Park data analysis system and method based on pipe gallery combined sensor

Abstract

The invention provides a park data analysis system and method based on a pipe gallery combined sensor, and belongs to the technical field of data processing and resource scheduling. The method comprises the following steps: collecting at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in a pipe gallery combined sensor; performing result output analysis on the at least two groups of park monitoring data sets by adopting a local prediction model to obtain at least two groups of prediction output results; based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results, determining a current pipe gallery data analysis mode; based on the current pipe gallery data analysis mode, sending an activation instruction to a back-end control platform, so that the back-end control platform starts or switches to a target cluster resource corresponding to the current pipe gallery data analysis mode; the analysis mode of the multi-dimensional to-be-analyzed data generated by the pipe gallery combined sensor is pre-judged in advance, then cluster resources matched with the multi-dimensional to-be-analyzed data are scheduled for processing, and resource waste can be avoided while result output delay is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing and resource scheduling, and in particular relates to a park data analysis system and method based on an integrated pipe gallery sensor, a non-volatile computer-readable storage medium for implementing the method, a computer program product, and an electronic device. Background Art

[0002] Pipe galleries are mainly divided into urban underground integrated pipe galleries and pipe galleries in large industrial parks / factories / workshops. Urban pipe galleries are generally underground hidden works and are located in areas where various signals and pipelines converge; pipe galleries in parks / workshops are arranged outside devices or factories, generally in the air, supported by brackets, forming a shape similar to a corridor, and a small number are also located underground.

[0003] To fully ensure the safe operation of the pipe gallery, it is necessary to monitor its internal operating conditions in real time. Accurately, continuously and in real time obtain environmental monitoring data, equipment operation information, and status data of physical machines, containers, clusters, and services. Specifically, the pipe gallery is usually equipped with various types of integrated sensors to simultaneously monitor the temperature, humidity, and concentrations of various gases in the park pipe gallery, the pressure of various fluid transportation pipelines, the water accumulation level in the pipe gallery, the loss of communication cables, and the charge change of natural gas pipelines. By comprehensively analyzing these parameters, the health status of the pipe gallery structure under the influence of external vibrations and other factors is evaluated. For example, Chinese invention patent CN114638557B proposes a method for collecting operation data of urban integrated pipe galleries, which can perform multi-sensor data fusion based on the temperature data of different pipeline temperature sensors, thereby avoiding a large amount of data duplication and ensuring that the fused data has anisotropy, so as to better monitor the operation status of urban integrated pipe galleries.

[0004] Since the number and types of integrated sensors in the pipe gallery structure are large, the acquisition frequency is high, and the acquisition period is continuous (usually round-the-clock monitoring), the amount of data in the acquired monitoring data set is very large. Even after relevant processing (such as the multi-sensor data fusion and avoiding a large amount of data duplication mentioned in the aforementioned patent), there is still a lot of data to be processed sent to the backend. There are also differences in its data sources, data dimensions, data formats, etc., and there are also different display dimension and display format requirements for the output results after data processing. Furthermore, it is necessary to schedule a large number of different types of device resources for processing. If all device resources are in the on state all the time to process the above data, there will be a waste of device resources; if some device resources are selectively turned on, there will be problems such as insufficient data resources or the need to frequently turn on different modes of cluster resources, which may cause a delay in the output of data processing results and pose a hidden danger to the safety and normal operation of the park. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a park data analysis system and method based on an integrated pipe gallery sensor, a non-volatile computer-readable storage medium for implementing the method, a computer program product, and an electronic device.

[0006] In a first aspect of the present invention, a park data analysis method based on an integrated pipe gallery sensor is provided. The method is implemented based on a data middleware, and the method includes:

[0007] Collecting at least two sets of park monitoring data sets sent by at least two groups of monitoring sensors in the integrated pipe gallery sensor;

[0008] Performing result output analysis on the at least two sets of park monitoring data sets using a local prediction model to obtain at least two sets of prediction output results;

[0009] Determining a current pipe gallery data analysis mode based on the at least two sets of park monitoring data sets and the at least two sets of prediction output results;

[0010] Based on the current pipe gallery data analysis mode, sending an activation instruction to a backend control platform to cause the backend control platform to activate or switch to a target cluster resource corresponding to the current pipe gallery data analysis mode.

[0011] The integrated pipe gallery sensor includes at least two of a temperature integrated sensor, a pressure integrated sensor, a natural gas pipeline charge integrated sensor, and a communication cable pipeline integrated sensor.

[0012] The prediction output result includes the delay of the output result; the pipe gallery data analysis mode includes a multi-cluster analysis mode or a single-cluster analysis mode;

[0013] When the delay of the output result is greater than a preset threshold, determining that the current pipe gallery data analysis mode is a multi-cluster analysis mode, and the delay of the output result after performing data analysis on the park monitoring data set sent by the data middleware using the target cluster resource corresponding to the multi-cluster analysis mode is less than the preset threshold.

[0014] The pipe gallery data analysis mode further includes a homomorphic analysis mode and a polymorphic analysis mode;

[0015] The cluster resources in the homomorphic analysis mode provide display resource support in the same format;

[0016] The cluster resources in the polymorphic analysis mode provide display resource support in at least two formats;

[0017] The cluster resources in the multi-cluster analysis mode include at least a first cluster and a second cluster, and the first cluster and the second cluster are distributed at different locations.

[0018] Each group of monitoring sensors in the pipe gallery combined sensor is equipped with an edge data detection unit;

[0019] When the edge data analysis unit detects that a park monitoring data set in a current time period of a certain monitoring sensor has an abnormality, the park monitoring data set with the abnormality is sent to the data middleware.

[0020] The two groups of park monitoring data sets include a first park monitoring data set and a second park monitoring data set;

[0021] The using of the local prediction model to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results specifically includes:

[0022] Using the first park monitoring data set, the length of the first park monitoring data set, and the generation time period of the first park monitoring data set as inputs of the local prediction model to obtain a first prediction output result;

[0023] Using the second park monitoring data set, the length of the second park monitoring data set, and the generation time period of the second park monitoring data set as inputs of the local prediction model to obtain a second prediction output result;

[0024] The first predicted output result or the second predicted output result includes: the size of the output result, the display format of the output result and the delay of the output result.

[0025] Based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results, determining the current pipeline corridor data analysis mode specifically includes:

[0026] When the display format of the first prediction output result is the same as that of the second prediction output result, it is determined that the current pipeline corridor data analysis mode is a homomorphic analysis mode; otherwise, it is determined that the current pipeline corridor data analysis mode is a polymorphic analysis mode.

[0027] Based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results, determining the current pipeline corridor data analysis mode specifically includes:

[0028] When the size of the output result or the delay of the output result in the first prediction output result or the second prediction output result meets the preset standard, the current pipeline corridor data analysis mode is determined to be a multi-cluster analysis mode; otherwise, the current pipeline corridor data analysis mode is determined to be a single cluster analysis mode.

[0029] In the second aspect, corresponding to the method of the first aspect, a park data analysis system based on a pipe gallery combined sensor is also proposed, which can be used to implement all or part of the steps of the aforementioned park data analysis method based on a pipe gallery combined sensor.

[0030] The system includes a data middleware and a backend control platform, wherein the data middleware configures a local prediction model and the backend control platform connects multiple cluster resources;

[0031] Each group of monitoring sensors in the pipe gallery combined sensor is equipped with an edge data detection unit, and the edge data detection unit performs anomaly detection on the park monitoring data generated by each group of monitoring sensors within the current preset time period, and when an anomaly is detected, sends the park monitoring data set with the anomaly to the data middleware;

[0032] The data middleware collects at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the pipe gallery combined sensor, uses a local prediction model to perform result output analysis on the at least two groups of park monitoring data sets, and obtains at least two groups of prediction output results;

[0033] Determine a current pipe gallery data analysis mode based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results;

[0034] Based on the current pipe gallery data analysis mode, an activation instruction is sent to the backend control platform so that the backend control platform starts or switches to the target cluster resources corresponding to the current pipe gallery data analysis mode;

[0035] The target cluster resources are used to perform data analysis on the campus monitoring data set sent by the data middleware.

[0036] The pipeline gallery combination sensor includes at least two of a temperature combination sensor, a pressure combination sensor, a natural gas pipeline charge combination sensor, and a communication cable pipeline combination sensor.

[0037] Alternatively, the pipe gallery combined sensor includes a natural gas pipeline sensor, a power cable sensor, a drainage pipeline sensor, a communication pipeline sensor, and a heating pipeline sensor;

[0038] The natural gas pipeline sensor is used to detect changes in charge around the natural gas pipeline;

[0039] The power cable sensor is used to detect whether the power cable is operating normally;

[0040] The drainage pipe sensor is used to detect various data indicators of water supply and drainage;

[0041] The heating pipe sensor is used to measure the temperature index of the heating pipe.

[0042] The prediction output result includes a delay of the output result; the pipeline corridor data analysis mode includes a multi-cluster analysis mode or a single cluster analysis mode;

[0043] When the delay of the output result is greater than a preset threshold, it is determined that the current corridor data analysis mode is a multi-cluster analysis mode, and the delay of the output result after performing data analysis on the park monitoring data set sent by the data middleware using the target cluster resources corresponding to the multi-cluster analysis mode is less than the preset threshold.

[0044] The aforementioned campus data analysis method based on the corridor combined sensor can also be connected to the cloud resource platform through various forms of electronic devices and automatically implemented through computer program instructions; the computer program instructions can be stored in different forms of storage media and loaded into computer electronic devices for execution.

[0045] In the third aspect of the present invention, a non-volatile computer-readable storage medium is also provided for storing computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes all or part of the steps of the aforementioned campus data analysis method based on the corridor combined sensor.

[0046] In the fourth aspect of the present invention, a computer device is also proposed, which includes a processor and a memory, the memory is used to store instructions, and the processor is used to call the instructions in the memory, so that the computer device executes the aforementioned campus data analysis method based on the corridor combined sensor.

[0047] In the fifth aspect of the present invention, a computer program product is also proposed, which includes a computer program. When the computer program is executed, all or part of the steps of the aforementioned campus data analysis method based on the corridor combined sensor are implemented.

[0048] The technical solution of the present invention is as follows: first, when the edge data analysis unit detects that there is an abnormality in the park monitoring data set within the current time period of a certain monitoring sensor, the park monitoring data set with the abnormality is sent to the data middleware; secondly, the data middleware collects at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the corridor combination sensor, and uses a local prediction model to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results; finally, based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results, the current corridor data analysis mode is determined; based on the current corridor data analysis mode, an activation instruction is sent to the back-end control platform to enable the back-end control platform to start or switch to the target cluster resources corresponding to the current corridor data analysis mode. Therefore, the present invention can predict the analysis mode of the multi-dimensional data to be analyzed generated by the corridor combination sensor in advance, and then schedule the cluster resources matching it for processing, thereby reducing the result output delay and avoiding resource waste.

[0049] Further advantages of the present invention will be further detailed in the specific embodiment part in combination with the accompanying drawings of the specification. Description of the Drawings

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for use in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0051] Figure 1 It is the main flowchart of the method for analyzing park data based on the combined sensors in the pipe gallery according to an embodiment of the present invention;

[0052] Figure 2 It is the schematic flowchart of the process of the combined sensors in the pipe gallery entering the local prediction model after data acquisition;

[0053] Figure 3 It is the schematic diagram of the composition of the functional units of a system for analyzing park data based on the combined sensors in the pipe gallery according to an embodiment of the present invention;

[0054] Figure 4 It is the schematic diagram of the combination of device resources of different clusters for processing the data of the combined sensors in the pipe gallery; Detailed Description of the Embodiments

[0055] In the specific implementation of the present application, for the embodiments of the relevant technical solutions, if they involve user-related data, when the embodiments of the present application are applied to specific products or technologies, user permission or consent needs to be obtained, and the collection, use, and processing of the relevant data need to comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0056] See Figure 1 , Figure 1 It is the main flowchart of the method for analyzing park data based on the combined sensors in the pipe gallery according to an embodiment of the present invention.

[0057] Figure 1 The method of the embodiment includes multiple steps. For convenience of description, the following numbers are assigned to each step (the numbers are omitted in the accompanying drawings):

[0058] S1: Collect at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the combined sensors in the pipe gallery;

[0059] S2: Use the local prediction model to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results;

[0060] S3: Determine a current pipeline corridor data analysis mode based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results;

[0061] S4: Based on the current pipe gallery data analysis mode, an activation instruction is sent to the backend control platform, so that the backend control platform starts or switches to the target cluster resources corresponding to the current pipe gallery data analysis mode.

[0062] Next, the implementation principle of the above steps S1-S4 will be specifically introduced with a more specific example.

[0063] In a specific park, in order to monitor the relevant environmental parameters of a workshop, such as pipeline temperature, pressure, charge value, etc., the corridor structure is arranged with a variety of combined sensors. Each combined sensor contains at least one specific type of physical sensor or sensor unit, which is used to monitor the corresponding type of target data volume.

[0064] For example, a pipeline is arranged with multiple temperature combination sensors, multiple pressure combination sensors, and multiple charge combination sensors, which are respectively used to monitor the temperature, pressure, and charge values ​​at multiple predetermined positions of the pipeline.

[0065] As a safety monitoring measure, when the temperature exceeds the first preset value, an acoustic warning signal needs to be issued; when the temperature exceeds the second preset value, an acoustic warning signal needs to be issued at the same time and a yellow alarm sign is displayed in the background. The yellow alarm sign needs to identify the specific location of the pipeline with abnormal temperature. The second preset value is greater than the first preset value.

[0066] As another safety monitoring shows, when the pipeline pressure exceeds the third preset value, it is necessary to display the specific locations of all pipelines where the pipeline pressure exceeds the third preset value in the background, and fit the pressure prediction value of the pipeline at the corresponding location for the next time period, and display the pressure prediction value on the background display in color from dark to light according to the size of the prediction value.

[0067] It can be seen that for the case of temperature monitoring (assuming case 1), the output result after the background processing data may only require audio resource support, or require audio resources and graphics card resources to support it together;

[0068] For the situation of pressure monitoring (assuming situation 2), the output results after background data processing require the support of intensive computing resources (such as CPU computing cards or even NPU (neural network processor) model training cards), as well as corresponding graphics card resources (including GPU resource cards).

[0069] Therefore, in Case 1, actually only audio resources or a combination of audio resources and graphics card resources need to be scheduled in the background, without the need for other resources (such as NPU); while in Case 2, the resources that actually need to be scheduled in the background include CPU, GPU, and even NPU, etc.

[0070] In other words, the types of resources that actually need to be scheduled in the background are different in different cases.

[0071] In the prior art, since each sensor of the in-pipe gallery sensor in the park basically operates all day long, and the multi-dimensional data generated by it is also uninterrupted. Therefore, in order to meet the diversity of data processing, usually all device resources are kept on all the time to process the above data, resulting in waste of device resources; if some device resources are selectively turned on, problems such as insufficient data resources or the need to frequently turn on different modes of cluster resources will be faced, which may cause delays in the output of data processing results, bringing potential hazards to the safety and normal operation of the park.

[0072] For this reason, the present application proposes corresponding improvement solutions.

[0073] It can be understood that the examples given in the above description are only illustrative. The types and quantities of actual in-pipe gallery sensors in the park, as well as the types and quantities of resources that can be called in the background, are variable and not limited by the above description.

[0074] As a more specific example, the in-pipe gallery combined sensor includes at least two of a temperature combined sensor, a pressure combined sensor, a natural gas pipeline charge combined sensor, and a communication cable pipeline combined sensor.

[0075] Alternatively, the in-pipe gallery combined sensor includes a natural gas pipeline sensor, a power cable sensor, a drainage pipeline sensor, a communication pipeline sensor, and a heating pipeline sensor;

[0076] The natural gas pipeline sensor is used to detect the charge change around the natural gas pipeline;

[0077] The power cable sensor is used to detect multiple index values of the operating state of the power cable;

[0078] The drainage pipeline sensor is used to detect various data indexes of the water supply and drainage;

[0079] The heating pipeline sensor is used to measure the temperature index of the heating pipeline.

[0080] In practical applications, the resources that can be called by the background are usually implemented in the form of a cluster in this embodiment. Each cluster can include various hardware resources and software resources. The hardware resources can include audio components (audio output cards), video components (video display cards / graphics cards, GPUs, etc.), model training cards (NPUs, etc.), and data processing cards (CPUs). The software resources include data processing models (such as data fitting visualization models, data prediction models), data analysis models (such as trend analysis models), etc. It can be understood that software resources and hardware resources are usually corresponding. For example, the CPU is usually used for data trend analysis models, the GPU is usually used for data visualization calculation models, and the NPU can be used for data prediction models, etc.

[0081] The resources that can be called by the background include not only the types of callable hardware resources and software resources, but also the quantity of each type of hardware resource and software resource. For example, calling 1 CPU card and a data trend analysis model, and calling 2 GPUs and a data visualization calculation model. These two belong to two different target cluster resources.

[0082] Based on the above introduction, first, look at step S1: Collect at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the corridor combined sensor.

[0083] During the actual operation process, the corridor combined sensor should be in a normal working state most of the time. The size, range, and change trend of the monitored data set should conform to certain normal rules. Therefore, although the corridor combined sensor is in a state of working around the clock without interruption, it is not necessary to store and send the data set collected by each corridor combined sensor at each moment to the backend for processing or display.

[0084] Based on this, each group of monitoring sensors in the corridor combined sensor is configured with an edge data detection unit; the edge data detection unit can perform basic data edge calculation and analysis locally, such as determining whether the data exceeds the preset standard, whether the current change trend of the data is abnormal, etc.

[0085] Step S1 specifically includes: When the edge data analysis unit detects that the park monitoring data set of a certain monitoring sensor is abnormal during the current period, send the abnormal park monitoring data set to the data middleware.

[0086] In a specific embodiment, the main body of the method is implemented by the data middleware. The data middleware is responsible for providing a storage queue to collect abnormal data sets from the monitoring sensors and sending the abnormal data sets to the backend platform. A specific implementation form of the data middleware can be a kafka queue. Of course, other message queues can also be used, such as Rabbit MQ message queues, etc.

[0087] Preferably, to prevent data overflow from causing an exception, the data middleware used in the embodiments of the present application adopts an N-ring queue, where N is an integer greater than or equal to 2. The N-ring queue includes N ring queues, and each ring queue stores the abnormal data sent by a combined sensor correspondingly.

[0088] For the convenience of description, in the subsequent embodiments, N = 2 is taken as an example, that is, the case of a double-ring queue is introduced. It can be understood that the cases of other N>2 can be deduced correspondingly.

[0089] Specifically, the double-ring queue includes a first ring queue and a second ring queue. The data middleware collects at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the corridor combined sensor.

[0090] The reason for collecting at least two groups of park monitoring data sets is that when the first data set collected by a certain group of sensors is abnormal, it means that the data index of a certain pipeline position monitored by this group of sensors may be abnormal; under actual working conditions, the phenomenon of abnormal data indexes in a certain dimension of a pipeline position will not occur alone. That is, when the first data set A1 collected by sensor A1 is abnormal (for example, the temperature is abnormal), it usually also means that the second data set 2A collected by sensor A2 is also abnormal (when the temperature sensors are redundantly configured, that is, two temperature sensors at the same position); or, the third data set B1 collected by sensor B1 is abnormal (for example, the pressure at the same position or adjacent positions is abnormal).

[0091] When the data middleware collects at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the corridor combined sensor, assuming that the two groups of park monitoring data sets include a first park monitoring data set and a second park monitoring data set, the first park monitoring data set is stored in the first ring queue, and the second park monitoring data set is stored in the second ring queue.

[0092] To avoid the data transmission pressure caused by frequent data sending, it is necessary to perform preliminary classification prediction processing on the first park monitoring data set and the second park monitoring data set, that is, enter step S2: use a local prediction model to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results.

[0093] Specifically, refer to Figure 2 , the schematic flow chart of the data collected by the corridor combined sensor entering the local prediction model.

[0094] In Figure 2 , sensor A is taken as an example of the monitoring sensor for introduction.

[0095] First, the monitoring sensor A continuously generates monitoring data {a1, a2,...};

[0096] However, it is determined whether the preset analysis period is reached, for example, whether the data collection continues for 45 seconds; if so, proceed to the next step, otherwise return to the previous step;

[0097] Next, the edge data detection unit determines whether there is an abnormality in the monitoring data {a1, a2,... a N}, and the determination of whether there is an abnormality here can be to determine whether each of the monitoring data {a1, a2,... a N} is abnormal (such as less than the low threshold, greater than the high threshold, abnormal change trend, etc.); if there is an abnormality, only proceed to the next step; otherwise, return to the first step;

[0098] Continuing, if there is an abnormality, the monitoring data {a1, a2,... a N} is sent to the data middleware for storage, for example, stored in the first circular queue of the data middleware;

[0099] The introduction of the monitoring sensor B is similar, and finally, when there is an abnormality in {b1, b2,..., b M}, it is sent to the data middleware for storage, for example, stored in the second circular queue of the data middleware.

[0100] As a further preference, when the first circular queue or the second circular queue is full, the data middleware stores it in the data middleware and at the same time sends it to the local prediction model; this ensures the completeness of the data and avoids data analysis delay;

[0101] The local prediction model is used to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results;

[0102] The use of the local prediction model to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results specifically includes:

[0103] Taking the first park monitoring data set, the length of the first park monitoring data set, and the generation time period of the first park monitoring data set as the input of the local prediction model to obtain the first prediction output result;

[0104] Taking the second park monitoring data set, the length of the second park monitoring data set, and the generation time period of the second park monitoring data set as the input of the local prediction model to obtain the second prediction output result;

[0105] The first prediction output result or the second prediction output result includes: the size of the output result, the display format of the output result, and the delay of the output result.

[0106] The local prediction model herein is a pre-trained preliminary classification model. The training samples for training this preliminary classification model are: the park monitoring data set within the historical time period, the length of the park monitoring data set, the generation time period of the park monitoring data set, and after sending the park monitoring data set, the length of the park monitoring data set, and the generation time period of the park monitoring data set within the historical time period to the back-end control platform, the size of the output result, the display format of the output result, and the delay of the output result output after analysis by the back-end cluster resources.

[0107] Preferably, the local prediction model may be an output result prediction model based on the Xgboost model.

[0108] As an example, assume that in a previous processing process, for a temperature monitoring data set with a length of 90 seconds (S) (assuming the temperature is sampled once every 3 seconds, for a total of 30 temperature values), the maximum value of this temperature monitoring data set is 75 °C, the minimum value is 3 °C, the average value is 48 °C, and the sampling time period is [10:00:00 - 10:01:30]. The output result output after analysis by the back-end cluster resources is:

[0109] Temperature anomaly: It is necessary to output an alarm signal in audio for 10 seconds (S) and display it in a yellow screen, and the processing duration is 3 seconds (S) (that is, the output result delay is 3s).

[0110] Then, when the pre-trained preliminary classification model subsequently receives a temperature monitoring data set with similar distribution values, it can quickly know that the resources that need to be scheduled for further analysis of this temperature monitoring data set in the future include:

[0111] Software: Data trend analysis model;

[0112] Hardware: Support for intensive computing resources (such as CPU computing cards or even NPU (neural network processor) model training cards), as well as support for corresponding graphics card resources (including GPU resource cards) and audio resource cards.

[0113] That is to say, it is possible to enter step S3: Determine the current utility tunnel data analysis mode based on the at least two groups of park monitoring data sets and the at least two groups of predicted output results.

[0114] In the technical solution of this application, different predicted output results mean that different software or hardware resources need to be scheduled in the future, that is, different cluster resource groups are scheduled, so as to determine the current utility tunnel data analysis mode.

[0115] In other words, different utility tunnel data analysis modes correspond to different schedulable cluster resources.

[0116] The data analysis mode of the pipe gallery includes a homomorphism analysis mode, a polymorphism analysis mode, a multi-cluster analysis mode, or a single-cluster analysis mode;

[0117] The cluster resources under the homomorphism analysis mode provide display resource support in the same format;

[0118] The cluster resources under the polymorphism analysis mode provide display resource support in at least two formats;

[0119] The cluster resources under the multi-cluster analysis mode include at least a first cluster and a second cluster, and the first cluster and the second cluster are distributed at different locations.

[0120] In one case, when the display formats of the first prediction output result and the second prediction output result are the same, it is determined that the current data analysis mode of the pipe gallery is the homomorphism analysis mode; otherwise, it is determined that the current data analysis mode of the pipe gallery is the polymorphism analysis mode.

[0121] For example, when the display formats of the first prediction output result and the second prediction output result both require graphics card support, it is determined that the current data analysis mode of the pipe gallery is the first homomorphism analysis mode. In the first homomorphism analysis mode, the graphics card resources in a certain or several cluster resources can be enabled only (of course, other basic resources such as the CPU are also included, but the audio resources can be not activated); when the display formats of the first prediction output result and the second prediction output result both require audio support, it is determined that the current data analysis mode of the pipe gallery is the second homomorphism analysis mode. In the second homomorphism analysis mode, the audio card resources in a certain or several cluster resources can be enabled only (of course, other basic resources such as the CPU are also included, but the graphics card resources can be not activated).

[0122] Similarly, when the display format of the first prediction output result requires graphics card support and the display format of the second prediction output result requires audio support, it is determined that the current data analysis mode of the pipe gallery is the polymorphism analysis mode, that is, the graphics card resources and the audio card resources in a certain or several cluster resources need to be enabled simultaneously.

[0123] In another aspect, when the size or delay of the output result in the first prediction output result or the second prediction output result meets a preset standard, it is determined that the current data analysis mode of the pipe gallery is the multi-cluster analysis mode; otherwise, it is determined that the current data analysis mode of the pipe gallery is the single-cluster analysis mode.

[0124] Continuing the above example (temperature anomaly: an alarm signal needs to be output in audio for 10 seconds (S) and displayed in a yellow screen, and the processing duration is 3 seconds (S) (i.e., the output result delay is 3s).

[0125] Assuming that the safety delay threshold is 5 seconds (s), when the delay of the output result does not meet the preset standard, it is determined that the current pipe gallery data analysis mode is a single cluster analysis mode, that is, there is no need to increase cluster resources;

[0126] On the contrary, if the safety delay threshold is 1 second (s), when the delay of the output result meets the preset standard, the current tunnel data analysis mode is determined to be a multi-cluster analysis mode, that is, cluster resources need to be increased to ensure that the results of the analysis and processing of the current data set are output faster and not delayed to affect safety.

[0127] That is to say, when the delay of the output result is greater than the preset threshold, it is determined that the current corridor data analysis mode is a multi-cluster analysis mode, and the delay of the output result after performing data analysis on the park monitoring data set sent by the data middleware using the target cluster resources corresponding to the multi-cluster analysis mode is less than the preset threshold.

[0128] Similarly, when the size of the output result of the first prediction output result or the second prediction output result meets the preset standard, the current pipeline corridor data analysis mode is determined to be a multi-cluster analysis mode; otherwise, the current pipeline corridor data analysis mode is determined to be a single cluster analysis mode.

[0129] For example, a larger output result will also result in a longer output time (larger delay), which will require more cluster resource support. Therefore, the scheduled corridor data analysis mode is a multi-cluster analysis mode, and the delay in the output result after the target cluster resources corresponding to the multi-cluster analysis mode perform data analysis on the park monitoring data set sent by the data middleware is less than the preset threshold.

[0130] It should be pointed out that the above-mentioned pipeline corridor data analysis modes including homomorphic analysis mode, polymorphic analysis mode, multi-cluster analysis mode or single cluster analysis mode are not completely mutually exclusive.

[0131] In simple terms, the current data analysis mode can be one of the homomorphic analysis mode or the polymorphic analysis mode, or one of the multi-cluster analysis mode or the single-cluster analysis mode.

[0132] However, the homomorphic analysis mode can also be a cluster analysis mode or a single cluster analysis mode, and the polymorphic analysis mode can also be a multi-cluster analysis mode or a single cluster analysis mode; this depends on the attributes of the output results and the analysis needs of the on-site working conditions.

[0133] After the current tunnel data analysis mode is determined, step S4 is entered: based on the current tunnel data analysis mode, an activation instruction is sent to the back-end control platform to enable the back-end control platform to start or switch to the target cluster resources corresponding to the current tunnel data analysis mode.

[0134] It can be seen that for the multi-dimensional data to be analyzed generated by the utility tunnel combined sensor, the analysis mode can be predicted in advance, and the present invention can schedule the matching cluster resources in advance for processing, reducing the result output delay and avoiding resource waste at the same time.

[0135] In Figure 1 - Figure 2 Based on the introduction of the method embodiment, next, the system embodiment will be introduced. Figure 3 The schematic diagram of the functional unit composition of a park data analysis system based on a utility tunnel combined sensor according to an embodiment of the present invention is shown.

[0136] The system includes a data middleware and a backend control platform. The data middleware configures a local prediction model, and the backend control platform is connected to multiple cluster resources;

[0137] Each group of monitoring sensors in the utility tunnel combined sensor is configured with an edge data detection unit. The edge data detection unit performs anomaly detection on the park monitoring data generated by each group of monitoring sensors within the current preset time period. When an anomaly is detected, the park monitoring data set with the anomaly is sent to the data middleware;

[0138] The data middleware collects at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the utility tunnel combined sensor, and performs result output analysis on the at least two groups of park monitoring data sets using the local prediction model to obtain at least two groups of prediction output results;

[0139] Based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results, determine the current utility tunnel data analysis mode;

[0140] Based on the current utility tunnel data analysis mode, send an activation instruction to the backend control platform so that the backend control platform activates or switches to the target cluster resources corresponding to the current utility tunnel data analysis mode;

[0141] Perform data analysis on the park monitoring data set sent by the data middleware using the target cluster resources.

[0142] The utility tunnel combined sensor includes at least two of a temperature combined sensor, a pressure combined sensor, a natural gas pipeline charge combined sensor, and a communication cable pipeline combined sensor.

[0143] Alternatively, the utility tunnel combined sensor includes a natural gas pipeline sensor, a power cable sensor, a drainage pipeline sensor, a communication pipeline sensor, and a heating pipeline sensor;

[0144] The natural gas pipeline sensor is used to detect the charge change around the natural gas pipeline;

[0145] The power cable sensor is used to detect whether the power cable is operating normally;

[0146] The drainage pipe sensor is used to detect various data indicators of water supply and drainage;

[0147] The heating pipe sensor is used to measure the temperature index of the heating pipe.

[0148] The prediction output result includes a delay of the output result; the pipeline corridor data analysis mode includes a multi-cluster analysis mode or a single cluster analysis mode;

[0149] When the delay of the output result is greater than a preset threshold, it is determined that the current corridor data analysis mode is a multi-cluster analysis mode, and the delay of the output result after performing data analysis on the park monitoring data set sent by the data middleware using the target cluster resources corresponding to the multi-cluster analysis mode is less than the preset threshold.

[0150] Figure 4 This is a schematic diagram of the equipment resource combination of different clusters for processing the combined sensor data of the corridor.

[0151] exist Figure 4 , three available clusters of the current system configuration are shown: cluster A, cluster B, and cluster C. It is assumed that at least two of cluster A, cluster B, and cluster C are located in different geographical locations.

[0152] Cluster A has the most abundant resources, including CPU, GPU, audio resources (such as microphone, headphones, voice playback unit, etc.) and graphics card resources (such as display); Cluster B only includes CPU and audio resources, and Cluster C only includes GPU and graphics card resources.

[0153] In homomorphic analysis mode, you may only need to enable cluster B or cluster C, or enable only the CPU and audio resources in cluster A, or enable only the GPU and graphics card resources in cluster A. In multi-device analysis mode, you may need to enable both cluster B and cluster C, or enable all device resources in cluster A, or "enable cluster B and some resources in cluster A at the same time", or "enable cluster C and some resources in cluster A at the same time".

[0154] Correspondingly, the homomorphic analysis mode of "only need to open cluster B or cluster C" is also a "single cluster analysis mode"; "cluster B and cluster C at the same time" is also a "multi-cluster analysis mode".

[0155] In the technical solution of the present invention, first, when the edge data analysis unit detects that there is an abnormality in the park monitoring data set of a certain monitoring sensor in the current period, the park monitoring data set with the abnormality is sent to the data middleware; secondly, the data middleware collects at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the corridor combination sensor, and uses a local prediction model to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results; finally, based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results, the current corridor data analysis mode is determined; based on the current corridor data analysis mode, an activation instruction is sent to the backend control platform so that the backend control platform can activate or switch to the target cluster resources corresponding to the current corridor data analysis mode. Therefore, the present invention can pre-judge the analysis mode of the multi-dimensional data to be analyzed generated by the corridor combination sensor in advance, and then schedule the matching cluster resources for processing, reducing the result output delay and avoiding resource waste at the same time.

[0156] It can be understood that the relevant principles and steps of the system embodiment correspond to those of the method embodiment, so there is no need to repeat the description in detail, and the two can be cited and referred to each other.

[0157] For other technologies, principles, algorithms or models not elaborated in detail in this application, reference can be made to the prior art.

[0158] In the foregoing embodiment part, the present invention provides multiple embodiments, and each embodiment can constitute an independent technical solution and may contribute to the prior art and solve corresponding technical problems. However, it should be noted that different embodiments can be combined with each other without violating logic; at the same time, each embodiment can solve at least one technical problem, but it is not required that each individual embodiment solve multiple or all technical problems.

[0159] The method embodiments and systems of the present invention have been shown and described above. However, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirits of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A park data analysis method based on a combined sensor of a pipe gallery, the method is implemented based on data middleware, and is characterized in that: The method comprises: Collect at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the pipe gallery combination sensor; Using a local prediction model to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results; Determine a current pipe gallery data analysis mode based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results; Based on the current pipe gallery data analysis mode, an activation instruction is sent to the backend control platform so that the backend control platform starts or switches to the target cluster resources corresponding to the current pipe gallery data analysis mode.

2. A park data analysis method based on a pipe gallery combined sensor as claimed in claim 1, characterized in that: Each group of monitoring sensors in the pipe gallery combined sensor is equipped with an edge data detection unit; When the edge data analysis unit detects that a park monitoring data set in a current time period of a certain monitoring sensor has an abnormality, the park monitoring data set with the abnormality is sent to the data middleware.

3. A park data analysis method based on a pipe gallery combined sensor as claimed in claim 1, characterized in that: The two groups of park monitoring data sets include a first park monitoring data set and a second park monitoring data set; The using of the local prediction model to perform result output analysis on the at least two groups of park monitoring data sets to obtain at least two groups of prediction output results specifically includes: Using the first park monitoring data set, the length of the first park monitoring data set, and the generation time period of the first park monitoring data set as inputs of the local prediction model to obtain a first prediction output result; Using the second park monitoring data set, the length of the second park monitoring data set, and the generation time period of the second park monitoring data set as inputs of the local prediction model to obtain a second prediction output result; The first predicted output result or the second predicted output result includes: the size of the output result, the display format of the output result and the delay of the output result.

4. A park data analysis method based on a pipe gallery combined sensor as claimed in claim 3, characterized in that: Based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results, determining the current pipeline corridor data analysis mode specifically includes: When the display format of the first prediction output result is the same as that of the second prediction output result, it is determined that the current pipeline corridor data analysis mode is a homomorphic analysis mode; otherwise, it is determined that the current pipeline corridor data analysis mode is a polymorphic analysis mode.

5. A park data analysis method based on a pipe gallery combined sensor as claimed in claim 3, characterized in that: Based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results, determining the current pipeline corridor data analysis mode specifically includes: When the size of the output result or the delay of the output result in the first prediction output result or the second prediction output result meets the preset standard, the current pipeline corridor data analysis mode is determined to be a multi-cluster analysis mode; otherwise, the current pipeline corridor data analysis mode is determined to be a single cluster analysis mode.

6. A park data analysis method based on a pipe gallery combined sensor as claimed in claim 1, characterized in that: The pipe gallery data analysis mode includes a homomorphic analysis mode, a polymorphic analysis mode, a multi-cluster analysis mode or a single cluster analysis mode; The cluster resources in the homomorphic analysis mode provide display resource support in the same format; The cluster resources in the polymorphic analysis mode provide display resource support in at least two formats; The cluster resources in the multi-cluster analysis mode include at least a first cluster and a second cluster, and the first cluster and the second cluster are distributed in different locations.

7. A park data analysis system based on a combined pipe gallery sensor, the system comprising a data middleware and a backend control platform, characterized in that: The data middleware configures a local prediction model, and the backend control platform connects multiple cluster resources; Each group of monitoring sensors in the pipe gallery combined sensor is equipped with an edge data detection unit, and the edge data detection unit performs anomaly detection on the park monitoring data generated by each group of monitoring sensors within the current preset time period, and when an anomaly is detected, sends the park monitoring data set with the anomaly to the data middleware; The data middleware collects at least two groups of park monitoring data sets sent by at least two groups of monitoring sensors in the pipe gallery combined sensor, uses a local prediction model to perform result output analysis on the at least two groups of park monitoring data sets, and obtains at least two groups of prediction output results; Determine a current pipe gallery data analysis mode based on the at least two groups of park monitoring data sets and the at least two groups of prediction output results; Based on the current pipe gallery data analysis mode, an activation instruction is sent to the backend control platform so that the backend control platform starts or switches to the target cluster resources corresponding to the current pipe gallery data analysis mode; The target cluster resources are used to perform data analysis on the campus monitoring data set sent by the data middleware.

8. A park data analysis system based on a pipe gallery combined sensor as claimed in claim 7, characterized in that: The pipeline gallery combination sensor includes at least two of a temperature combination sensor, a pressure combination sensor, a natural gas pipeline charge combination sensor, and a communication cable pipeline combination sensor.

9. A park data analysis system based on a pipe gallery combined sensor as claimed in claim 7, characterized in that: The prediction output result includes a delay of the output result; the pipeline corridor data analysis mode includes a multi-cluster analysis mode or a single cluster analysis mode; When the delay of the output result is greater than a preset threshold, it is determined that the current corridor data analysis mode is a multi-cluster analysis mode, and the delay of the output result after performing data analysis on the park monitoring data set sent by the data middleware using the target cluster resources corresponding to the multi-cluster analysis mode is less than the preset threshold.

10. A non-volatile computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, executes a campus data analysis method based on a corridor combined sensor as described in any one of claims 1 to 6.

Images