A city gas monitoring method based on narrowband internet of things

Abstract

This invention discloses a method for urban gas monitoring based on narrowband Internet of Things (IoT), comprising the following steps: Terminals installed at various monitoring points in the gas pipeline network collect gas concentration, pressure, and flow data, construct a monitoring data sequence, and after preprocessing the data, perform recursive calculations of statistical mean and cumulative deviation at the terminal side to generate a mutation statistical state quantity. Based on the mutation statistical state quantity, Page-Hinkley online detection is performed to obtain anomaly judgment results and anomaly intensity parameters. The terminal communication status is determined according to the anomaly judgment results, and a corresponding set of data to be uploaded is generated. A multi-terminal scheduling model is constructed based on the status information of each terminal, calculating the priority of the Whittle index of each terminal and controlling the data upload order. After receiving the uploaded data, the remote monitoring platform generates a gas operation status assessment result and anomaly alarm result for the corresponding monitoring point. This invention achieves unified processing of multi-terminal collaborative monitoring and data transmission control.

Description

Technical Field

[0001] This invention relates to the field of narrowband Internet of Things (IoT) technology, and in particular to a method for monitoring urban gas supply based on narrowband IoT. Background Technology

[0002] During the operation of urban gas pipeline networks, it is necessary to continuously monitor key parameters such as gas concentration, pressure, and flow rate to ensure the safety and stability of gas transmission and distribution. Current technologies typically employ narrowband Internet of Things (IoT) based terminal devices to collect data from each monitoring point and upload it to a remote monitoring platform for unified analysis and management via a preset wireless network.

[0003] In practical applications, existing technologies generally use fixed thresholds or simple rules to judge anomalies in monitoring data, making it difficult to accurately identify gradual anomalies or early abnormal states, resulting in insufficient anomaly detection sensitivity or a high false alarm rate. At the same time, existing data transmission methods mostly use periodic reporting or uniform frequency reporting mechanisms, failing to differentiate processing based on the importance or degree of anomaly of the data, resulting in a large amount of invalid data consuming wireless transmission resources.

[0004] In scenarios with multiple terminals accessing concurrently, existing technologies lack dynamic scheduling mechanisms for multiple terminals. They typically employ fixed polling or uniform priority strategies for data upload control, failing to effectively differentiate the urgency of data from different terminals. This results in critical abnormal data not being uploaded in a timely manner, impacting the real-time performance and reliability of remote monitoring. Furthermore, the terminal side's failure to dynamically adjust communication behavior based on abnormal states leads to higher overall power consumption, which is detrimental to the long-term operation of narrowband IoT terminals.

[0005] Therefore, how to provide a method for urban gas monitoring based on narrowband Internet of Things is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a method for urban gas monitoring based on narrowband Internet of Things (IoT). This invention introduces a mutation detection mechanism at the terminal side to perform real-time analysis of gas monitoring data and combines a multi-terminal dynamic scheduling strategy to uniformly control the data transmission process. It describes in detail the complete processing flow from data acquisition, anomaly identification, communication status determination to priority scheduling and remote evaluation, and has the advantages of sensitive anomaly detection, orderly data transmission, high resource utilization efficiency, and strong real-time remote monitoring.

[0007] A method for monitoring urban gas supply based on narrowband Internet of Things according to an embodiment of the present invention includes the following steps: Narrowband IoT terminals deployed at various monitoring points in the urban gas pipeline network collect gas concentration, pressure, and flow data, and obtain corresponding terminal remaining power information and wireless link status information to construct the original time-series data sequence; The original time-series data sequence is subjected to invalid value removal and filtering to obtain a continuous monitoring sequence; The statistical mean and cumulative deviation are recursively calculated on the terminal side based on the continuous monitoring sequence, and the historical minimum value is determined based on the cumulative deviation to generate the statistical state of abrupt change. The mutation statistics state quantity is subjected to Page-Hinkley online detection, the offset between the cumulative deviation and the historical minimum value is calculated, and the anomaly judgment result and anomaly intensity parameter are generated based on the offset. The terminal communication status is determined based on the anomaly judgment result and anomaly intensity parameter, and a corresponding set of data to be uploaded is generated based on the terminal communication status. Based on the set of data to be uploaded, a corresponding data waiting time is generated, and the abnormal intensity parameter, data waiting time, terminal remaining power information and wireless link status information are constructed into a terminal status vector. A multi-terminal dynamic scheduling model is established based on the terminal state vector, and the Whittle index of each terminal is calculated in the multi-terminal dynamic scheduling model. A transmission priority sequence is generated based on the Whittle index. Based on the transmission priority sequence, each terminal is controlled to upload data from the data set to be uploaded through the narrowband IoT wireless access channel. After receiving the uploaded data, the remote monitoring platform generates the gas operation status assessment result and abnormal alarm result for the corresponding monitoring point.

[0008] Optionally, a narrowband IoT terminal collects gas concentration, pressure, and flow data, and obtains corresponding information on remaining terminal battery power and wireless link status, including: At each sampling time, the narrowband IoT terminal acquires the corresponding gas concentration data through the gas concentration sensor. The gas concentration data is the measured value of the concentration of gas components per unit volume. The narrowband IoT terminal acquires the corresponding pressure data through the pressure sensor at the sampling time corresponding to the gas concentration data. The pressure data is the instantaneous pressure measurement value of the gas in the pipeline. The narrowband IoT terminal acquires the corresponding flow data through the flow sensor at the sampling time corresponding to the gas concentration data and pressure data. The flow data is the measured value of the gas volume or mass flow rate through the pipe cross section per unit time. The gas concentration data, pressure data, and flow data obtained at the same sampling time are correlated to form the monitoring data item at the sampling time; The remaining power information of the terminal at the corresponding sampling time is read based on the monitoring data items. The remaining power information of the terminal is the remaining power percentage or remaining capacity value output by the terminal power module. The wireless link status information at the corresponding sampling time is obtained based on the monitoring data item, and the wireless link status information is used as the link status data corresponding to the monitoring data item. The monitoring data items, terminal power data, and link status data are combined to obtain the original time-series data sequence.

[0009] Optionally, invalid value removal and filtering of the original time-series data sequence includes: According to the sampling time order in the original time series data sequence, the monitoring data items corresponding to each sampling time are extracted sequentially, and the gas concentration data, pressure data and flow data in the monitoring data items are read respectively; The gas concentration data, pressure data, and flow data are compared with their respective preset effective ranges. Data that exceeds the preset effective range is identified as invalid values, and monitoring data items containing invalid values ​​are deleted. After deleting monitoring data items containing invalid values, data sequences were constructed for the gas concentration data, pressure data, and flow data in the remaining monitoring data items according to the sampling time order. Each data sequence is processed using a sliding window filtering method. At each sampling time, the data at the sampling time and the data at the sampling times before and after it are selected to form a filtering window. The data within the filtering window are averaged to obtain the filtering result at the corresponding sampling time. The filtered results corresponding to the gas concentration data, the pressure data, and the flow data are correlated according to the original sampling time to obtain the filtered monitoring data items. The filtered monitoring data items are arranged in order of sampling time to generate a continuous monitoring sequence.

[0010] Optionally, the generation of mutation statistical states includes: According to the sampling time sequence in the continuous monitoring sequence, the monitoring data items corresponding to each sampling time are extracted sequentially, and the gas concentration data, pressure data and flow data in the monitoring data items corresponding to each sampling time are used as the current sampling data. Based on the current sampling data, recursive calculations are performed on the narrowband IoT terminal side to obtain the average concentration, average pressure, and average flow rate corresponding to the current sampling time. The average concentration, average pressure, and average flow rate are then used as the average statistical value at the current sampling time. Based on the difference between the current sampled data and the corresponding statistical mean, the cumulative deviations of concentration, pressure, and flow rate corresponding to the current sampling time are recursively accumulated to obtain the cumulative deviations of concentration, pressure, and flow rate at the current sampling time. The cumulative deviations of concentration, pressure, and flow rate are then used as the cumulative deviations at the current sampling time. The cumulative deviation at the current sampling time is compared with the historical minimum value determined at the previous sampling time. When the cumulative deviation at the current sampling time is less than the historical minimum value determined at the previous sampling time, the cumulative deviation at the current sampling time is determined as the historical minimum value at the current sampling time. When the cumulative deviation at the current sampling time is greater than or equal to the historical minimum value determined at the previous sampling time, the historical minimum value determined at the previous sampling time is kept as the historical minimum value at the current sampling time. By correlating the statistical mean, cumulative deviation, and historical minimum value at the current sampling time, a sudden change statistical state quantity at the current sampling time is generated.

[0011] Optional, Page-Hinkley online detection includes: Read the mutation statistics state quantities corresponding to each sampling time in sequence according to the sampling time order, and extract the statistical mean, cumulative deviation and historical minimum value of the corresponding sampling time from the mutation statistics state quantities. A stable monitoring segment is constructed based on the monitoring data items in the continuous monitoring sequence arranged in order of sampling time prior to the current sampling time, and a stable reference state is generated based on the gas concentration data, pressure data, and flow data in the stable monitoring segment; The gas concentration data, pressure data, and flow data in the monitoring data items at the current sampling time are compared with the stable reference state to generate the residual detection value corresponding to the current sampling time. The residual detection value is used as the Page-Hinkley online detection input value corresponding to the current sampling time, and the detection offset corresponding to the current sampling time is generated based on the input value and the statistical mean, cumulative deviation and historical minimum value corresponding to the current sampling time. The detection offset is compared with a preset detection threshold. When the detection offset is less than the preset detection threshold, a normal judgment result for the current sampling time is generated. When the detection offset is greater than or equal to the preset detection threshold, the current sampling time is determined as the candidate anomaly start time, and the historical minimum value corresponding to the candidate anomaly start time is latched. At the same time, an anomaly segment cache is established, and the detection offset corresponding to the candidate anomaly start time is written into the anomaly segment cache. After establishing the abnormal segment cache, continue to read the mutation statistical state quantity corresponding to the sampling time in the order of sampling time, and extract the statistical mean, cumulative deviation and historical minimum value from the mutation statistical state quantity. At the same time, generate the residual detection value corresponding to the sampling time based on the stable reference state and the monitoring data item corresponding to the sampling time. The residual detection value corresponding to the sampling time is used as the Page-Hinkley online detection input value for the corresponding sampling time. Based on the input value, the corresponding statistical mean, the corresponding cumulative deviation, and the latched historical minimum value, the detection offset corresponding to the sampling time is generated. The detection offsets that are greater than or equal to the preset detection threshold are written into the abnormal segment cache in the order of sampling time. Continuous abnormal segments are generated based on the detection offsets written in the abnormal segment cache in the order of sampling time, and the continuous abnormal segments are determined as the abnormal judgment results. The maximum detection offset in the continuous abnormal segments is determined as the abnormal intensity parameter. When the detection offset in the abnormal segment cache does not generate a continuous abnormal segment, the abnormal segment cache is cleared and the historical minimum value is released from latching. After generating the anomaly determination result, the stable monitoring segment is reconstructed based on the monitoring data items in the continuous monitoring sequence arranged in the order of sampling time after the anomaly determination result, and the stable reference state corresponding to the next sampling time is generated based on the reconstructed stable monitoring segment.

[0012] Optionally, generating the corresponding data set to be uploaded includes: Read the anomaly judgment result and anomaly intensity parameter corresponding to the current sampling time, and determine the terminal communication status corresponding to the current sampling time based on the anomaly judgment result; When the anomaly determination result corresponding to the current sampling time is normal, the terminal communication status corresponding to the current sampling time is determined as the summary reporting status, and the summary data corresponding to the current sampling time is generated based on the monitoring data item corresponding to the current sampling time. When the anomaly determination result corresponding to the current sampling time is an anomaly, the terminal communication status corresponding to the current sampling time is determined as the event reporting status, and the event data corresponding to the current sampling time is generated based on the monitoring data item corresponding to the current sampling time and the anomaly intensity parameter corresponding to the current sampling time. Associate the terminal communication status at the current sampling time with the summary data and event data at the current sampling time to generate the data unit at the current sampling time; Read the data units corresponding to each sampling time sequentially according to the sampling time order, and merge multiple consecutive data units with the same terminal communication status to generate a data segment corresponding to the terminal communication status. The summary data and event data corresponding to each data unit in the data segment are arranged in the order of sampling time to generate a set of data to be uploaded corresponding to the terminal communication status.

[0013] Optionally, constructing the terminal state vector includes: Read the data set to be uploaded corresponding to the current terminal, and extract the data units arranged in order of sampling time from the data set to be uploaded; Based on the data units arranged in order of sampling time in the dataset to be uploaded, extract the data unit corresponding to the earliest sampling time, and determine the sampling time of the data unit corresponding to the earliest sampling time as the starting time of the dataset to be uploaded. The data waiting time for the current terminal is generated based on the time difference between the current time and the start time. Read the abnormal intensity parameter, remaining battery power information and wireless link status information corresponding to the current terminal, and determine the abnormal intensity parameter, data waiting time, remaining battery power information and wireless link status information as the abnormal state component, waiting state component, battery status component and link status component corresponding to the terminal state vector, respectively. Associate the abnormal state components, waiting state components, power status components, and link status components with each other to generate the state component group corresponding to the current terminal. The state component groups are arranged in the order of abnormal state component, waiting state component, power status component, and link status component to generate the terminal state vector corresponding to the current terminal.

[0014] Optionally, calculating the Whittle index for each terminal and generating a transmission priority sequence includes: Read the terminal state vector corresponding to each terminal, and use the terminal state vector corresponding to each terminal as the terminal state input in the multi-terminal dynamic scheduling model; Based on the terminal status input corresponding to each terminal, extract the abnormal status component, waiting status component, battery status component and link status component corresponding to each terminal, and determine the abnormal status component, waiting status component, battery status component and link status component as the status set of the corresponding terminal. The state sets corresponding to each terminal are aggregated to generate the multi-terminal state set corresponding to the multi-terminal dynamic scheduling model. Based on the multi-terminal state set, activation actions and maintenance actions are defined for each terminal, and the activation actions and maintenance actions are determined as the action set of each terminal in the multi-terminal dynamic scheduling model; Based on the state set and action set corresponding to each terminal, the state transition results corresponding to each terminal when performing activation action and maintenance action are generated, and the state transition results corresponding to each terminal are collected to generate the state transition relationship corresponding to the multi-terminal dynamic scheduling model. Based on the state set, action set and state transition relationship of each terminal, the first benefit value and the second benefit value of each terminal at the current sampling time are generated. The first benefit value corresponds to the activation action and the second benefit value corresponds to the hold action. Based on the first and second revenue values ​​corresponding to each terminal, the revenue difference corresponding to each terminal at the current sampling time is generated, and the revenue difference corresponding to each terminal is determined as the Whittle index corresponding to each terminal. Read the Whittle index corresponding to each terminal, and sort the terminals in descending order of Whittle index to generate an initial transmission priority sequence. When there are terminals with the same Whittle exponent in the initial transmission priority sequence, the waiting state components corresponding to the terminals with the same Whittle exponent are read, and the terminals with the same Whittle exponent are reordered in descending order of waiting state components to generate a transmission priority sequence.

[0015] Optionally, the generated gas operation status assessment results and abnormal alarm results for the corresponding monitoring points include: Read the transmission priority sequence and select the corresponding terminal in the order of the terminals in the transmission priority sequence. The data set to be uploaded corresponding to the selected terminal is determined as the current data set to be uploaded. Based on the data units arranged in the sampling time order in the current uploaded data set, extract the summary data and event data corresponding to each data unit, and encapsulate the summary data and event data corresponding to each data unit into a data frame to be sent in the sampling time order; The selected terminal is controlled to send data frames to be sent sequentially through the narrowband IoT wireless access channel. After the data frames to be sent are sent, the next set of data to be uploaded corresponding to the terminal is selected and sent according to the transmission priority sequence until the data sets to be uploaded corresponding to each terminal in the transmission priority sequence are sent. The remote monitoring platform receives the data frames to be sent from each terminal, parses the received data frames, and extracts the terminal identifier, sampling time, gas concentration data, pressure data, flow data and abnormal intensity parameters corresponding to each data frame to be sent. Based on the terminal identifier obtained by parsing, the data corresponding to the same terminal identifier is identified as the uploaded data corresponding to the same monitoring point. Based on the sampling time in the uploaded data, the gas concentration data, pressure data, flow data and abnormal intensity parameters corresponding to the same monitoring point are arranged in order to generate the monitoring data sequence of the corresponding monitoring point. Based on the gas concentration data, pressure data and flow data in the monitoring data sequence of the corresponding monitoring point, the operation status judgment result of the corresponding monitoring point is generated, and based on the abnormal intensity parameter in the monitoring data sequence, the abnormal judgment result of the corresponding monitoring point is generated. When the anomaly determination result of the corresponding monitoring point is normal, the gas operation status assessment result of the corresponding monitoring point is determined to be normal operation status; When the anomaly determination result of the corresponding monitoring point is abnormal, the gas operation status assessment result of the corresponding monitoring point is determined as an abnormal operation status, and an abnormal alarm result of the corresponding monitoring point is generated based on the anomaly intensity parameter of the corresponding monitoring point. The gas operation status assessment results and abnormal alarm results of the corresponding monitoring points will be output accordingly.

[0016] The beneficial effects of this invention are: This invention introduces an online detection method based on abrupt statistical state quantities at the narrowband IoT terminal side, enabling dynamic analysis of monitoring data such as gas concentration, pressure, and flow rate. This allows the system to continuously identify and determine abnormal states, thereby improving its responsiveness to changes in monitoring data. Simultaneously, by establishing a correlation between the anomaly determination results and the communication status, a data feature-driven mechanism for generating a set of data to be uploaded is constructed, achieving integrated processing of data acquisition and transmission. This allows different types of data to be organized and managed according to their corresponding states.

[0017] This invention constructs a multi-terminal dynamic scheduling model and calculates the Whittle index of each terminal based on the terminal state vector in the model, thereby realizing unified scheduling control of the data upload order of multiple terminals. This ensures that the upload behavior of each terminal is consistent with its state characteristics, thus improving the consistency and orderliness of data scheduling under multi-terminal concurrency conditions. At the same time, by introducing state transition relationships and reward calculation mechanisms, the scheduling process can be dynamically adjusted in combination with changes in terminal state, improving the system's adaptability in complex network environments.

[0018] This invention applies a transmission priority sequence to the data transmission control of a narrowband IoT wireless access channel, and combines it with a remote monitoring platform to parse and reconstruct the uploaded data. This enables a complete processing flow from terminal data generation to platform-side status assessment and anomaly alarm, allowing the system to uniformly assess and output the operating status of each monitoring point. This improves the data processing continuity, monitoring consistency, and overall operating efficiency of the urban gas monitoring system. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a city gas monitoring method based on narrowband Internet of Things proposed in this invention; Figure 2 This is a schematic diagram of the Page-Hinkley online detection process based on mutation statistical state quantities in this invention; Figure 3 This is a schematic diagram of the multi-terminal dynamic scheduling model and Whittle index calculation process in this invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0021] refer to Figures 1-3 A method for monitoring urban gas supply based on narrowband Internet of Things includes the following steps: Narrowband IoT terminals deployed at various monitoring points in the urban gas pipeline network collect gas concentration, pressure, and flow data, and obtain corresponding terminal remaining power information and wireless link status information to construct the original time-series data sequence; The original time-series data sequence is subjected to invalid value removal and filtering to obtain a continuous monitoring sequence; The statistical mean and cumulative deviation are recursively calculated on the terminal side based on the continuous monitoring sequence, and the historical minimum value is determined based on the cumulative deviation to generate the statistical state of abrupt change. The mutation statistics state quantity is subjected to Page-Hinkley online detection, the offset between the cumulative deviation and the historical minimum value is calculated, and the anomaly judgment result and anomaly intensity parameter are generated based on the offset. The terminal communication status is determined based on the anomaly judgment result and anomaly intensity parameter, and a corresponding set of data to be uploaded is generated based on the terminal communication status. Based on the set of data to be uploaded, a corresponding data waiting time is generated, and the abnormal intensity parameter, data waiting time, terminal remaining power information and wireless link status information are constructed into a terminal status vector. A multi-terminal dynamic scheduling model is established based on the terminal state vector, and the Whittle index of each terminal is calculated in the multi-terminal dynamic scheduling model. A transmission priority sequence is generated based on the Whittle index. Based on the transmission priority sequence, each terminal is controlled to upload data from the data set to be uploaded through the narrowband IoT wireless access channel. After receiving the uploaded data, the remote monitoring platform generates the gas operation status assessment result and abnormal alarm result for the corresponding monitoring point.

[0022] In this embodiment, a narrowband IoT terminal collects gas concentration, pressure, and flow data, and obtains corresponding terminal remaining battery power information and wireless link status information, including: At each sampling time, the narrowband IoT terminal acquires the corresponding gas concentration data through the gas concentration sensor. The gas concentration data is the measured value of the concentration of gas components per unit volume. The narrowband IoT terminal acquires the corresponding pressure data through the pressure sensor at the sampling time corresponding to the gas concentration data. The pressure data is the instantaneous pressure measurement value of the gas in the pipeline. The narrowband IoT terminal acquires the corresponding flow data through the flow sensor at the sampling time corresponding to the gas concentration data and pressure data. The flow data is the measured value of the gas volume or mass flow rate through the pipe cross section per unit time. The gas concentration data, pressure data, and flow data obtained at the same sampling time are correlated to form the monitoring data item at the sampling time; The remaining power information of the terminal at the corresponding sampling time is read based on the monitoring data items. The remaining power information of the terminal is the remaining power percentage or remaining capacity value output by the terminal power module. The wireless link status information at the corresponding sampling time is obtained based on the monitoring data item, and the wireless link status information is used as the link status data corresponding to the monitoring data item. The monitoring data items, terminal power data, and link status data are combined to obtain the original time-series data sequence.

[0023] In this embodiment, the invalid value removal and filtering process for the original time-series data sequence includes: According to the sampling time order in the original time series data sequence, the monitoring data items corresponding to each sampling time are extracted sequentially, and the gas concentration data, pressure data and flow data in the monitoring data items are read respectively; The gas concentration data, pressure data, and flow data are compared with their respective preset effective ranges. Data that exceeds the preset effective range is identified as invalid values, and monitoring data items containing invalid values ​​are deleted. Specifically, the gas concentration data, pressure data, and flow data are compared with their corresponding preset effective range intervals. This includes: for each type of sensor, the corresponding upper and lower limits of the measurement range are pre-stored on the terminal side. After acquiring the gas concentration data, pressure data, and flow data at each sampling time, each data is compared with the corresponding lower and upper limits of the range one by one. When any data is less than its corresponding lower limit or greater than its corresponding upper limit, the data is determined to be invalid. After deleting monitoring data items containing invalid values, data sequences were constructed for the gas concentration data, pressure data, and flow data in the remaining monitoring data items according to the sampling time order. Each data sequence is processed using a sliding window filtering method. At each sampling time, the data at the sampling time and the data at the sampling times before and after it are selected to form a filtering window. The data within the filtering window are averaged to obtain the filtering result at the corresponding sampling time. Specifically, each data sequence is processed using a sliding window filtering method, including: for the gas concentration data sequence, pressure data sequence, and flow data sequence arranged in the order of sampling time, the data corresponding to the current sampling time and its previous and next sampling times are selected as the data in the filtering window. The data in the filtering window are summed and divided by the number of data in the filtering window to obtain the filtering result corresponding to the current sampling time. The filtering result is then used as the updated data for that sampling time. The filtered results corresponding to the gas concentration data, the pressure data, and the flow data are correlated according to the original sampling time to obtain the filtered monitoring data items. The filtered monitoring data items are arranged in order of sampling time to generate a continuous monitoring sequence.

[0024] In this embodiment, the generation of mutation statistical state quantities includes: According to the sampling time sequence in the continuous monitoring sequence, the monitoring data items corresponding to each sampling time are extracted sequentially, and the gas concentration data, pressure data and flow data in the monitoring data items corresponding to each sampling time are used as the current sampling data. Based on the current sampling data, recursive calculations are performed on the narrowband IoT terminal side to obtain the average concentration, average pressure, and average flow rate corresponding to the current sampling time. The average concentration, average pressure, and average flow rate are then used as the average statistical value at the current sampling time. Specifically, based on the current sampled data, recursive calculations are performed on the narrowband IoT terminal side, including: for the current sampled data obtained in the order of sampling time, storing the statistical mean corresponding to the previous sampling time on the terminal side, calculating the difference between the current sampled data and the statistical mean of the previous sampling time at the current sampling time, weighting the difference according to a preset update coefficient and adding it to the statistical mean of the previous sampling time to obtain the updated statistical mean corresponding to the current sampling time, and using the updated statistical mean as the input for the recursive calculation at the next sampling time; Based on the difference between the current sampled data and the corresponding statistical mean, the cumulative deviations of concentration, pressure, and flow rate corresponding to the current sampling time are recursively accumulated to obtain the cumulative deviations of concentration, pressure, and flow rate at the current sampling time. The cumulative deviations of concentration, pressure, and flow rate are then used as the cumulative deviations at the current sampling time. Specifically, the difference between the current sampled data and the corresponding statistical mean is used to perform recursive accumulation, including: obtaining the difference between the current sampled data and the corresponding statistical mean at the current sampling time, reading the accumulated deviation stored at the previous sampling time on the terminal side, directly adding the difference to the accumulated deviation at the previous sampling time to obtain the updated accumulated deviation corresponding to the current sampling time, and using the updated accumulated deviation as the input for the recursive calculation at the next sampling time; The cumulative deviation at the current sampling time is compared with the historical minimum value determined at the previous sampling time. When the cumulative deviation at the current sampling time is less than the historical minimum value determined at the previous sampling time, the cumulative deviation at the current sampling time is determined as the historical minimum value at the current sampling time. When the cumulative deviation at the current sampling time is greater than or equal to the historical minimum value determined at the previous sampling time, the historical minimum value determined at the previous sampling time is kept as the historical minimum value at the current sampling time. By correlating the statistical mean, cumulative deviation, and historical minimum value at the current sampling time, a sudden change statistical state quantity at the current sampling time is generated.

[0025] In this embodiment, Page-Hinkley online detection includes: Read the mutation statistics state quantities corresponding to each sampling time in sequence according to the sampling time order, and extract the statistical mean, cumulative deviation and historical minimum value of the corresponding sampling time from the mutation statistics state quantities. A stable monitoring segment is constructed based on the monitoring data items in the continuous monitoring sequence arranged in order of sampling time prior to the current sampling time, and a stable reference state is generated based on the gas concentration data, pressure data, and flow data in the stable monitoring segment; The gas concentration data, pressure data, and flow data in the monitoring data items at the current sampling time are compared with the stable reference state to generate the residual detection value corresponding to the current sampling time. Specifically, the gas concentration data, pressure data, and flow rate data in the monitoring data items at the current sampling time are compared with the stable reference state, including: for the gas concentration data, pressure data, and flow rate data obtained in the monitoring data items at the current sampling time, the corresponding concentration reference value, pressure reference value, and flow rate reference value in the stable reference state are read respectively, the difference between the gas concentration data and the concentration reference value is calculated, the difference between the pressure data and the pressure reference value is calculated, the difference between the flow rate data and the flow rate reference value is calculated, and the difference result is used as the corresponding comparison result; The residual detection value is used as the Page-Hinkley online detection input value corresponding to the current sampling time, and the detection offset corresponding to the current sampling time is generated based on the input value and the statistical mean, cumulative deviation and historical minimum value corresponding to the current sampling time. The detection offset is compared with a preset detection threshold. When the detection offset is less than the preset detection threshold, a normal judgment result for the current sampling time is generated. When the detection offset is greater than or equal to the preset detection threshold, the current sampling time is determined as the candidate anomaly start time, and the historical minimum value corresponding to the candidate anomaly start time is latched. At the same time, an anomaly segment cache is established, and the detection offset corresponding to the candidate anomaly start time is written into the anomaly segment cache. After establishing the abnormal segment cache, continue to read the mutation statistical state quantity corresponding to the sampling time in the order of sampling time, and extract the statistical mean, cumulative deviation and historical minimum value from the mutation statistical state quantity. At the same time, generate the residual detection value corresponding to the sampling time based on the stable reference state and the monitoring data item corresponding to the sampling time. The residual detection value corresponding to the sampling time is used as the Page-Hinkley online detection input value for the corresponding sampling time. Based on the input value, the corresponding statistical mean, the corresponding cumulative deviation, and the latched historical minimum value, the detection offset corresponding to the sampling time is generated. The detection offsets that are greater than or equal to the preset detection threshold are written into the abnormal segment cache in the order of sampling time. Continuous abnormal segments are generated based on the detection offsets written in the abnormal segment cache in the order of sampling time, and the continuous abnormal segments are determined as the abnormal judgment results. The maximum detection offset in the continuous abnormal segments is determined as the abnormal intensity parameter. When the detection offset in the abnormal segment cache does not generate a continuous abnormal segment, the abnormal segment cache is cleared and the historical minimum value is released from latching. After generating the anomaly determination result, the stable monitoring segment is reconstructed based on the monitoring data items in the continuous monitoring sequence arranged in the order of sampling time after the anomaly determination result, and the stable reference state corresponding to the next sampling time is generated based on the reconstructed stable monitoring segment.

[0026] In this embodiment, generating the corresponding set of data to be uploaded includes: Read the anomaly judgment result and anomaly intensity parameter corresponding to the current sampling time, and determine the terminal communication status corresponding to the current sampling time based on the anomaly judgment result; Specifically, determining the terminal communication status corresponding to the current sampling time based on the anomaly determination result includes: reading the anomaly determination result corresponding to the current sampling time and matching the anomaly determination result with the pre-set communication status mapping rules. When the anomaly determination result is normal, the terminal communication status at the current sampling time is determined to be the summary reporting status. When the anomaly determination result is abnormal, the terminal communication status at the current sampling time is determined to be the event reporting status. When the anomaly determination result corresponding to the current sampling time is normal, the terminal communication status corresponding to the current sampling time is determined as the summary reporting status, and the summary data corresponding to the current sampling time is generated based on the monitoring data item corresponding to the current sampling time. When the anomaly determination result corresponding to the current sampling time is an anomaly, the terminal communication status corresponding to the current sampling time is determined as the event reporting status, and the event data corresponding to the current sampling time is generated based on the monitoring data item corresponding to the current sampling time and the anomaly intensity parameter corresponding to the current sampling time. Associate the terminal communication status at the current sampling time with the summary data and event data at the current sampling time to generate the data unit at the current sampling time; Specifically, the terminal communication state corresponding to the current sampling time is associated with the summary data and event data corresponding to the current sampling time. This includes: after determining the terminal communication state corresponding to the current sampling time, reading the data content generated at the current sampling time; when the terminal communication state corresponding to the current sampling time is the summary reporting state, binding the summary data generated at the current sampling time with the summary reporting state; when the terminal communication state corresponding to the current sampling time is the event reporting state, binding the event data generated at the current sampling time with the event reporting state; and using the bound terminal communication state and the corresponding data content as the data unit corresponding to the current sampling time. Read the data units corresponding to each sampling time sequentially according to the sampling time order, and merge multiple consecutive data units with the same terminal communication status to generate a data segment corresponding to the terminal communication status. The summary data and event data corresponding to each data unit in the data segment are arranged in the order of sampling time to generate a set of data to be uploaded corresponding to the terminal communication status.

[0027] In this embodiment, constructing the terminal state vector includes: Read the data set to be uploaded corresponding to the current terminal, and extract the data units arranged in order of sampling time from the data set to be uploaded; Based on the data units arranged in order of sampling time in the dataset to be uploaded, extract the data unit corresponding to the earliest sampling time, and determine the sampling time of the data unit corresponding to the earliest sampling time as the starting time of the dataset to be uploaded. The data waiting time for the current terminal is generated based on the time difference between the current time and the start time. Read the abnormal intensity parameter, remaining battery power information and wireless link status information corresponding to the current terminal, and determine the abnormal intensity parameter, data waiting time, remaining battery power information and wireless link status information as the abnormal state component, waiting state component, battery status component and link status component corresponding to the terminal state vector, respectively. Associate the abnormal state components, waiting state components, power status components, and link status components with each other to generate the state component group corresponding to the current terminal. Specifically, the abnormal state components, waiting state components, power status components, and link status components are associated accordingly. This includes: at the same time corresponding to the same terminal, the abnormal state components corresponding to the abnormal intensity parameter, the waiting state components corresponding to the data waiting time, the power status components corresponding to the terminal's remaining power information, and the link status components corresponding to the wireless link status information are uniformly identified and combined according to the same terminal identifier and the same time identifier, so that each status component corresponds in the same data structure, thereby forming an association result that can reflect the comprehensive status of the current terminal at that time. The state component groups are arranged in the order of abnormal state component, waiting state component, power status component, and link status component to generate the terminal state vector corresponding to the current terminal.

[0028] In this embodiment, calculating the Whittle index of each terminal and generating a transmission priority sequence includes: Read the terminal state vector corresponding to each terminal, and use the terminal state vector corresponding to each terminal as the terminal state input in the multi-terminal dynamic scheduling model; Based on the terminal status input corresponding to each terminal, extract the abnormal status component, waiting status component, battery status component and link status component corresponding to each terminal, and determine the abnormal status component, waiting status component, battery status component and link status component as the status set of the corresponding terminal. The state sets corresponding to each terminal are aggregated to generate the multi-terminal state set corresponding to the multi-terminal dynamic scheduling model. Based on the multi-terminal state set, activation actions and maintenance actions are defined for each terminal, and the activation actions and maintenance actions are determined as the action set of each terminal in the multi-terminal dynamic scheduling model; Specifically, activation actions and hold actions are defined, including: based on the terminal state vector corresponding to each terminal, two types of scheduling actions are set for each terminal in the multi-terminal dynamic scheduling model. When the activation action is selected, the corresponding terminal is controlled to perform a data upload operation on its data set to be uploaded, and the waiting state component corresponding to the terminal is updated after the data upload operation is performed; when the hold action is selected, the corresponding terminal is controlled not to perform a data upload operation at the current sampling time, and its data set to be uploaded remains unchanged, while the waiting state component corresponding to the terminal is updated at subsequent sampling times. Based on the state set and action set corresponding to each terminal, the state transition results corresponding to each terminal when performing activation action and maintenance action are generated, and the state transition results corresponding to each terminal are collected to generate the state transition relationship corresponding to the multi-terminal dynamic scheduling model. Specifically, based on the state set and action set corresponding to each terminal, the state transition results corresponding to each terminal when performing activation and hold actions are generated. This includes: for the state set corresponding to the current sampling time of each terminal, when selecting the activation action, the waiting state component is updated to the initial value based on the terminal state vector corresponding to that terminal, and the updated abnormal state component, power state component, and link state component are generated based on the abnormal intensity parameter, terminal remaining power information, and wireless link state information corresponding to the current sampling time, thus forming the state transition result after performing the activation action; when selecting the hold action, the abnormal state component, power state component, and link state component corresponding to that terminal are kept unchanged, and the waiting state component is incrementally updated based on the data waiting time corresponding to the current sampling time, thereby forming the state transition result after performing the hold action. Based on the state set, action set and state transition relationship of each terminal, the first benefit value and the second benefit value of each terminal at the current sampling time are generated. The first benefit value corresponds to the activation action and the second benefit value corresponds to the hold action. Based on the first and second revenue values ​​corresponding to each terminal, the revenue difference corresponding to each terminal at the current sampling time is generated, and the revenue difference corresponding to each terminal is determined as the Whittle index corresponding to each terminal. Specifically, based on the first and second revenue values ​​corresponding to each terminal, a revenue difference value corresponding to each terminal at the current sampling time is generated, and the revenue difference value corresponding to each terminal is determined as the Whittle index corresponding to each terminal. This includes: for each terminal, at the current sampling time, reading the first revenue value corresponding to the terminal when performing the activation action and the second revenue value corresponding to the terminal when performing the hold action, calculating the difference between the first revenue value and the second revenue value to obtain the revenue difference value corresponding to the terminal at the current sampling time, and using the revenue difference value as the Whittle index corresponding to the terminal at the current sampling time. Read the Whittle index corresponding to each terminal, and sort the terminals in descending order of Whittle index to generate an initial transmission priority sequence. When there are terminals with the same Whittle exponent in the initial transmission priority sequence, read the waiting state component corresponding to the terminal with the same Whittle exponent, and reorder the terminals with the same Whittle exponent according to the order of waiting state components from largest to smallest to generate a transmission priority sequence. Specifically, when there are terminals with the same Whittle index in the initial transmission priority sequence, the waiting state components corresponding to the terminals with the same Whittle index are read, and the terminals with the same Whittle index are reordered in descending order of waiting state components to generate the transmission priority sequence. This includes: after sorting the Whittle indexes of each terminal, traversing the adjacent terminals in the initial transmission priority sequence. When multiple terminals are detected to have the same Whittle index value, the terminals with the same Whittle index are grouped, and the waiting state components of the corresponding terminals in each group are read. The waiting state components of each terminal in the same group are compared, and the terminals in the group are rearranged in descending order of waiting state component value. The rearranged terminal order replaces the corresponding position in the original initial transmission priority sequence to generate the final transmission priority sequence.

[0029] In this embodiment, the generation of gas operation status assessment results and abnormal alarm results for the corresponding monitoring points includes: Read the transmission priority sequence and select the corresponding terminal in the order of the terminals in the transmission priority sequence. The data set to be uploaded corresponding to the selected terminal is determined as the current data set to be uploaded. Based on the data units arranged in the sampling time order in the current uploaded data set, extract the summary data and event data corresponding to each data unit, and encapsulate the summary data and event data corresponding to each data unit into a data frame to be sent in the sampling time order; The selected terminal is controlled to send data frames to be sent sequentially through the narrowband IoT wireless access channel. After the data frames to be sent are sent, the next set of data to be uploaded corresponding to the terminal is selected and sent according to the transmission priority sequence until the data sets to be uploaded corresponding to each terminal in the transmission priority sequence are sent. The remote monitoring platform receives the data frames to be sent from each terminal, parses the received data frames, and extracts the terminal identifier, sampling time, gas concentration data, pressure data, flow data and abnormal intensity parameters corresponding to each data frame to be sent. Based on the terminal identifier obtained by parsing, the data corresponding to the same terminal identifier is identified as the uploaded data corresponding to the same monitoring point. Based on the sampling time in the uploaded data, the gas concentration data, pressure data, flow data and abnormal intensity parameters corresponding to the same monitoring point are arranged in order to generate the monitoring data sequence of the corresponding monitoring point. Based on the gas concentration data, pressure data and flow data in the monitoring data sequence of the corresponding monitoring point, the operation status judgment result of the corresponding monitoring point is generated, and based on the abnormal intensity parameter in the monitoring data sequence, the abnormal judgment result of the corresponding monitoring point is generated. When the anomaly determination result of the corresponding monitoring point is normal, the gas operation status assessment result of the corresponding monitoring point is determined to be normal operation status; When the anomaly determination result of the corresponding monitoring point is abnormal, the gas operation status assessment result of the corresponding monitoring point is determined as an abnormal operation status, and an abnormal alarm result of the corresponding monitoring point is generated based on the anomaly intensity parameter of the corresponding monitoring point. The gas operation status assessment results and abnormal alarm results of the corresponding monitoring points will be output accordingly.

[0030] Example 1: To verify the feasibility of this invention in practice, it was applied to the operation of a city's gas pipeline network. Due to the wide coverage, complex operating environment, and large fluctuations in gas load, problems such as gas leaks, abnormal pressure fluctuations, and sudden changes in flow are prone to occur. Existing systems typically use periodic sampling and timed uploading for monitoring. When there are many monitoring terminals in the network, a large amount of normal data consumes transmission resources, causing delays or even loss of abnormal data during the uploading process. Furthermore, fixed threshold-based anomaly detection methods are insufficient to promptly identify early, slowly changing anomalies, thus affecting the safety monitoring capabilities of the gas system. This embodiment uses a city's medium-pressure gas transmission and distribution network as an application scenario to verify the effectiveness of the method of this invention in a real-world operating environment.

[0031] In this scenario, multiple narrowband IoT terminals are deployed along the pipeline network. Each terminal is equipped with a gas concentration sensor, a pressure sensor, and a flow sensor, and operates continuously via battery power. Each terminal collects gas concentration, pressure, and flow data at a fixed sampling period, while simultaneously acquiring remaining battery power and wireless link status information to construct the original time-series data sequence. Taking a specific monitoring point as an example, the gas concentration data collected during continuous operation ranges from 0.5% to 2.1%, the pressure data ranges from 0.18 MPa to 0.26 MPa, and the flow data ranges from 120 m³ / h. 3 / h to 310m 3 / h, after removing invalid values ​​and applying sliding window filtering to the raw data, a continuous monitoring sequence was obtained. The data fluctuation amplitude was significantly reduced after filtering, with concentration fluctuations stabilizing between 0.8% and 1.6%, pressure fluctuations stabilizing between 0.20 MPa and 0.24 MPa, and flow rate fluctuations stabilizing within 150 m³ / h. 3 / h to 280m 3 / h.

[0032] Based on continuous monitoring sequences, the terminal side performs statistical mean recursive calculation on each sampled data, and further calculates the cumulative deviation and historical minimum value to generate abrupt change statistical state variables. In the initial stage of operation, the statistical mean changes steadily, and the cumulative deviation fluctuates close to zero. When a minor leak occurs during system operation, the gas concentration shows a slow upward trend, for example, gradually increasing from 1.2% to 1.6%. Traditional fixed threshold methods cannot identify anomalies before exceeding the alarm threshold. However, this method gradually accumulates the cumulative deviation, causing the detection offset to reach the set threshold after approximately 40 sampling cycles, thereby triggering anomaly detection. Actual test results show that this method can identify anomalies before the concentration reaches the traditional alarm threshold of 2.0%, issuing an anomaly signal approximately 15 to 20 minutes earlier than traditional methods.

[0033] After an anomaly detection, the terminal determines the communication status based on the anomaly detection result. When in a normal state, the terminal enters summary reporting mode, uploading only key statistical data, with an average data upload rate of approximately 1.2KB per minute. When an anomaly is detected, the terminal switches to event reporting mode, uploading complete monitoring data and anomaly intensity parameters, increasing the data upload rate to approximately 6.8KB per minute. In actual operation, the overall average data upload volume of the system is reduced by approximately 62% compared to the traditional periodic full upload method, significantly reducing wireless link occupancy.

[0034] In a multi-terminal concurrent scenario, this embodiment deploys 320 terminals, with multiple terminals simultaneously generating data sets to be uploaded within the same time period. A terminal state vector is constructed, combining anomaly intensity parameters, data waiting time, remaining battery power, and wireless link status information, and inputting this data into a multi-terminal dynamic scheduling model. During scheduling, a Whittle index is calculated for each terminal. For example, the Whittle index of an abnormal terminal is calculated to be 3.8, while the Whittle indices of normal terminals are mostly between 0.5 and 1.2. After sorting by index, the abnormal terminal receives priority for uploading. In actual testing, the average upload latency for abnormal terminals is 2.3 seconds, while the upload latency under the traditional polling method is approximately 8.7 seconds, reducing the transmission latency of critical abnormal data by approximately 73%.

[0035] Even when the wireless link is unstable, such as when the link signal strength drops to -110dBm, the system can still adjust the scheduling order based on the link state components in the terminal state vector, prioritizing abnormal terminals with better link quality to send data, thereby improving the success rate of uploading critical data. Test results show that in a fluctuating link environment, the success rate of abnormal data upload is increased from 85% with traditional methods to over 96%.

[0036] After receiving data uploaded from various terminals, the remote monitoring platform parses the data frames and reconstructs the monitoring data sequence. It associates the data with terminal identifiers and sampling times to form a continuous data stream for each monitoring point. During data analysis, the platform generates operational status judgment results based on gas concentration, pressure, and flow data, and generates anomaly alarms in conjunction with anomaly intensity parameters. When the anomaly intensity parameter at a monitoring point reaches a set threshold, the platform immediately generates an alarm message and triggers subsequent safety procedures. In actual operation, the system detected a total of 12 abnormal events, including 7 slow leaks and 5 instantaneous pressure fluctuations. All anomalies were accurately identified without any false alarms.

[0037] In terms of power consumption, since the terminal adopts a low-frequency summary upload mode under normal conditions, the average current of the terminal is reduced from 18mA in the traditional solution to 7mA, and the battery life is extended from about 18 months to about 42 months, which significantly improves the long-term operation capability of the system.

[0038] As can be seen from the above embodiments, the present invention effectively solves the problems of insensitive anomaly detection, unstable data transmission, and high terminal power consumption in the prior art in practical applications. It realizes integrated processing of anomaly detection, data organization, communication scheduling, and remote monitoring, and can operate stably in multi-terminal, large-scale deployment environments, thereby improving the overall performance of the gas monitoring system.

[0039] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for monitoring urban gas supply based on narrowband Internet of Things, characterized in that, include: Narrowband IoT terminals deployed at various monitoring points in the urban gas pipeline network collect gas concentration, pressure, and flow data, and obtain corresponding terminal remaining power information and wireless link status information to construct the original time-series data sequence; The original time-series data sequence is subjected to invalid value removal and filtering to obtain a continuous monitoring sequence; The statistical mean and cumulative deviation are recursively calculated on the terminal side based on the continuous monitoring sequence, and the historical minimum value is determined based on the cumulative deviation to generate the statistical state of abrupt change. The mutation statistics state quantity is subjected to Page-Hinkley online detection, the offset between the cumulative deviation and the historical minimum value is calculated, and the anomaly judgment result and anomaly intensity parameter are generated based on the offset. The terminal communication status is determined based on the anomaly judgment result and anomaly intensity parameter, and a corresponding set of data to be uploaded is generated based on the terminal communication status. Based on the set of data to be uploaded, a corresponding data waiting time is generated, and the abnormal intensity parameter, data waiting time, terminal remaining power information and wireless link status information are constructed into a terminal status vector. A multi-terminal dynamic scheduling model is established based on the terminal state vector, and the Whittle index of each terminal is calculated in the multi-terminal dynamic scheduling model. A transmission priority sequence is generated based on the Whittle index. Based on the transmission priority sequence, each terminal is controlled to upload data from the data set to be uploaded through the narrowband IoT wireless access channel. After receiving the uploaded data, the remote monitoring platform generates the gas operation status assessment result and abnormal alarm result for the corresponding monitoring point. 2.The method of claim 1, wherein, The narrowband IoT terminal collects gas concentration, pressure, and flow data, and obtains corresponding information such as remaining terminal battery power and wireless link status, including: At each sampling time, the narrowband IoT terminal acquires the corresponding gas concentration data through the gas concentration sensor. The gas concentration data is the measured value of the concentration of gas components per unit volume. The narrowband IoT terminal acquires the corresponding pressure data through the pressure sensor at the sampling time corresponding to the gas concentration data. The pressure data is the instantaneous pressure measurement value of the gas in the pipeline. The narrowband IoT terminal acquires the corresponding flow data through the flow sensor at the sampling time corresponding to the gas concentration data and pressure data. The flow data is the measured value of the gas volume or mass flow rate through the pipe cross section per unit time. The gas concentration data, pressure data, and flow data obtained at the same sampling time are correlated to form the monitoring data item at the sampling time; The remaining power information of the terminal at the corresponding sampling time is read based on the monitoring data items. The remaining power information of the terminal is the remaining power percentage or remaining capacity value output by the terminal power module. The wireless link status information at the corresponding sampling time is obtained based on the monitoring data item, and the wireless link status information is used as the link status data corresponding to the monitoring data item. The monitoring data items, terminal power data, and link status data are combined to obtain the original time-series data sequence. 3.The method of claim 1, wherein, The invalid value removal and filtering process for the original time series data sequence includes: According to the sampling time order in the original time series data sequence, the monitoring data items corresponding to each sampling time are extracted sequentially, and the gas concentration data, pressure data and flow data in the monitoring data items are read respectively; The gas concentration data, pressure data, and flow data are compared with their respective preset effective ranges. Data that exceeds the preset effective range is identified as invalid values, and monitoring data items containing invalid values ​​are deleted. After deleting monitoring data items containing invalid values, data sequences were constructed for the gas concentration data, pressure data, and flow data in the remaining monitoring data items according to the sampling time order. Each data sequence is processed using a sliding window filtering method. At each sampling time, the data at the sampling time and the data at the sampling times before and after it are selected to form a filtering window. The data within the filtering window are averaged to obtain the filtering result at the corresponding sampling time. The filtered results corresponding to the gas concentration data, the pressure data, and the flow data are correlated according to the original sampling time to obtain the filtered monitoring data items. The filtered monitoring data items are arranged in order of sampling time to generate a continuous monitoring sequence. 4.The method of claim 1, wherein, The generation of mutation statistical state variables includes: According to the sampling time sequence in the continuous monitoring sequence, the monitoring data items corresponding to each sampling time are extracted sequentially, and the gas concentration data, pressure data and flow data in the monitoring data items corresponding to each sampling time are used as the current sampling data. Based on the current sampling data, recursive calculations are performed on the narrowband IoT terminal side to obtain the average concentration, average pressure, and average flow rate corresponding to the current sampling time. The average concentration, average pressure, and average flow rate are then used as the average statistical value at the current sampling time. Based on the difference between the current sampled data and the corresponding statistical mean, the cumulative deviations of concentration, pressure, and flow rate corresponding to the current sampling time are recursively accumulated to obtain the cumulative deviations of concentration, pressure, and flow rate at the current sampling time. The cumulative deviations of concentration, pressure, and flow rate are then used as the cumulative deviations at the current sampling time. The cumulative deviation at the current sampling time is compared with the historical minimum value determined at the previous sampling time. When the cumulative deviation at the current sampling time is less than the historical minimum value determined at the previous sampling time, the cumulative deviation at the current sampling time is determined as the historical minimum value at the current sampling time. When the cumulative deviation at the current sampling time is greater than or equal to the historical minimum value determined at the previous sampling time, the historical minimum value determined at the previous sampling time is kept as the historical minimum value at the current sampling time. By correlating the statistical mean, cumulative deviation, and historical minimum value at the current sampling time, a sudden change statistical state quantity at the current sampling time is generated. 5.The urban gas monitoring method based on the narrowband Internet of Things according to claim 1, wherein, Page-Hinkley online detection includes: Read the mutation statistics state quantities corresponding to each sampling time in sequence according to the sampling time order, and extract the statistical mean, cumulative deviation and historical minimum value of the corresponding sampling time from the mutation statistics state quantities. A stable monitoring segment is constructed based on the monitoring data items in the continuous monitoring sequence arranged in order of sampling time prior to the current sampling time, and a stable reference state is generated based on the gas concentration data, pressure data, and flow data in the stable monitoring segment; The gas concentration data, pressure data, and flow data in the monitoring data items at the current sampling time are compared with the stable reference state to generate the residual detection value corresponding to the current sampling time. The residual detection value is used as the Page-Hinkley online detection input value corresponding to the current sampling time, and the detection offset corresponding to the current sampling time is generated based on the input value and the statistical mean, cumulative deviation and historical minimum value corresponding to the current sampling time. The detection offset is compared with a preset detection threshold. When the detection offset is less than the preset detection threshold, a normal judgment result for the current sampling time is generated. When the detection offset is greater than or equal to the preset detection threshold, the current sampling time is determined as the candidate anomaly start time, and the historical minimum value corresponding to the candidate anomaly start time is latched. At the same time, an anomaly segment cache is established, and the detection offset corresponding to the candidate anomaly start time is written into the anomaly segment cache. After establishing the abnormal segment cache, continue to read the mutation statistical state quantity corresponding to the sampling time in the order of sampling time, and extract the statistical mean, cumulative deviation and historical minimum value from the mutation statistical state quantity. At the same time, generate the residual detection value corresponding to the sampling time based on the stable reference state and the monitoring data item corresponding to the sampling time. The residual detection value corresponding to the sampling time is used as the Page-Hinkley online detection input value for the corresponding sampling time. Based on the input value, the corresponding statistical mean, the corresponding cumulative deviation, and the latched historical minimum value, the detection offset corresponding to the sampling time is generated. The detection offsets that are greater than or equal to the preset detection threshold are written into the abnormal segment cache in the order of sampling time. Continuous abnormal segments are generated based on the detection offsets written in the abnormal segment cache in the order of sampling time, and the continuous abnormal segments are determined as the abnormal judgment results. The maximum detection offset in the continuous abnormal segments is determined as the abnormal intensity parameter. When the detection offset in the abnormal segment cache does not generate a continuous abnormal segment, the abnormal segment cache is cleared and the historical minimum value is released from latching. After generating the anomaly determination result, the stable monitoring segment is reconstructed based on the monitoring data items in the continuous monitoring sequence arranged in the order of sampling time after the anomaly determination result, and the stable reference state corresponding to the next sampling time is generated based on the reconstructed stable monitoring segment. 6.The method of claim 1, wherein, The corresponding collection of data to be uploaded includes: Read the anomaly judgment result and anomaly intensity parameter corresponding to the current sampling time, and determine the terminal communication status corresponding to the current sampling time based on the anomaly judgment result; When the anomaly determination result corresponding to the current sampling time is normal, the terminal communication status corresponding to the current sampling time is determined as the summary reporting status, and the summary data corresponding to the current sampling time is generated based on the monitoring data item corresponding to the current sampling time. When the anomaly determination result corresponding to the current sampling time is an anomaly, the terminal communication status corresponding to the current sampling time is determined as the event reporting status, and the event data corresponding to the current sampling time is generated based on the monitoring data item corresponding to the current sampling time and the anomaly intensity parameter corresponding to the current sampling time. Associate the terminal communication status at the current sampling time with the summary data and event data at the current sampling time to generate the data unit at the current sampling time; Read the data units corresponding to each sampling time sequentially according to the sampling time order, and merge multiple consecutive data units with the same terminal communication status to generate a data segment corresponding to the terminal communication status. The summary data and event data corresponding to each data unit in the data segment are arranged in the order of sampling time to generate a set of data to be uploaded corresponding to the terminal communication status. 7.The method of claim 1, wherein the method further comprises: Constructing the terminal state vector includes: Read the data set to be uploaded corresponding to the current terminal, and extract the data units arranged in order of sampling time from the data set to be uploaded; Based on the data units arranged in order of sampling time in the dataset to be uploaded, extract the data unit corresponding to the earliest sampling time, and determine the sampling time of the data unit corresponding to the earliest sampling time as the starting time of the dataset to be uploaded. The data waiting time for the current terminal is generated based on the time difference between the current time and the start time. Read the abnormal intensity parameter, remaining battery power information and wireless link status information corresponding to the current terminal, and determine the abnormal intensity parameter, data waiting time, remaining battery power information and wireless link status information as the abnormal state component, waiting state component, battery status component and link status component corresponding to the terminal state vector, respectively. Associate the abnormal state components, waiting state components, power status components, and link status components with each other to generate the state component group corresponding to the current terminal. The state component groups are arranged in the order of abnormal state component, waiting state component, power status component, and link status component to generate the terminal state vector corresponding to the current terminal. 8.The method of claim 1, wherein, Calculating the Whittle index for each terminal and generating a transmission priority sequence includes: Read the terminal state vector corresponding to each terminal, and use the terminal state vector corresponding to each terminal as the terminal state input in the multi-terminal dynamic scheduling model; Based on the terminal status input corresponding to each terminal, extract the abnormal status component, waiting status component, battery status component and link status component corresponding to each terminal, and determine the abnormal status component, waiting status component, battery status component and link status component as the status set of the corresponding terminal. The state sets corresponding to each terminal are aggregated to generate the multi-terminal state set corresponding to the multi-terminal dynamic scheduling model. Based on the multi-terminal state set, activation actions and maintenance actions are defined for each terminal, and the activation actions and maintenance actions are determined as the action set of each terminal in the multi-terminal dynamic scheduling model; Based on the state set and action set corresponding to each terminal, the state transition results corresponding to each terminal when performing activation action and maintenance action are generated, and the state transition results corresponding to each terminal are collected to generate the state transition relationship corresponding to the multi-terminal dynamic scheduling model. Based on the state set, action set and state transition relationship of each terminal, the first and second revenue values ​​of each terminal at the current sampling time are generated. Based on the first and second revenue values ​​corresponding to each terminal, the revenue difference corresponding to each terminal at the current sampling time is generated, and the revenue difference corresponding to each terminal is determined as the Whittle index corresponding to each terminal. Read the Whittle index corresponding to each terminal, and sort the terminals in descending order of Whittle index to generate an initial transmission priority sequence. When there are terminals with the same Whittle exponent in the initial transmission priority sequence, the waiting state components corresponding to the terminals with the same Whittle exponent are read, and the terminals with the same Whittle exponent are reordered in descending order of waiting state components to generate a transmission priority sequence. 9.The urban gas monitoring method based on the narrowband Internet of Things according to claim 1, wherein, The generated gas operation status assessment results and abnormal alarm results for the corresponding monitoring points include: Read the transmission priority sequence and select the corresponding terminal in the order of the terminals in the transmission priority sequence. The data set to be uploaded corresponding to the selected terminal is determined as the current data set to be uploaded. Based on the data units arranged in the sampling time order in the current uploaded data set, extract the summary data and event data corresponding to each data unit, and encapsulate the summary data and event data corresponding to each data unit into a data frame to be sent in the sampling time order; The selected terminal is controlled to send data frames to be sent sequentially through the narrowband IoT wireless access channel. After the data frames to be sent are sent, the next set of data to be uploaded corresponding to the terminal is selected and sent according to the transmission priority sequence until the data sets to be uploaded corresponding to each terminal in the transmission priority sequence are sent. The remote monitoring platform receives the data frames to be sent from each terminal, parses the received data frames, and extracts the terminal identifier, sampling time, gas concentration data, pressure data, flow data and abnormal intensity parameters corresponding to each data frame to be sent. Based on the terminal identifier obtained by parsing, the data corresponding to the same terminal identifier is identified as the uploaded data corresponding to the same monitoring point. Based on the sampling time in the uploaded data, the gas concentration data, pressure data, flow data and abnormal intensity parameters corresponding to the same monitoring point are arranged in order to generate the monitoring data sequence of the corresponding monitoring point. Based on the gas concentration data, pressure data and flow data in the monitoring data sequence of the corresponding monitoring point, the operation status judgment result of the corresponding monitoring point is generated, and based on the abnormal intensity parameter in the monitoring data sequence, the abnormal judgment result of the corresponding monitoring point is generated. When the anomaly determination result of the corresponding monitoring point is normal, the gas operation status assessment result of the corresponding monitoring point is determined to be normal operation status; When the anomaly determination result of the corresponding monitoring point is abnormal, the gas operation status assessment result of the corresponding monitoring point is determined as an abnormal operation status, and an abnormal alarm result of the corresponding monitoring point is generated based on the anomaly intensity parameter of the corresponding monitoring point. The gas operation status assessment results and abnormal alarm results of the corresponding monitoring points will be output accordingly.

Images