deviation threshold, determine that there is an abnormality in the target emission source, and perform abnormal source tracing and classification processing based on the temporal distribution characteristics of the real-time deviation index within a preset time window; wherein, if the temporal distribution characteristics conform to a preset random discrete distribution pattern, determine that the abnormality is an equipment abnormality and trigger a fault alarm signal; if the temporal distribution characteristics conform to a preset directional offset pattern or logical conflict pattern, determine that the abnormality is human interference and trigger a cheating alarm signal.

[0044] In a preferred embodiment, the step of performing a consistency residual analysis between the instantaneous flue gas concentration measurement and the theoretically predicted flue gas pollutant concentration to generate a real-time deviation index includes: The instantaneous flue gas concentration measurement value and the theoretical prediction value of flue gas pollutant concentration are subjected to synchronous differential calculation to obtain a real-time residual value containing positive and negative directional characteristics; The absolute value of the real-time residual is calculated to generate a real-time deviation index that characterizes the magnitude of the prediction deviation.

[0045] In a preferred embodiment, the step of determining that the target emission source has an anomaly when the real-time deviation index exceeds the dynamic allowable deviation threshold, and performing anomaly source tracing and classification processing based on the temporal distribution characteristics of the real-time deviation index within a preset time window, includes: The real-time deviation index is compared numerically with the dynamic allowable deviation threshold; If the real-time deviation index is greater than the dynamic allowable deviation threshold, it is determined that there is an abnormality in the target emission source and the current moment is locked as the abnormal starting point. Extract the time-series deviation data of the abnormal starting point within the preset time window, and calculate the continuous duration and waveform change rate of the time-series deviation data; If the continuous duration is less than the preset duration threshold, the waveform change rate exhibits high-frequency oscillation, and the waveform change rate does not exceed the preset abrupt change gradient threshold, the time sequence distribution characteristics are confirmed to conform to the random discrete distribution pattern, the abnormal situation is determined to be an equipment abnormality, and a fault alarm signal is triggered. If the duration of continuous maintenance is greater than or equal to the duration threshold and the waveform change rate does not exceed the abrupt gradient threshold, the timing distribution characteristics are confirmed to conform to the preset directional offset pattern, the abnormal situation is determined to be human interference and a cheating alarm signal is triggered. If the waveform change rate exceeds the preset abrupt change gradient threshold, the timing distribution characteristics are confirmed to conform to the preset logical conflict mode, the abnormal situation is determined to be human interference, and a cheating alarm signal is triggered.

[0046] Specifically, to accurately quantify the discrepancy between monitoring data and theoretical benchmarks, the flue gas emission monitoring data verification method first performs a consistency residual analysis. The flue gas emission monitoring data verification method uses instantaneous flue gas concentration measurements at the same sampling time... Compared with theoretical predicted values ​​of flue gas pollutant concentrations Synchronous differential calculation is performed. Since instantaneous flue gas concentration measurements may be higher or lower than theoretical values, the direct difference results in values ​​with both positive and negative directional characteristics. To standardize the measurement of the magnitude of deviation, the flue gas emission monitoring data verification method performs absolute value calculations on the real-time residual values ​​containing positive and negative directional characteristics, generating a real-time deviation index characterizing the magnitude of the prediction deviation. The calculation formula is as follows: Real-time deviation index This intuitively reflects the extent to which the current monitoring data deviates from the physical generation mechanism.

[0047] Subsequently, the method for verifying flue gas emission monitoring data will include real-time deviation indicators. Compared with the dynamic allowable deviation threshold generated in step S3 Perform numerical comparison. If the real-time deviation index... Greater than the dynamic allowable deviation threshold This indicates that the current data deviation has exceeded the reasonable error range allowed by operating condition fluctuations. Based on this, the flue gas emission monitoring data verification method determines that there is an anomaly in the target emission source and immediately identifies the current moment as the starting point of the anomaly. .

[0048] In response to the anomaly determination results, and in order to determine the cause of the anomaly, the flue gas emission monitoring data verification method extracts data from the point of origin of the anomaly. The data sequence of time-series deviations is selected at the beginning and within a preset time window (e.g., the next 5 minutes or 10 sampling points). The method for verifying flue gas emission monitoring data extracts features from the time-series deviation data sequence, calculating features in two core dimensions: continuous duration. and waveform change rate .

[0049] Among them, continuous duration This refers to the length of time that the real-time deviation index remains continuously above the dynamic allowable deviation threshold. Waveform change rate. The gradient, used to characterize the drasticness of data changes, can be approximated using first-order differences or derivatives. The value of each point is Then the waveform change rate The calculation formula is as follows: Based on the above two characteristics, the flue gas emission monitoring data verification method performs anomaly source tracing and classification processing. To ensure the accuracy and logical mutual exclusion of anomaly cause determination, this embodiment adopts a priority-based determination logic, as follows: First, determine if a "logical conflict pattern" exists. The flue gas emission monitoring data verification method detects the waveform change rate. Does the data exceed a preset abrupt change threshold? If it does, it indicates a precipitous change in the data that violates physical inertia (e.g., instantaneously returning to zero or jumping to a fixed value). Regardless of the duration of this change, the time-series distribution characteristics are confirmed to conform to a preset logical conflict pattern. This pattern typically corresponds to blatant tampering, such as unplugging the sampling probe, directly setting data in software, or cutting the signal line. Therefore, the flue gas emission monitoring data verification method determines the anomaly as human interference and triggers a cheating alarm signal.

[0050] Secondly, if the waveform change rate does not exceed the abrupt gradient threshold, it is further determined whether a "directional shift mode" exists. The flue gas emission monitoring data verification method detects the continuous maintenance duration. Is the duration greater than or equal to a preset duration threshold? If this duration condition is met and the waveform change rate does not exceed the limit, it indicates that the data has a long-term, stable deviation (i.e., a "boiling frog" style cheating method), confirming that the time-series distribution characteristics conform to a preset directional offset pattern. This pattern typically corresponds to manually modifying the slope parameter, adding a fixed resistance, or physically blocking the sampling pipeline, causing the monitored value to be consistently lower than the true value for a long period. Therefore, the flue gas emission monitoring data verification method determines the abnormal situation as human interference and triggers a cheating alarm signal.

[0051] Finally, if the waveform change rate does not exceed the abrupt change gradient threshold and the duration of continuous maintenance is less than the persistence threshold, it is determined whether a "random discrete distribution pattern" exists. Under this condition, if the waveform change rate exhibits high-frequency oscillation (i.e., the data jumps rapidly near the threshold, but the amplitude does not reach the abrupt change standard, and the derivative sign switches frequently), it is confirmed that the time-series distribution characteristics conform to the preset random discrete distribution pattern. This pattern indicates that the anomaly is temporary and random, usually corresponding to non-human factors such as sensor electromagnetic interference, signal jitter caused by probe dust accumulation, or air bubbles in the pipeline. Therefore, the flue gas emission monitoring data verification method determines the abnormal situation as equipment malfunction and triggers a fault alarm signal, prompting maintenance personnel to check the hardware equipment.

[0052] By executing step S4, the judgment can be upgraded from a single value exceeding the standard to intelligent identification of abnormal patterns. Through the above mutually exclusive classification logic, it can effectively distinguish between "false anomalies" caused by equipment aging or environmental interference and "real cheating" caused by deliberate human tampering, which significantly improves the accuracy of regulatory enforcement and the credibility of evidence.

[0053] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0054] like Figure 2 As shown, an embodiment of the present invention provides a flue gas emission monitoring data verification device, including: a data acquisition module, an intelligent prediction module, a dynamic threshold module, and an anomaly verification module; The data acquisition module is used to acquire instantaneous flue gas concentration measurements, cumulative power generation, and cumulative coal consumption during the operation of the target emission source. The intelligent prediction module is used to perform feature item quantization transformation operation based on the cumulative power generation and the cumulative coal consumption to generate a multi-dimensional feature vector; and input the multi-dimensional feature vector into a preset long short-term memory network model so that the long short-term memory network model generates a theoretical prediction value of flue gas pollutant concentration based on the multi-dimensional feature vector. The dynamic threshold module is used to determine the operating condition correction coefficient of the target emission source based on the cumulative power generation; and to perform a weighted correction operation on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold. The anomaly verification module is used to perform consistency residual analysis on the instantaneous flue gas concentration measurement value and the theoretical predicted value of flue gas pollutant concentration to generate a real-time deviation index. When the real-time deviation index exceeds the dynamic allowable deviation threshold, it is determined that there is an anomaly in the target emission source, and anomaly source tracing and classification processing is performed based on the temporal distribution characteristics of the real-time deviation index within a preset time window. Specifically, if the temporal distribution characteristics conform to a preset random discrete distribution pattern, the anomaly is determined to be an equipment anomaly and a fault alarm signal is triggered. If the temporal distribution characteristics conform to a preset directional offset pattern or logical conflict pattern, the anomaly is determined to be human interference and a cheating alarm signal is triggered.

[0055] In a preferred embodiment, the intelligent prediction module performs a feature term quantization transformation operation based on the cumulative power generation and the cumulative coal consumption to generate a multi-dimensional feature vector, including: The cumulative power generation and the cumulative coal consumption are time-aligned according to a preset time interval to generate their respective time-aligned initial data sequences. Perform missing value interpolation and completion processing on the initial time-aligned data sequence to obtain the corresponding complete time-synchronized sequence data; Perform a differential operation on the data corresponding to the cumulative power generation in the complete time-synchronized sequence data to generate an instantaneous power generation value; perform a differential operation on the data corresponding to the cumulative coal consumption in the complete time-synchronized sequence data to generate an instantaneous coal consumption rate value; Calculate the ratio of the instantaneous coal consumption rate to the instantaneous power generation value to obtain the unit coal consumption for power generation. The instantaneous power generation value, the instantaneous coal consumption rate value, and the unit coal consumption value are combined and normalized to generate a multi-dimensional feature vector.

[0056] In a preferred embodiment, the flue gas emission monitoring data verification device further includes: a long short-term memory network model module; The Long Short-Term Memory (LSTM) network model module is used to acquire a training dataset for predicting flue gas pollutant concentrations. This training dataset includes several historical time-series samples, each containing a historical multi-dimensional feature vector as input and a historical measured flue gas concentration value as a label. All historical time-series samples are divided into several batches of training sequence samples according to a preset time step and batch size. Each batch of training sequence samples is sequentially input into the LTM network model to be trained for iterative training until a preset training termination condition is met. In each iteration, the LTM network model performs time-dependent feature extraction on the historical multi-dimensional feature vectors in the current batch of training sequence samples, outputting the corresponding predicted flue gas pollutant concentration value. A preset loss function is used to calculate the prediction error between the predicted flue gas pollutant concentration value and the historical measured flue gas concentration value, generating a loss function value. A preset optimizer is used to backpropagate and update the network weight parameters of the LTM network model based on the loss function value.

[0057] In a preferred embodiment, the dynamic threshold module determines the operating condition correction coefficient for the target emission source based on the cumulative power generation, including: Perform time-series differential processing on the cumulative power generation to construct a real-time load data sequence; Calculate the statistical dispersion of the real-time load data sequence within a preset sliding time window to generate a load fluctuation intensity index characterizing the stability of the current operating condition; Retrieve a preset working condition association mapping relationship; wherein the working condition association mapping relationship represents the corresponding logic between load fluctuation intensity and correction coefficient value; Based on the load fluctuation intensity index, the operating condition correction coefficient for the target emission source is determined using the operating condition correlation mapping relationship.

[0058] In a preferred embodiment, the dynamic threshold module performs a weighted correction operation on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold, including: The product of the preset historical prediction error standard deviation and the working condition correction coefficient is calculated to obtain the corrected standard deviation under the current working condition. Based on the preset confidence coefficient, the corrected standard deviation is calculated by interval expansion, and the calculation result is determined as the dynamic allowable deviation threshold.

[0059] In a preferred embodiment, the anomaly verification module performs a consistency residual analysis on the instantaneous flue gas concentration measurement and the theoretical predicted value of the flue gas pollutant concentration to generate a real-time deviation index, including: The instantaneous flue gas concentration measurement value and the theoretical prediction value of flue gas pollutant concentration are subjected to synchronous differential calculation to obtain a real-time residual value containing positive and negative directional characteristics; The absolute value of the real-time residual is calculated to generate a real-time deviation index that characterizes the magnitude of the prediction deviation.

[0060] In a preferred embodiment, the anomaly verification module determines that the target emission source has an anomaly when the real-time deviation index exceeds the dynamic allowable deviation threshold, and performs anomaly source tracing and classification processing based on the temporal distribution characteristics of the real-time deviation index within a preset time window, including: The real-time deviation index is compared numerically with the dynamic allowable deviation threshold; If the real-time deviation index is greater than the dynamic allowable deviation threshold, it is determined that there is an abnormality in the target emission source and the current moment is locked as the abnormal starting point. Extract the time-series deviation data of the abnormal starting point within the preset time window, and calculate the continuous duration and waveform change rate of the time-series deviation data; If the continuous duration is less than the preset duration threshold, the waveform change rate exhibits high-frequency oscillation, and the waveform change rate does not exceed the preset abrupt change gradient threshold, the time sequence distribution characteristics are confirmed to conform to the random discrete distribution pattern, the abnormal situation is determined to be an equipment abnormality, and a fault alarm signal is triggered. If the duration of continuous maintenance is greater than or equal to the duration threshold and the waveform change rate does not exceed the abrupt gradient threshold, the timing distribution characteristics are confirmed to conform to the preset directional offset pattern, the abnormal situation is determined to be human interference and a cheating alarm signal is triggered. If the waveform change rate exceeds the preset abrupt change gradient threshold, the timing distribution characteristics are confirmed to conform to the preset logical conflict mode, the abnormal situation is determined to be human interference, and a cheating alarm signal is triggered.

[0061] It should be noted that the embodiments of the device described above correspond to the embodiments of the present invention described above, and can realize the flue gas emission monitoring data verification method described in any one of the above embodiments of the present invention. Furthermore, the embodiments of the device described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without creative effort.

[0062] Based on the above-described method embodiments of the present invention, a corresponding embodiment of an electronic device is provided.

[0063] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the flue gas emission monitoring data verification method according to any one of the present invention, or, when the processor executes the computer program, it implements the functions of each module in the above-described device embodiments.

[0064] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device.

[0065] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0066] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0067] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0068] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments; Another embodiment of the present invention provides a storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute any of the above-described flue gas emission monitoring data verification methods of the present invention.

[0069] The aforementioned storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0070] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0071] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for verifying flue gas emission monitoring data, characterized in that, include: Acquire instantaneous flue gas concentration measurements, cumulative power generation, and cumulative coal consumption during the operation of the target emission source; Based on the cumulative power generation and the cumulative coal consumption, a feature term quantization transformation operation is performed to generate a multi-dimensional feature vector; the multi-dimensional feature vector is then input into a preset long short-term memory network model so that the long short-term memory network model generates a theoretical prediction value of flue gas pollutant concentration based on the multi-dimensional feature vector. The operating condition correction coefficient for the target emission source is determined based on the cumulative power generation; a weighted correction operation is performed on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold. A consistency residual analysis is performed between the instantaneous flue gas concentration measurement value and the theoretical prediction value of the flue gas pollutant concentration to generate a real-time deviation index. When the real-time deviation index exceeds the dynamic allowable deviation threshold, an abnormal situation is determined to exist in the target emission source, and an abnormal source tracing and classification process is performed based on the temporal distribution characteristics of the real-time deviation index within a preset time window; wherein, if the temporal distribution characteristics conform to a preset random discrete distribution pattern, the abnormal situation is determined to be an equipment abnormality and a fault alarm signal is triggered; if the temporal distribution characteristics conform to a preset directional offset pattern or logical conflict pattern, the abnormal situation is determined to be human interference and a cheating alarm signal is triggered.

2. The method for verifying flue gas emission monitoring data as described in claim 1, characterized in that, The step of performing feature term quantization transformation operation based on the cumulative power generation and the cumulative coal consumption to generate a multi-dimensional feature vector includes: The cumulative power generation and the cumulative coal consumption are time-aligned according to a preset time interval to generate their respective time-aligned initial data sequences. Perform missing value interpolation and completion processing on the initial time-aligned data sequence to obtain the corresponding complete time-synchronized sequence data; Perform a differential operation on the data corresponding to the cumulative power generation in the complete time-synchronized sequence data to generate an instantaneous power generation value; perform a differential operation on the data corresponding to the cumulative coal consumption in the complete time-synchronized sequence data to generate an instantaneous coal consumption rate value; Calculate the ratio of the instantaneous coal consumption rate to the instantaneous power generation value to obtain the unit coal consumption for power generation. The instantaneous power generation value, the instantaneous coal consumption rate value, and the unit coal consumption value are combined and normalized to generate a multi-dimensional feature vector.

3. The method for verifying flue gas emission monitoring data as described in claim 2, characterized in that, The Long Short-Term Memory network model was trained using the following method: Obtain a training dataset for predicting flue gas pollutant concentrations; wherein the training dataset for predicting flue gas pollutant concentrations includes several historical time-series samples, each of which includes a historical multi-dimensional feature vector as input and a historical measured value of flue gas concentration as a label. According to the preset time step and batch size, all the historical time series samples are divided into several batches of training sequence samples; The training sequence samples from each batch are sequentially input into the long short-term memory network model to be trained for iterative training until the preset training termination condition is met. In each iteration, the long short-term memory network model performs time-dependent feature extraction on the historical multi-dimensional feature vectors in the current batch of training sequence samples, and outputs the corresponding predicted value of flue gas pollutant concentration. The prediction error between the predicted value of the flue gas pollutant concentration and the measured value of the historical flue gas concentration is calculated by a preset loss function, and a loss function value is generated; a preset optimizer is used to backpropagate and update the network weight parameters of the long short-term memory network model according to the loss function value.

4. The method for verifying flue gas emission monitoring data as described in claim 3, characterized in that, The determination of the operating condition correction factor for the target emission source based on the cumulative power generation includes: Perform time-series differential processing on the cumulative power generation to construct a real-time load data sequence; Calculate the statistical dispersion of the real-time load data sequence within a preset sliding time window to generate a load fluctuation intensity index characterizing the stability of the current operating condition; Retrieve a preset working condition association mapping relationship; wherein the working condition association mapping relationship represents the corresponding logic between load fluctuation intensity and correction coefficient value; Based on the load fluctuation intensity index, the operating condition correction coefficient for the target emission source is determined using the operating condition correlation mapping relationship.

5. The method for verifying flue gas emission monitoring data as described in claim 4, characterized in that, The step of performing a weighted correction operation on the preset historical prediction error standard deviation based on the operating condition correction coefficient to generate a dynamic allowable deviation threshold includes: The product of the preset historical prediction error standard deviation and the working condition correction coefficient is calculated to obtain the corrected standard deviation under the current working condition. Based on the preset confidence coefficient, the corrected standard deviation is calculated by interval expansion, and the calculation result is determined as the dynamic allowable deviation threshold.

6. The method for verifying flue gas emission monitoring data as described in claim 5, characterized in that, The step of performing a consistency residual analysis between the instantaneous flue gas concentration measurement and the theoretical predicted value of the flue gas pollutant concentration to generate a real-time deviation index includes: The instantaneous flue gas concentration measurement value and the theoretical prediction value of flue gas pollutant concentration are subjected to synchronous differential calculation to obtain a real-time residual value containing positive and negative directional characteristics; The absolute value of the real-time residual is calculated to generate a real-time deviation index that characterizes the magnitude of the prediction deviation.

7. The method for verifying flue gas emission monitoring data as described in claim 6, characterized in that, When the real-time deviation index exceeds the dynamic allowable deviation threshold, it is determined that there is an anomaly in the target emission source, and anomaly source tracing and classification processing is performed based on the temporal distribution characteristics of the real-time deviation index within a preset time window, including: The real-time deviation index is compared numerically with the dynamic allowable deviation threshold; If the real-time deviation index is greater than the dynamic allowable deviation threshold, it is determined that there is an abnormality in the target emission source and the current moment is locked as the abnormal starting point. Extract the time-series deviation data of the abnormal starting point within the preset time window, and calculate the continuous duration and waveform change rate of the time-series deviation data; If the continuous duration is less than the preset duration threshold, the waveform change rate exhibits high-frequency oscillation, and the waveform change rate does not exceed the preset abrupt change gradient threshold, the time sequence distribution characteristics are confirmed to conform to the random discrete distribution pattern, the abnormal situation is determined to be an equipment abnormality, and a fault alarm signal is triggered. If the duration of continuous maintenance is greater than or equal to the duration threshold and the waveform change rate does not exceed the abrupt gradient threshold, the timing distribution characteristics are confirmed to conform to the preset directional offset pattern, the abnormal situation is determined to be human interference and a cheating alarm signal is triggered. If the waveform change rate exceeds the preset abrupt change gradient threshold, the timing distribution characteristics are confirmed to conform to the preset logical conflict mode, the abnormal situation is determined to be human interference, and a cheating alarm signal is triggered.

8. A device for verifying flue gas emission monitoring data, characterized in that, include: The module includes a data acquisition module, an intelligent prediction module, a dynamic threshold module, and an anomaly verification module. The data acquisition module is used to acquire instantaneous flue gas concentration measurements, cumulative power generation, and cumulative coal consumption during the operation of the target emission source. The intelligent prediction module is used to perform feature item quantization transformation operation based on the cumulative power generation and the cumulative coal consumption to generate a multi-dimensional feature vector; and input the multi-dimensional feature vector into a preset long short-term memory network model so that the long short-term memory network model generates a theoretical prediction value of flue gas pollutant concentration based on the multi-dimensional feature vector. The dynamic threshold module is used to determine the operating condition correction coefficient of the target emission source based on the cumulative power generation. Based on the operating condition correction coefficient, a weighted correction operation is performed on the preset historical prediction error standard deviation to generate a dynamic allowable deviation threshold. The anomaly verification module is used to perform consistency residual analysis on the instantaneous flue gas concentration measurement value and the theoretical predicted value of flue gas pollutant concentration to generate a real-time deviation index. When the real-time deviation index exceeds the dynamic allowable deviation threshold, it is determined that there is an anomaly in the target emission source, and anomaly source tracing and classification processing is performed based on the temporal distribution characteristics of the real-time deviation index within a preset time window. Specifically, if the temporal distribution characteristics conform to a preset random discrete distribution pattern, the anomaly is determined to be an equipment anomaly and a fault alarm signal is triggered. If the temporal distribution characteristics conform to a preset directional offset pattern or logical conflict pattern, the anomaly is determined to be human interference and a cheating alarm signal is triggered.

9. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the flue gas emission monitoring data verification method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the storage medium is located to perform the flue gas emission monitoring data verification method as described in any one of claims 1 to 7.

Images