Manhole cover comprehensive environment monitoring system based on Internet of Things

Abstract

The invention relates to the technical field of manhole cover monitoring, in particular to a manhole cover comprehensive environment monitoring system based on the Internet of Things, which is characterized in that a manhole cover displacement value is calculated point by point in a time window and is subjected to weighted stacking with a preorder trend curve to form a fusion sequence, and then the fusion sequence is input into a long short-term memory network to complete nonlinear prediction; an output result is compared with a threshold value interval section by section, it is ensured that abnormity judgment is achieved in a continuous fragment, through combination of time sequence prediction and nonlinear learning, unified modeling is achieved on displacement trend change and environment trend, the continuity and accuracy of abnormity judgment are improved, the displacement and the inclination angle of the well lid are synchronously judged in a trend set, and the safety of the well lid is improved. According to the method, the initial signal is triggered, the signal and the gas sudden change difference value are accumulated, features are extracted through the convolutional neural network, matching is completed, and the high-risk signal is output, so that sensitive capture of sudden abnormity is enhanced, the probability of missing report is reduced, and the response timeliness under the sudden situation is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of well lid monitoring, and in particular to a well lid comprehensive environment monitoring system based on the Internet of Things. BACKGROUND

[0002] The technical field of well lid monitoring aims to improve the safety management level of urban underground pipe networks and road facilities through informatization and automation, realize real-time monitoring, remote transmission and intelligent early warning of well lid operation status and underground environment conditions, avoid environmental risks such as underground toxic gas leakage, abnormal water level, fire, etc., which endanger public safety, and improve municipal management efficiency and emergency response capability.

[0003] The purpose of a well lid comprehensive environment monitoring system based on the Internet of Things is to continuously monitor and remotely manage the physical state of the well lid and the underground environment parameters, to ensure the safe operation of urban roads and underground pipe networks, to realize real-time detection of the opening, displacement, inclination and damage of the well lid, and accurate monitoring of a plurality of environmental parameters such as underground gas concentration, temperature and humidity, water level, etc., to timely discover abnormalities and issue warnings, to avoid traffic accidents and public safety hazards caused by missing well lids or environmental abnormalities, to reduce operation and maintenance costs through Internet of Things communication and centralized platform management, and to improve the work efficiency of municipal departments in facility supervision and emergency response.

[0004] Although the prior art can realize real-time monitoring and remote transmission of the operation status and environmental parameters of the well lid, there are still some deficiencies in actual application. The monitoring relies on single-point data and threshold triggering, the judgment in the time dimension often stays in single numerical comparison, the trend is difficult to present continuously, leading to misjudgment or omission under edge conditions. The data organization is centered on devices, and the spatial dimension lacks consideration of node connectivity and geographical distance. The risk is often left on an isolated point, which cannot reveal the chain transmission of abnormal diffusion from one place to multiple places. Cross-zone early warning relies on administrative or operation and maintenance partition management, lacks linkage judgment based on risk diffusion rules, and is prone to cause repeated dispatching or waste of disposal resources. The processing capacity on the edge side is insufficient, and sudden high-risk signals often need to be identified by a centralized platform. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and to provide a well lid comprehensive environment monitoring system based on the Internet of Things.

[0006] In order to achieve the above purpose, the present application adopts the following technical scheme: a well lid comprehensive environment monitoring system based on the Internet of Things comprises:

[0007] A well lid state monitoring module: based on a well lid water level probe, a gas sensor, a temperature and humidity element, an inclinometer and a vibration meter, the water level value and the gas value are subtracted, the temperature and humidity sequence is accumulated, the inclination value and the vibration value are integrated, and a monitoring value set is generated.

[0008] time series prediction module: based on the monitoring value set, heat and humidity sequence, deduce displacement amount, and generate prediction sequence by weighted fusion, use long short-term memory network to compare prediction sequence with threshold value and continuously judge, generate multi-parameter trend set;

[0009] Edge fast identification module: based on the multi-parameter trend set, execute synchronous discrimination on manhole cover displacement and inclination and trigger signal, use convolutional neural network to process trigger signal and gas mutation difference value and execute accumulation, get high-risk signal, system match high-risk signal with trend set, form abnormal trigger set;

[0010] Space correlation analysis module: based on the abnormal trigger set, build manhole cover node, and weight the pipeline connection value and the geographical distance to superimpose the trigger set, calculate node risk and execute neighborhood propagation operation, weight risk score and propagation probability, generate node propagation set;

[0011] Linkage early warning execution module: based on the node propagation set, calculate upstream risk value and compare with proportional threshold value, output link signal, and cross compare with node propagation set, determine cross-region risk, execute region division and calculate average value, compare with standard, generate linkage early warning set.

[0012] As a further scheme of the application, the monitoring value set includes water-gas difference amount, heat and humidity change amount, manhole cover displacement amount, the multi-parameter trend set includes water-gas trend prediction, heat and humidity trend prediction, displacement trend prediction, the abnormal trigger set includes structure trigger signal, high-risk signal, high-level trigger signal, the node propagation set includes node risk score, propagation probability value, weighted result value, and the linkage early warning set includes link early warning signal, cross-region risk value, block risk value.

[0013] As a further scheme of the application, the manhole cover state monitoring module includes:

[0014] Water-gas difference sub-module: based on the water level probe and gas sensor value of the manhole cover, calculate the water level value minus the gas value and generate the difference sequence, continuously store the difference sequence, calculate the average deviation of the value sequence, and execute the stabilization processing on the deviation sequence, output the water-gas difference amount;

[0015] Heat and humidity displacement sub-module: based on the water-gas difference amount, process temperature and humidity sequence and gradually superimpose to form a curve, execute interval integration on the inclination counter value and vibration counter value to get integral vector, collect the curve and the integral vector set, and establish the monitoring value set.

[0016] As a further scheme of the application, the time series prediction module includes:

[0017] Water-gas trend sub-module: based on the monitoring value set, subtract the water level value from the gas value at the same time point to form a difference sequence, and construct a time curve by point-by-point extension, perform continuous cumulative smoothing processing within the section, output the trend change trajectory, and generate water-gas trend prediction;

[0018] Heat and humidity trend sub-module: based on the water-gas trend prediction, segment the temperature sequence and humidity sequence according to the time interval, and perform repeated stacking of the segmented results according to the cycle length to form a fluctuation pattern, compare the peak and valley differences of the fluctuation pattern to determine the change form, and generate heat and humidity trend prediction;

[0019] Displacement fusion sub-module: based on the heat and humidity trend prediction, calculate the manhole cover displacement value within the time window point by point, superimpose the calculation result with the previous trend curve point by point and perform weighted operation to form a fusion sequence, and input the fusion sequence into a long short-term memory network to complete nonlinear learning, compare the output sequence with the threshold interval in time sequence, and perform judgment within the continuous segment to generate a multi-parameter trend set.

[0020] As a further scheme of the application, the long short-term memory network first splits the sequence into multiple input segments according to time steps, inputs each segment into a long short-term memory unit, sequentially completes input gate control, forget gate screening, state update and output gate filtering in each unit, superimposes the state of the previous time segment with the current input to form a memory state, and passes the state to the next time segment, outputs the prediction sequence after all segment processing is completed, the prediction sequence maintains a time-dependent relationship, and outputs a multi-dimensional prediction result.

[0021] As a further scheme of the application, the edge rapid identification module comprises:

[0022] Displacement inclination sub-module: based on the multi-parameter trend set, synchronously compare the manhole cover displacement value and the inclination value at the same time, mark the trigger point when both values exceed the limited range at the same time, and store the trigger point sequence to generate a structure trigger signal;

[0023] Gas accumulation sub-module: based on the structure trigger signal, subtract the gas value point by point in the continuous time segment to form a difference sequence, and generate an accumulation sequence by sequentially accumulating the difference sequence, input the accumulation sequence into a convolutional neural network to complete feature extraction, and compare the output result with the limit value segment by segment and filter the abnormal segment to generate a high-risk signal;

[0024] Risk trigger sub-module: based on the high-risk signal, compare the prediction values in the corresponding time period in the multi-parameter trend set, output a trigger judgment when the comparison ratio exceeds the limited standard, convert the judgment result into a sequence and aggregate it to generate an abnormal trigger set.

[0025] As a further scheme of the present application, the convolutional neural network first cuts the cumulative sequence into multiple input segments of fixed length, each segment is sequentially sent into the convolutional layer, and the local feature extraction is completed under the sliding window of the convolution kernel, the extraction result is output through nonlinear transformation feature map, and the feature map is further input into the pooling layer for down-sampling compression to reserve the main change information, the multi-layer convolution and pooling are repeatedly superimposed to form deep feature expression, the deep feature is sent into the full connection layer to form output result point by point mapping, and the output result retains the significant features of abnormal changes, which is used for subsequent comparison with the limit value and screening of abnormal sections.

[0026] As a further scheme of the present application, the spatial correlation analysis module comprises:

[0027] The node construction submodule: based on the abnormal trigger set, a node is established through the geographic position mapping of the well lid, the pipe communication value and the geographic distance are weighted and accumulated, the abnormal trigger set is superimposed item by item, the node corresponding risk value is output, and the node risk score is generated;

[0028] The propagation weighting submodule: based on the node risk score, the risk values of adjacent nodes are diffused and transmitted, the propagation proportion is calculated, the node risk value and the propagation proportion are weighted point by point, the coverage area weighted set is output, and the node propagation set is generated.

[0029] As a further scheme of the present application, the abnormal trigger set is superimposed item by item, each trigger signal in the set is arranged in time sequence, the amplitude normalization processing is performed on multiple trigger signals in the same time slice, then the node time trigger value is obtained by adding each trigger signal item by item according to the node position, the node current comprehensive value is formed by adding the node time trigger value and the historical trigger baseline value of the corresponding node, if the node exists trigger in continuous multiple time slices, the node comprehensive values of each time slice are accumulated in turn, and the superimposed node comprehensive trigger sequence is output.

[0030] As a further scheme of the present application, the linkage early warning execution module comprises:

[0031] The link accumulation submodule: based on the node propagation set, the risk values of upstream nodes are accumulated level by level, the proportion threshold is compared point by point, the link overrun identification signal is output, the signal sequence is uniformly converted into a link trigger instruction, and the link early warning signal is obtained;

[0032] The area determination submodule: based on the link early warning signal, the link early warning signal and the node propagation set are cross-compared, the cross-region risk value is generated, the average risk value of the area division is extracted, the standard threshold is compared step by step, and the linkage early warning set is generated.

[0033] Compared with the prior art, the present application has the advantages and positive effects that:

[0034] In the application, the displacement value of the well lid is calculated point by point in a time window, and is superimposed with a previous trend curve to form a fusion sequence, and then the fusion sequence is input into a long short-term memory network to complete nonlinear prediction, and the output result is compared with a threshold interval in sections to ensure that abnormality determination is realized in continuous segments, and through the combination of time series prediction and nonlinear learning, the displacement trend change and the environmental trend are uniformly modeled, and the continuity and accuracy of abnormality determination are improved.

[0035] In the application, the displacement and the inclination of the well lid are synchronously discriminated in the trend set to trigger an initial signal, and the signal is accumulated with a gas mutation difference value, and then the characteristics are extracted and matched through a convolutional neural network to output a high-risk signal, thereby enhancing the sensitive capture of sudden abnormalities, enabling high-risk events under complex conditions to be quickly locked, reducing the false alarm probability, and improving the timeliness of response under a sudden situation. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The system flowchart of the application. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application.

[0038] Example 1

[0039] Please refer to Figure 1 The application provides a technical scheme: a well lid comprehensive environmental monitoring system based on the Internet of Things, comprising:

[0040] The well lid state monitoring module: based on the well lid water level probe, gas sensor, temperature and humidity element, inclinometer and vibration meter, the water level value and the gas value are subtracted, the temperature and humidity sequence is accumulated, the inclination value and the vibration value are integrated, and a monitoring value set is generated;

[0041] The time series prediction module: based on the monitoring value set, the heat and humidity sequence is combined, the displacement amount is deduced, and a prediction sequence is generated by performing weighted fusion, the long short-term memory network is used to compare the prediction sequence with the threshold value and continuously judge, and a multi-parameter trend set is generated;

[0042] The edge rapid identification module: based on the multi-parameter trend set, the displacement of the well lid and the inclination are synchronously discriminated and a signal is triggered, the convolutional neural network is used to process the triggered signal and the gas mutation difference value and perform accumulation to obtain a high-risk signal, and the system matches the high-risk signal with the trend set to form an abnormal trigger set;

[0043] Spatial correlation analysis module: based on the abnormal trigger set, build the manhole cover node, and superimpose the pipe connectivity value and the geographical distance weighted trigger set, calculate the node risk and perform the neighborhood propagation operation, weight the risk score and the propagation probability, and generate the node propagation set;

[0044] Linkage early warning execution module: based on the node propagation set, calculate the upstream risk value and compare it with the proportion threshold value, output the link signal, and cross compare with the node propagation set to determine the cross-region risk, perform regional division and calculate the average value, compare with the standard to generate the linkage early warning set.

[0045] The monitoring value set includes water-gas difference, heat and humidity change, and manhole cover displacement; the multi-parameter trend set includes water-gas trend prediction, heat and humidity trend prediction, and displacement trend prediction; the abnormal trigger set includes structure trigger signal, high-risk signal, and high-level trigger signal; the node propagation set includes node risk score, propagation probability value, and weighted result value; and the linkage early warning set includes link early warning signal, cross-region risk value, and block risk value.

[0046] The manhole cover state monitoring module includes:

[0047] Water-gas difference sub-module: based on the water level probe and gas sensor values of the manhole cover, calculate the water level value minus the gas value and generate the difference sequence, continuously store the difference sequence, calculate the average deviation of the value sequence, and perform stabilization processing on the deviation sequence, output the water-gas difference;

[0048] Heat and humidity displacement sub-module: based on the water-gas difference, process the temperature and humidity sequence and gradually superimpose to form a curve, perform interval integration on the tilt angle and vibration count values to obtain an integral vector, and set the curve and integral vector set to establish the monitoring value set;

[0049] Water-gas difference sub-module: based on the water level probe and gas sensor values of the manhole cover, use the difference algorithm to perform point-by-point subtraction processing on the water level value and the gas value, the subtraction operation aligns the data with the timestamp and calculates the difference value item by item, and the difference result is written to the cache area address 0x1000 starting position, the address is stored in sequence, the mean square deviation calculation method is used to perform point-by-point subtraction of the sequence in the cache and accumulation, a data block with a fixed window size of 100 is used for batch calculation, and the result of each window is written to the deviation sequence buffer, and the exponential weighted smoothing method is used to stabilize the deviation sequence, the smoothing parameter is set to 0.3, the previous value and the current input are recursively called in time sequence to perform weighted operation, and finally the water-gas difference is output;

[0050] The hot and wet displacement sub-module: based on the water vapor difference amount, a cumulative curve generation method is used to perform point-by-point superposition operation on the temperature sequence and the humidity sequence, the superposition operation is in units of one second of sampling interval, the superposition value is written into the curve array step by step, the curve is continuously stored in the starting address 0x2000 of the cache area in time sequence, the interval integration method is used to perform integration processing on the inclination count value and the vibration count value, the integration range is set to one segment for every 60 sampling points, the cumulative value in the segment is calculated and written into the integral array in turn, the array number order is incremented, the set method is used to index match the curve array and the integral array, the set structure uses a hash table for index allocation and is stored in a unified cache space, and finally a monitoring value set is established.

[0051] The time series prediction module comprises:

[0052] The water vapor trend sub-module: based on the monitoring value set, the water level value and the gas value at the same time point are subtracted to form a difference sequence, and a time curve is constructed by point-by-point extension, continuous cumulative smoothing processing is performed in the segment, a trend change trajectory is output, and a water vapor trend prediction is generated;

[0053] The hot and wet trend sub-module: based on the water vapor trend prediction, the temperature sequence and the humidity sequence are segmented according to the time interval, and the segmented results are repeatedly stacked according to the cycle length to form a fluctuation mode, the peak and valley difference of the fluctuation mode is compared to determine the change form, and a hot and wet trend prediction is generated;

[0054] The displacement fusion sub-module: based on the hot and wet trend prediction, the well lid displacement value in the time window is calculated point by point, the calculation result is superimposed with the previous trend curve point by point and weighted operation is performed to form a fusion sequence, and the fusion sequence is input into a long short-term memory network to complete nonlinear learning, the output sequence is compared in segments according to time sequence and threshold interval, and the judgment is performed in the continuous segment to generate a multi-parameter trend set;

[0055] Water gas trend sub-module: based on the monitoring value set, the difference algorithm is used to subtract the water level value and the gas value at the same time point, the subtraction processing is completed by setting a unified timestamp index method, the timestamp step is one second, the starting index is zero, each subtraction result is written into the difference sequence cache in order, the cache address starts from 0x3000 and increases by four bytes, the generated difference sequence is processed by point-by-point extension using the sequence extension method, the extension processing is realized by copying the adjacent values at the end of the sequence and inserting the interpolation points, each interpolation point is calculated according to the linear interpolation of the previous and next two items, the number of interpolation points is set to five, the extended sequence is written into the curve array point by point, the array number starts from 001 and increases in turn, the cumulative smoothing method is used in the curve array to perform continuous cumulative operation on the values in the section, the section length is set to 100 sampling points, each section result is stored in the smoothing array, the array position increases according to the section number, all smoothing results are reorganized into trend change track according to time sequence, and finally water gas trend prediction is output;

[0056] Hot and humid trend sub-module: based on water gas trend prediction, the temperature sequence and humidity sequence are segmented by using the segmented superposition method, the segmentation size is set to 60 seconds, each interval is independently numbered and arranged in time sequence, the segmented results are stacked by using the period stacking method according to the period length, the period length is set to five segments, five segments are sequentially stacked in the same array space, the stacking operation is written by point-by-point addition through the index pointer, the pointer step is 1, the storage address starts from 0x4000, the formed stacking result is reorganized into fluctuation mode array, the fluctuation mode is analyzed by using the peak-valley difference comparison method, the local maximum value and the local minimum value of each period stacking sequence are searched, the difference between the maximum value and the minimum value is written into the difference table, the difference table number starts from 100 and increases, the change form is numbered according to the record of the difference table, the numbering result is stored in association with the fluctuation mode, and finally hot and humid trend prediction is output;

[0057] Displacement fusion submodule: based on the prediction of thermal and humidity trends, the displacement value of the manhole cover is calculated point by point within the time window using the window calculation method, the size of the time window is set to 120 seconds, the displacement value in each window is calculated in turn and written into the calculation array, the starting address of the array is 0x5000, the calculation result and the previous trend curve are superimposed point by point using the weighted superposition method, the weight parameters are set to 0.6 and 0.4, the superimposed value is written into the fusion array, and the fusion array number is sequentially incremented. The fusion array is used as input to the long short-term memory network algorithm for nonlinear learning, the network structure is set to three layers, the input layer node number is 64, the hidden layer node number is 128, the output layer node number is 32, the activation function uses tanh, the learning rate is set to 0.001, the batch size is 32, and the training round number is 500. The training process uses sequential gradient descent to update the parameters, and the output prediction sequence is compared with the threshold interval in time sequence using the piecewise comparison method. The threshold interval is stored in the threshold table, and the table number starts from 200 and increments. Each segment of the output sequence is compared with the threshold table one by one, and the determination function is called to complete the determination and write into the determination array. Finally, a multi-parameter trend set is generated.

[0058] The long short-term memory network first splits the sequence into multiple input segments according to the time step, inputs each segment into the long short-term memory unit, and sequentially completes the input gate control, forget gate filtering, state update, and output gate filtering in each unit. The state of the previous time segment is superimposed with the current input to form a memory state, and this state is passed to the next time segment. After processing all segments, the prediction sequence is output, which maintains the time-dependent relationship and outputs a multi-dimensional prediction result.

[0059] The edge rapid identification module includes:

[0060] Displacement and inclination sub-module: based on the multi-parameter trend set, the displacement value and the inclination value of the manhole cover are compared at the same time. When both values exceed the limited range at the same time, a trigger point is marked, and a trigger point sequence is stored to generate a structure trigger signal.

[0061] Gas accumulation submodule: based on the structure trigger signal, the gas value is subtracted point by point in the continuous time slice to form a difference sequence, and the difference sequence is sequentially accumulated to generate an accumulation sequence. The accumulation sequence is input into the convolutional neural network to complete feature extraction, and the output result is compared with the limit value in segments to filter abnormal segments and generate a high-risk signal.

[0062] Risk trigger submodule: based on the high-risk signal, compare the prediction values of the corresponding time period in the multi-parameter trend set. When the comparison ratio exceeds the limited standard, output the trigger judgment, convert the judgment result into a sequence and aggregate it to generate an abnormal trigger set.

[0063] Displacement Inclination Submodule: Based on the multi-parameter trend set, the synchronous comparison algorithm is used to synchronize the displacement value and the inclination value of the manhole cover at the same time. The comparison process aligns the two sequences through a unified timestamp. The timestamp sampling interval is one second, the starting point is zero, and the end point is the maximum length of the sequence. The displacement value and the inclination value are compared at the same index position in a point-by-point traversal manner. The comparison threshold is set to displacement greater than 5 mm and inclination greater than 2 degrees. When the conditions are met at the same time, write a mark value of 1 in the trigger mark array, and write a mark value of 0 when the conditions are not met. The mark array address starts at 0x6000, and the address step is 0x0004. The generated trigger marks are written into the trigger point sequence point by point. The trigger point sequence is numbered in ascending order from 001. Each trigger point contains three parameters: timestamp, displacement value, and inclination value. The trigger point sequence is sequentially stored, and the final output structure is the trigger signal;

[0064] Gas Accumulation Submodule: Based on the structure trigger signal, the difference accumulation method is used to subtract the gas value point by point in the continuous time slice. The difference method is to subtract the previous time slice value from the current time slice value. The time slice length is set to five seconds. The difference value is written into the difference sequence array point by point. The array address starts at 0x7000, and the address increment step is 0x0004. The difference sequence is sequentially accumulated to obtain the accumulation sequence. The accumulation operation is realized by sequential traversal. The result is written into the accumulation array, which is stored in an incremental manner. The accumulation array is input into the convolutional neural network algorithm to complete feature extraction. The convolutional neural network structure is input layer size 64, convolution layer number 3, convolution kernel size 3x3, step size 1, padding method same, pooling layer uses maximum pooling, pooling window size 2, activation function relu, full connection layer node number 128, output layer node number 32, batch size 64, learning rate 0.001, and iteration number 300. The output result is compared with the limit value in time sequence, and the limit value is stored in the limit table. The table number starts from 300. Each output segment is compared with the limit segment. When the output value is greater than the limit value, the abnormal section is recorded, and the high-risk signal is finally generated.

[0065] Risk trigger submodule: based on high-risk signals, using proportional comparison method to compare the prediction value of the corresponding time period in the multi-parameter trend set, the comparison process aligns the high-risk signals and the prediction sequence through the time stamp, the time interval is set to one second, after alignment, calculate the proportion of the number of high-risk signals in the number of prediction values in this time period, the proportion calculation is completed by the counter accumulation method, the limited standard is set to the proportion greater than 0.4, when the comparison ratio exceeds the limited standard, output the trigger judgment, the trigger judgment result is written into the judgment sequence, the judgment sequence number starts from 400, each record includes three fields of time stamp, comparison ratio and judgment result, the judgment sequence is executed Set operation, set operation is completed by index mapping through hash index structure, all results are written into a unified cache area, and finally an abnormal trigger set is generated.

[0066] Convolutional neural network, first, the cumulative sequence is cut into multiple input segments according to the fixed length, each segment is sent into the convolution layer in turn, and the local feature extraction is completed under the sliding window of the convolution kernel. The extraction result is output as a feature map through a nonlinear transformation. The feature map is then input into the pooling layer for down-sampling compression to retain the main change information. The multi-layer convolution and pooling are repeatedly superimposed to form a deep feature expression. The deep feature is input into the fully connected layer for point-by-point mapping to form an output result. The output result retains the significant features of abnormal changes, which are used for subsequent segment-by-segment comparison with the limit value and screening of abnormal sections.

[0067] Spatial correlation analysis module includes:

[0068] Node construction submodule: based on the abnormal trigger set, nodes are established through the geographic location mapping of the manhole cover. The pipeline connectivity value and the geographic distance are weighted and accumulated. The abnormal trigger set is overlapped item by item. The node corresponds to the risk value output, and the node risk score is generated.

[0069] Propagation weighting submodule: based on the node risk score, the risk value of adjacent nodes is diffused and transmitted. The propagation ratio is calculated. The node risk value and the propagation ratio are weighted point by point. The coverage area weighted set is output, and the node propagation set is generated.

[0070] Node construction submodule: based on the abnormal trigger set, the geographic mapping method is used for mapping operation on the manhole cover coordinates, the coordinate data format is longitude and latitude with six decimal places, the mapping process binds and stores each manhole unique identifier with the longitude and latitude coordinates through the hash table, the initial capacity of the hash table is set to 1024, the load factor is 0.75, the node construction function is called for each mapping point to establish node index, the node number starts from N001 and sequentially increases, the weighted accumulation algorithm is executed on the pipe connectivity value and the geographic distance, the pipe connectivity value ranges from 0 to 1, the geographic distance unit is meter, the weight parameter is set to 0.7 and 0.3, the calculation method is to call the accumulation function node by node, multiply the connectivity value of each node by the weight and add the geographic distance multiplied by the weight to write into the weighted array, the starting address of the weighted array is 0x8000, the storage step is 0x0008, the weighted array is overlaid with the abnormal trigger set item by item, the trigger events in the abnormal trigger set are arranged in ascending order according to the timestamp, the risk factor value of each event is added to the weighted value of the corresponding node to write into the node risk array, the node risk array number starts from R001, the node risk array is outputted item by item, and finally the node risk score is generated;

[0071] Propagation weighting submodule: based on the node risk score, the propagation proportion calculation method is used to transfer the risk diffusion between adjacent nodes, the adjacent node relationship is described by an adjacency matrix, the matrix size is n x n, n is the number of nodes, and the matrix elements are 0 or 1 to represent whether there is a connectivity relationship between nodes, the propagation operation is performed on each adjacent node pair, the propagation proportion parameter is set to 0.5, which means that the risk value is transmitted to the adjacent node by half in a single propagation, the node risk value is weighted point by point with the propagation proportion, the weighted operation is completed through loop iteration, the risk values of the current node and the adjacent nodes are updated each time, and the update result is written into the propagation array, the propagation array number starts from P001, the storage location is at address 0x9000, and the address step is 0x0008. The weighted set is obtained by traversing the propagation array and merging in order of node index, and the weighted set of the coverage area is outputted after merging, and the node propagation set is generated.

[0072] The abnormal trigger set is overlaid item by item, the trigger signals in the set are arranged in chronological order, the amplitude normalization processing is performed on the multiple trigger signals in the same time slice, then the node moment trigger value is obtained by adding the trigger signals item by item according to the node position, the node moment trigger value is added to the historical trigger baseline value of the corresponding node to form the node current comprehensive value, if the node exists trigger in continuous multiple time slices, the node comprehensive values of the time slices are accumulated in turn, and the superimposed node comprehensive trigger sequence is outputted.

[0073] The linkage early warning execution module comprises:

[0074] Link accumulation submodule: based on the node propagation set, the upstream node risk value is accumulated step by step, and the proportional threshold is compared point by point, the link overrun identification signal is output, the signal sequence is uniformly converted into link trigger instruction, and the link early warning signal is obtained;

[0075] Region determination submodule: based on the link early warning signal, the link early warning signal and the node propagation set are cross-compared, the cross-region risk value is generated, the average risk value of the region division is extracted, the standard threshold is compared step by step, and the linkage early warning set is generated;

[0076] Link accumulation submodule: based on the node propagation set, the upstream node risk value is accumulated step by step, and the proportional threshold is compared point by point, the link overrun identification signal is output, the signal sequence is uniformly converted into link trigger instruction, and the link early warning signal is obtained;

[0077] The regional determination sub-module: based on the link early warning signal, the cross comparison method is used to cross compare the link early warning signal and the node propagation set. The comparison process is completed by comparing the same time stamp and node index. If the time stamp is consistent, the trigger state of the link early warning signal is bound to the risk value in the node propagation set, and the binding result is written into the comparison table. The comparison table address starts from 0xB000, and the address step is 0x0008. The cross-regional risk generation method is executed on the comparison table. The method divides the node set by region index. The index is arranged from R001 to Rn according to the region number. The risk value sum in each region is calculated and stored in the region array. The region array number starts from Z001. The average risk value is extracted from the region array by using the average calculation method. The average calculation is completed by accumulating the risk value in the region and dividing by the number of nodes. The average result is written into the average array point by point. The average array storage address is 0xC000, and the step is 0x0004. The average array value is compared with the standard threshold value step by step. The standard threshold value table is stored in the fixed address 0xD000. Each average value is compared with the corresponding threshold value according to the index. When the average value is greater than the threshold value, it is marked as 1, otherwise it is marked as 0. All results are written into the early warning set array in order, and the linkage early warning set is finally generated.

[0078] The above is only the preferred embodiment of the present application, and does not limit the present application in other forms. Any skilled person in the art can modify or change the above disclosed technical content to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments without departing from the technical solution content of the present application still belongs to the protection scope of the present application.

Claims

1. An Internet of Things-based comprehensive environmental monitoring system for a well lid, characterized in that, The system comprises: a well lid state monitoring module: based on a well lid water level probe, a gas sensor, a temperature and humidity element, an inclinometer, and a vibration meter, the water level value and the gas value are subtracted, the temperature and humidity sequence is accumulated, the incline value and the vibration value are integrated, and a monitoring value set is generated; a time series prediction module: based on the monitoring value set, the heat and humidity sequence is combined, the displacement amount is deduced, and a prediction sequence is generated by weighted fusion, the prediction sequence is compared with the threshold value by using a long short-term memory network, and a multi-parameter trend set is generated by continuous judgment; an edge rapid identification module: based on the multi-parameter trend set, the well lid displacement and the incline are synchronously discriminated and a signal is triggered, the triggered signal and the gas mutation difference value are processed by using a convolutional neural network and accumulation is performed to obtain a high-risk signal, the system matches the high-risk signal with the trend set to form an abnormal trigger set; a spatial correlation analysis module: based on the abnormal trigger set, a well lid node is constructed, and the pipeline connection value and the geographical distance are weighted and superimposed on the trigger set, the node risk is calculated and neighborhood propagation operation is performed, the risk score and the propagation probability are weighted, and a node propagation set is generated; a linkage early warning execution module: based on the node propagation set, the upstream risk value is calculated and compared with the proportional threshold value, a link signal is output, and cross comparison is performed with the node propagation set to determine the cross-region risk, regional division is performed and the average value is calculated, compared with the standard, and a linkage early warning set is generated. 2.The Internet-of-Things based comprehensive environment monitoring system for manhole covers according to claim 1, characterized in that, The monitoring value set includes water-gas difference, heat and humidity change, well lid displacement, the multi-parameter trend set includes water-gas trend prediction, heat and humidity trend prediction, displacement trend prediction, the abnormal trigger set includes structure trigger signal, high-risk signal, high-level trigger signal, the node propagation set includes node risk score, propagation probability value, weighted result value, and the linkage early warning set includes link early warning signal, cross-region risk value, and block risk value. 3.The Internet-of-Things based comprehensive environment monitoring system for manhole covers according to claim 1, characterized in that, The well lid state monitoring module comprises: a water-gas difference submodule: based on the well lid water level probe and the gas sensor value, the water level value is subtracted from the gas value to generate a difference sequence, the difference sequence is continuously stored, the average deviation of the value sequence is calculated, and the deviation sequence is stabilized to output the water-gas difference; a heat and humidity displacement submodule: based on the water-gas difference, the temperature and humidity sequences are processed and gradually superimposed to form a curve, the incline value and the vibration value are integrated in the interval to obtain an integral vector, the curve and the integral vector set are combined to establish the monitoring value set. 4.The Internet-of-Things based comprehensive environment monitoring system for manhole covers according to claim 1, characterized in that, The time series prediction module comprises: a water-gas trend submodule: based on the monitoring value set, the water level value and the gas value are subtracted at the same time point to form a difference sequence, and a time curve is constructed by point-by-point extension, continuous accumulation smoothing is performed in the section to output the trend change trajectory, and water-gas trend prediction is generated; a heat and humidity trend submodule: based on the water-gas trend prediction, the temperature sequence and the humidity sequence are segmented according to the time interval, and the segmented results are repeatedly stacked according to the period length to form a fluctuation mode, the fluctuation mode is compared for peak and valley difference to determine the change form, and heat and humidity trend prediction is generated; The displacement fusion submodule: based on the heat and humidity trend prediction, the displacement value of the manhole cover in the time window is calculated point by point, the calculation result is superimposed on the previous trend curve point by point and weighted operation is performed, a fusion sequence is formed, the fusion sequence is input into the long short-term memory network to complete nonlinear learning, the output sequence is compared in sections in time sequence and threshold interval, the judgment is performed in the continuous section, and a multi-parameter trend set is generated. 5.The IoT-based comprehensive environment monitoring system for manhole covers according to claim 4, characterized in that, The long short-term memory network first divides the sequence into multiple input segments according to time steps, inputs each segment into a long short-term memory unit, sequentially completes input gate control, forget gate screening, state update and output gate filtering in each unit, superimposes the state of the previous time segment and the current input to form a memory state, and transmits the state to the next time segment. After processing all segments, the prediction sequence is output, the prediction sequence maintains the time dependence, and a multi-dimensional prediction result is output. 6.The IoT-based comprehensive environment monitoring system for manhole covers according to claim 1, wherein, The edge rapid identification module comprises: The displacement angle submodule: based on the multi-parameter trend set, the displacement value and the angle value of the manhole cover at the same time are compared synchronously, when both values exceed the limited range at the same time, a trigger point is marked, and a trigger point sequence is stored to generate a structure trigger signal; The gas accumulation submodule: based on the structure trigger signal, the gas value is subtracted point by point in the continuous time piece to form a difference sequence, the difference sequence is sequentially accumulated to generate an accumulation sequence, and the accumulation sequence is input into a convolutional neural network to complete feature extraction. The output result is compared with the limit value in sections and the abnormal section is screened to generate a high-risk signal; The risk trigger submodule: based on the high-risk signal, the prediction value of the corresponding time period in the multi-parameter trend set is compared, and when the comparison ratio exceeds the limited standard, a trigger judgment is output. The judgment result is converted into a sequence and is aggregated to generate an abnormal trigger set. 7.The IoT-based comprehensive environment monitoring system for manhole covers according to claim 6, wherein, The convolutional neural network first divides the accumulation sequence into multiple input segments according to a fixed length, each segment is sequentially sent into a convolutional layer, local feature extraction is completed under the sliding window of the convolution kernel, the extraction result is output as a feature map through nonlinear transformation, the feature map is further input into a pooling layer for down-sampling compression to reserve main change information, multiple layers of convolution and pooling are repeatedly superimposed to form deep feature expression, the deep feature is input into a fully connected layer for point-by-point mapping to form an output result. The output result retains the significant features of abnormal changes, which is used for subsequent comparison with the limit value in sections and screening of abnormal sections. 8.The IoT-based comprehensive environment monitoring system for manhole covers according to claim 1, wherein, The spatial correlation analysis module comprises: The node construction submodule: based on the abnormal trigger set, a node is established by mapping the geographic location of the manhole cover, the pipe connection value and the geographic distance are weighted and accumulated, the abnormal trigger set is superimposed item by item, the node corresponding risk value is output, and a node risk score is generated; The propagation weighting submodule: based on the node risk score, the risk values of adjacent nodes are diffused and transmitted, the propagation ratio is calculated, the node risk value and the propagation ratio are weighted point by point, the coverage area weighted set is output, and a node propagation set is generated. 9.The IoT-based comprehensive environment monitoring system for manhole covers according to claim 8, wherein, The abnormal trigger set is overlapped item by item, trigger signals in the set are arranged in time sequence, amplitude normalization is performed on multiple trigger signals of the same time slice, node time trigger values are obtained by adding items according to node positions, node current comprehensive values are formed by adding node time trigger values and historical trigger baseline values of corresponding nodes, if a node exists trigger in continuous multiple time slices, node comprehensive values of the time slices are accumulated in turn, and a node comprehensive trigger sequence after superposition is output. 10.The Internet-of-Things based comprehensive environment monitoring system for manhole covers according to claim 1, wherein, The linkage early warning execution module comprises: a link accumulation submodule: based on the node propagation set, risk values of upstream nodes are accumulated level by level, a proportional threshold value is compared point by point, a link overrun identification signal is output, a signal sequence is uniformly converted into a link trigger instruction, and a link early warning signal is obtained; a region determination submodule: based on the link early warning signal, the link early warning signal and the node propagation set are cross-compared, cross-region risk values are generated, average risk values are extracted according to region division, a standard threshold value is compared step by step, and a linkage early warning set is generated.

Images