Terminal acquisition system and data acquisition method for judging abnormal sensor data acquisition

Abstract

The present invention discloses a terminal acquisition system and data acquisition method for judging abnormal sensor data acquisition, including: a collector and a centralized controller; the collector is used to obtain valid acquisition values ​​according to the current sensor acquisition data and the sequence data judgment algorithm, and to remove abnormal values ​​and provide sensor warnings; the centralized controller is used to judge whether the current acquisition data is valid based on the effective acquisition value, the prediction model and the sensor values ​​of other acquisition points. The technical solution of the present invention is adopted to ensure the correctness of the sent data, provide data support for the correctness of the platform-side big data, the iteration validity of the artificial intelligence model and the correctness of management decisions, and at the same time have memory and association functions to provide alarms for abnormal terminal acquisition systems.

Description

Technical Field

[0001] The present invention relates to the field of information industrial automation technology, in particular to the field of information communication in the field of heating and the field of industrial automation technology, and in particular to a terminal acquisition system and a data acquisition method for judging abnormal sensor data acquisition. Background Art

[0002] High-power execution equipment is often used in the fields of heating, water supply, chemical industry, thermal power, etc. With the continuous development of Industry 4.0, artificial intelligence continues to drive the technological revolution and upgrade of various industries. Compared with traditional industrial sites, various sensors need to be installed continuously. The signal acquisition of sensors is generally weak. In addition, the limitations of the site and the non-standardization of the construction party often cause the sensors to receive strong electrical interference, resulting in abnormal data collected by the sensors. If the abnormal value collection value is sent to the platform at this time, it will not only increase the platform data cleaning intensity notification, cause abnormal iteration of the artificial intelligence model, and output the wrong model, but also cause wrong decisions by the management, and even cause accidents; in addition, traditional sensors now only play the role of current data acquisition, and there is no memory association function to determine whether the system is abnormal. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a terminal acquisition system and a data acquisition method for judging abnormalities in sensor data acquisition, ensuring the correctness of the sent data, providing data support for the correctness of the big data on the platform side, the effectiveness of artificial intelligence model iteration and the correctness of management decisions, and at the same time having memory and association functions to provide alarms for abnormalities in the terminal acquisition system.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A terminal acquisition system for judging abnormal sensor data acquisition includes: a collector and a centralized controller; the collector is used to obtain valid acquisition values ​​according to the current sensor acquisition data and a sequence data judgment algorithm, and to remove abnormal values ​​and issue sensor warnings; the centralized controller is used to judge whether the current acquisition data is valid based on the valid acquisition value, a prediction model and sensor values ​​of other acquisition points.

[0006] Preferably, the collector is used to determine the effective value of the sensor's current collection sequence data through the first-order partial derivative, mean square error and deviation calculation of discrete data to obtain the effective collection value, and re-determine its final effectiveness and sensor abnormality judgment warning based on the historical data change rate.

[0007] Preferably, the centralized controller predicts the sensor collection value of the upload point based on the monitoring parameters of other collection points based on the prediction model, and determines whether it is valid data based on the comparison between the actual value and the predicted calculated value; wherein the prediction model is obtained by self-learning based on the LSTM-seq2seq neural network through the actual site historical collection data.

[0008] Preferably, the sensors include: temperature sensors, pressure sensors, motor current sensors, pipeline flow sensors, etc.

[0009] The present invention also provides a data collection method, comprising the following steps:

[0010] Step S1: The collector collects data currently collected by the sensor, obtains valid collected values ​​according to the sequence data judgment algorithm, removes abnormal values ​​and issues sensor warnings;

[0011] Step S2: The centralized controller determines whether the current collected data is valid based on the effective collected value, the prediction model and the sensor values ​​of other collection points.

[0012] Preferably, in step S1, the collector determines the effective value of the current collected sequence data of the sensor by calculating the first-order partial derivative, mean square error and deviation of discrete data to obtain the effective collected value, and re-judges its final effectiveness and sensor abnormality judgment warning based on the historical data change rate.

[0013] Preferably, in step S2, the centralized controller predicts the sensor collection value of the upload point through the monitoring parameters of other collection points based on the prediction model, and determines whether it is valid data based on the comparison between the actual value and the predicted calculated value; wherein the prediction model is obtained by self-learning based on the LSTM-seq2seq neural network through the actual site historical collection data.

[0014] Preferably, the sensors include: a temperature sensor, a pressure sensor, a motor current sensor, and a pipeline flow sensor.

[0015] The present invention ensures the correctness of the sent data through double screening, that is, the collector and the centralized controller screen the data, thereby providing data support for the correctness of the big data on the platform side, the effectiveness of the iteration of the artificial intelligence model and the correctness of the management decision. It also has memory and association functions and provides alarms for abnormalities in the terminal collection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0017] Figure 1 A schematic diagram of the structure of a terminal acquisition system for determining abnormal sensor data acquisition according to an embodiment of the present invention;

[0018] Figure 2 It is a flow chart for data determination of the collector;

[0019] Figure 3 Flowchart for centralized controller data determination. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Embodiment 1:

[0023] like Figure 1 As shown, an embodiment of the present invention provides a terminal acquisition system for determining abnormal sensor data acquisition, including: a collector and a centralized controller; wherein,

[0024] The collector is used to collect data from the sensor, obtain valid collected values ​​based on the sequence data judgment algorithm, remove abnormal values ​​and issue sensor warnings;

[0025] The centralized controller is used to determine whether the current collected data is valid based on the effective collected values, the prediction model and the sensor values ​​of other collection points.

[0026] As an implementation method of an embodiment of the present invention, the collector includes: multiple sensors, an acquisition controller and a first communication module; the sensors include: temperature sensors, pressure sensors, motor current sensors, pipeline flow sensors, which are used to detect parameters of each operating point; the first communication module sends the sensor detection data to the centralized controller in a wired or wireless manner; the collector controller is used to determine the effective value of the current acquisition sequence data of the sensor through the first-order partial derivative, mean square error and deviation calculation of discrete data to obtain the effective acquisition value, and re-determine its final effectiveness and sensor abnormality judgment warning based on the historical data change rate; the acquisition controller includes: a preprocessing module, a data acquisition module, a data processing module, a data screening module and a data abnormality judgment module. Among them,

[0027] The preprocessing module is used to determine the data beyond the collection range as abnormal data through the abnormal data elimination algorithm and eliminate the data;

[0028] The data acquisition module is used to collect data according to the set acquisition frequency and the sampling rate of 10 times the acquisition frequency, and form a vector table of the basic data of this data;

[0029] The data processing module is used to discretize and derive the basic data to form a derivative vector table; sort the derivative vector table, remove the three largest values ​​to form a new derivative vector table; and perform mean value processing and mean square error processing on the new derivative vector table;

[0030] The data screening module is used to make a threshold judgment based on the mean square error. If it is greater than the threshold, the data samples collected this time are too discrete and are interference data, so they are discarded. If they are within the threshold, they are reversed based on the three largest derivative values ​​to remove abnormal data in the basic data vector table and perform mean processing at the same time.

[0031] The data anomaly judgment module is used to calculate the system abnormal state reference value based on the absolute deviation / collection cycle of the current collection value and the last collection value; if the reference value is greater than the threshold, the system is judged to be abnormal.

[0032] The data sending module is used to send the valid collected values ​​of this collection to the centralized controller.

[0033] Further, if Figure 2 As shown, the data determination steps of the collector are as follows:

[0034] 11. Register the collector;

[0035] 12. Timeout to determine whether to obtain the collector's configuration parameters for the collection point, that is, limited range X min , X max ; Discrete rate threshold X th d ; Collection frequency FC ; Effective data jump rate R th d ;

[0036] 13. If obtained, save the configuration parameters;

[0037] 14. If not obtained, read the locally saved configuration parameters;

[0038] 15. Calculate the equipment sampling frequency according to the data collection frequency requirements;

[0039] F S =F C *10

[0040] 16. Determine whether it is time to sample;

[0041] 17. Collect data X i ;

[0042] 18. Determine whether this collection is in an effective atmosphere

[0043] T min <X i <X max

[0044] 19. If it is not within the range, re-collect, and the cumulative number of errors is +1. If the cumulative number of errors reaches 10 times, the sensor is considered abnormal and the platform sensor error code 01 is sent;

[0045] 110. If within a finite range, then X i Stored in vector X, the sampling interval is stored in T i Store in vector T;

[0046] 111. Determine whether the number of samples is >= 10;

[0047] 112. If the vector X is discretely differentiated to obtain X';

[0048] 113. Sort X' by size and remove the three largest elements to obtain XI';

[0049] 114. Find the mean of vector XI'

[0050]

[0051] 115. Calculate the mean square error of the vector XI';

[0052]

[0053] 116. Determine whether XI > discrete threshold X th d ;

[0054] 117. If abnormal, send the central controller sensor abnormality 02;

[0055] 118. Reversely search for abnormal jump values ​​in vector X according to the three maximum values ​​in vector X';

[0056] 119. Average the remaining values

[0057] 120. Save this data;

[0058] 121. Calculate the acquisition point jump rate R, X N Indicates the value collected this time, X L Indicates the last collected value, and T indicates the interval between two valid values

[0059]

[0060] 122. Determine whether the jump rate is greater than the jump threshold R th d ;

[0061] 123. If it is greater than the threshold, it means that the collection point system is abnormal or the sensor is under-calibrated. The sensor state is set to 03 and sent to the centralized controller;

[0062] 124. If normal, set the sensor status to 00 and send to the centralized controller.

[0063] As an implementation method of an embodiment of the present invention, the centralized controller includes: a collection module, a second communication module and an edge computing controller. The collection module is used to communicate with the collector to obtain the effective collection value of each collection point; the second communication module is used to communicate with the platform, upload local data, and receive platform instructions; the edge computing controller is used to determine the sensor collection data according to the effective collection value, specifically including:

[0064] Establish a database locally and integrate the data from each collection point of the collector, send it to the platform and store it locally; integrate the prediction model of each collection point based on deep learning locally;

[0065] According to the data received from the collection point, the corresponding prediction model is triggered, such as the secondary supply temperature of the heat exchange station of the heating system. The latest data of other collection points except the collection point in the database are read and input into the model to generate the corresponding collection point such as the secondary supply temperature of the heat exchange station of the heating system. The calculation temperature and dynamic upper and lower limits; if the value is greater than the range value of the system where the point is located, it is judged that the uploaded data is abnormal and the sensor is faulty;

[0066] Calculate the absolute deviation between the temperature collection point and the actual collection value, compare it with the set threshold of the point, and determine whether the sensor is abnormal; if normal, store the data and send it to the platform;

[0067] The absolute deviation / collection cycle is calculated based on the last saved collection value as a reference value for the system abnormal state; if the reference value is greater than the threshold, the system is judged to be abnormal and a system warning is sent to the collection point.

[0068] When the centralized controller receives a data sent by a sensor, it calls the corresponding prediction model according to the type of data sent. The prediction model is obtained through self-learning based on the LSTM-seq2seq neural network through historical data (such as the historical data of this heat exchange station). The prediction model predicts the sensor collection value of the upload point through the monitoring parameters of other collection points (other collection points in the heat exchange station of the heating system), and judges whether it is valid data based on the comparison between the actual value and the predicted calculated value. For example, when the centralized controller receives the primary side heating data, it will call the prediction model of the primary side heating to predict the data and determine whether it is within the normal range. If it exceeds the set threshold, it indicates an abnormality, such as Figure 3 As shown, the data determination of the centralized controller includes the following steps:

[0069] 21. The centralized controller first registers on the platform to obtain the configuration of the terminal acquisition system;

[0070] 22. Determine the type of collector device that actively reports registration and issue the corresponding configuration parameters;

[0071] 23. The platform queries the prediction model for each acquisition parameter;

[0072] 24. If there is an update, update the prediction model;

[0073] 25. Receive the effective collection value T of the collector N ;

[0074] 26. Determine the detection terminal and measurement point data;

[0075] 27. Read the values ​​of other sensors in the system in addition to the values ​​uploaded this time, such as R (secondary pressure measurement), S (secondary temperature measurement), P (primary pressure), W (motor power consumption), and Q (primary flow rate) in the database;

[0076] 28. Input R, S, P, W, and Q into the prediction model with the acquisition parameter T (primary side temperature) to obtain the effective range T of parameter T. min , T max and the evaluation value T c ;

[0077] 29. Judge T N Is the value of Tmin , T max within the scope;

[0078] 210. If it is not within the range, the data is determined to be abnormal, the sensor is determined to be abnormal, the sensor state is set to 05, and the abnormal data is discarded;

[0079] 211. If within the range, calculate the deviation:

[0080] ΔT=|T N -T c |

[0081] 212. Determine whether ΔT exceeds T th d Set thresholds;

[0082] 213. If the threshold is exceeded, the sensor or system collection point is judged to be abnormal and the sensor state is set to 06;

[0083] 214. If normal, set the sensor status to 00;

[0084] 215. Calculate the data change rate ΔT of the system at this collection point D , ΔTim represents the time difference between the current acquisition time and the last saved number, T l Indicates the last saved value;

[0085]

[0086] 216. Calculate the data variation rate ΔT of this collection point D Is it greater than the set threshold? If so, the site system abnormality is set to T01;

[0087] 217. If normal, the system is T00;

[0088] 218. The local database saves and sends the system and sensor status values ​​to the platform.

[0089] The embodiment of the present invention can effectively ensure the correctness of the data sent by the acquisition system, thereby ensuring the validity of the big data on the platform side, providing data support for the iteration of the artificial intelligence model and the correctness of management decisions, and at the same time being able to judge in real time whether the system is operating reliably, so as to achieve an early warning effect. Double screening is performed in the terminal acquisition system, that is, the data is screened and judged in the collector and the centralized controller. The collector obtains the effective acquisition value according to the current data collected by the sensor; the centralized controller takes the return temperature, pressure, flow rate of the primary side pipe network, the return temperature, pressure of the secondary side pipe network, the frequency and current of the water supply pump, and the frequency and current of the circulation pump as the data source based on the effective acquisition value, taking the heat exchange station as an example, and learns through the LSTM-seq2seq neural network prediction model to obtain the prediction model of each acquisition parameter. When the sensor uploads a collection data, the prediction model will be triggered. Taking the secondary side heating as an example, the current latest data is brought into the secondary side heating prediction model to output a prediction value. When the prediction value exceeds the set threshold with the actual uploaded value, it means that the secondary side heating uploaded this time is abnormal.

[0090] Compared with the prior art, it has the following beneficial effects:

[0091] 1. Double screening to ensure the correctness of uploaded data;

[0092] 2. Sensor abnormality self-diagnosis function;

[0093] 3. Self-diagnosis function of system collection points;

[0094] 4. The collected data is processed based on the derivation method, which is applicable not only to the inert system but also to the fast response system;

[0095] 5. Remove abnormal values ​​that cause fluctuations based on the derivation method to make the uploaded data more accurate.

[0096] Embodiment 2:

[0097] The embodiment of the present invention also provides a data collection method, comprising the following steps:

[0098] Step S1: The collector collects data currently collected by the sensor, obtains valid collected values ​​according to the sequence data judgment algorithm, removes abnormal values ​​and issues sensor warnings;

[0099] Step S2: The centralized controller determines whether the current collected data is valid based on the effective collected value, the prediction model and the sensor values ​​of other collection points.

[0100] As an implementation mode of an embodiment of the present invention, in step S1, the collector performs effective value judgment on the current collected sequence data of the sensor through the first-order partial derivative, mean square error and deviation calculation of discrete data to obtain the effective collection value, and re-judges its final validity and sensor abnormality judgment warning based on the historical data change rate.

[0101] As an implementation mode of an embodiment of the present invention, in step S2, the centralized controller predicts the sensor collection value of the upload point based on the prediction model through the monitoring parameters of other collection points, and determines whether it is valid data based on the comparison between the actual value and the predicted calculated value; wherein the prediction model is obtained by self-learning based on the LSTM-seq2seq neural network through the actual site historical collection data.

[0102] As an implementation mode of the embodiment of the present invention, the sensor includes: a temperature sensor, a pressure sensor, a motor current sensor, and a pipeline flow sensor.

[0103] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A terminal acquisition system for determining abnormal sensor data acquisition, characterized in that: include: Collector and centralized controller; The collector is used to collect data from the sensor, obtain valid collected values ​​based on the sequence data judgment algorithm, remove abnormal values ​​and issue sensor warnings; The centralized controller is used to determine whether the current collected data is valid based on the effective collected values, the prediction model and the sensor values ​​of other collection points; The collector is used to determine the effective value of the sensor's current collection sequence data through the first-order partial derivative, mean square error and deviation calculation of discrete data, obtain the effective collection value, and re-determine its final effectiveness and sensor abnormality judgment warning based on the historical data change rate; The centralized controller predicts the sensor collection value of the upload point based on the prediction model through the monitoring parameters of other collection points, and determines whether it is valid data based on the comparison between the actual value and the predicted calculated value; the prediction model is obtained through self-learning of the LSTM-seq2seq neural network based on the actual site historical collection data; Sensors include: temperature sensor, pressure sensor, motor current sensor, pipeline flow sensor, etc.

2. A data collection method, characterized in that: The following steps are involved: Step S1, using the collector to collect data currently collected by the sensor, and using the sequence data judgment algorithm to obtain valid collection values ​​and remove abnormal values ​​and issue sensor warnings; Step S2: The centralized controller determines whether the current collected data is valid based on the effective collected values, the prediction model and the sensor values ​​of other collection points; In step S1, the collector determines the effective value of the current sensor collection sequence data by calculating the first-order partial derivative, mean square error and deviation of discrete data to obtain the effective collection value, and re-determines its final effectiveness and sensor abnormality judgment warning according to the historical data change rate; In step S2, the centralized controller predicts the sensor collection value of the upload point based on the monitoring parameters of other collection points based on the prediction model, and determines whether it is valid data based on the comparison between the actual value and the predicted calculated value; wherein the prediction model is obtained by self-learning based on the LSTM-seq2seq neural network through the actual site historical collection data; The sensors include: temperature sensor, pressure sensor, motor current sensor, and pipeline flow sensor.

Images