CN-117009906-B - Intelligent sensing system framework for environment monitoring based on machine learning technology

CN117009906BCN 117009906 BCN117009906 BCN 117009906BCN-117009906-B

Abstract

The invention discloses an intelligent sensing system framework for environment monitoring based on a machine learning technology, which comprises a data collection layer, a data processing layer and an information layer, wherein an abnormal event detection and clustering method is designed, and the original sensor measured value is automatically processed by the machine learning technology and is converted into organized information which is easy to understand and access by an end user, so that the problems of an environment monitoring space being not wide, unreliable and intelligent are effectively solved, and an ecological system is better managed and protected.

Inventors

ZHANG DIAN
XIAO ZHENYU
LIU ZIXUAN
ZHOU XIAOSHANG
GAO DAYONG

Assignees

海南大学

Dates

Publication Date: 20260505
Application Date: 20230821

Claims (4)

1. The intelligent sensing system for environment monitoring based on the machine learning technology is characterized by comprising a data collection layer, a data processing layer and an information layer; The system comprises a data collection layer, a data processing layer, a data storage library and a data storage library, wherein the front end of the data collection layer is in communication connection with an observation site for environment monitoring, and the rear end of the data collection layer is in communication connection with the data processing layer; the data processing layer extracts required data segments from the data storage library, and converts the in-situ sensor measured values into organized information which is easy to understand and access by a user through machine learning, abnormal event detection and clustering; the information layer is used for providing a graphical user interface for the terminal user, displaying the abnormal event detected by the in-situ sensor and the reading of the in-situ sensor and providing the inquiry for the terminal user; The data processing layer realizes abnormal event detection and clustering by executing an abnormal event detection flow, an abnormal feature extraction flow, an event construction flow, an event feature extraction flow and an event clustering flow; the abnormal event detection process is used for monitoring abnormal sensor values, and comprises the following steps: step 2.1, establishing a measured value trend model and abnormal classification; Defining a measured value trend model as B (t), wherein B (t) is an array consisting of N measured values B i (t) acquired from a data storage in a recent preset time period, namely: b (t) = { B 1 (t),...,B i (t),...,B N (t) } obtains a new measurement value I (t) from the data store, the new measurement value I (t) is classified according to the following: ; Wherein dist (I (T), B i (T)) represents the distance between the new measurement value I (T) and the known measurement value B i (T), min (dist (I (T), B i (T))) represents the minimum value of the distances between all elements in the new measurement value I (T) and B (T), T (T) represents the distance threshold, I e (1, 2,..once, N), the new measurement value I (T) is considered abnormal if it is 1, and step 2.2 is performed; Step 2.2, calculating a distance Dist (), and carrying out abnormal classification according to a calculation result; ; The abnormal feature extraction process is used for capturing the similarity of the detected abnormality and further clustering the similarity into abnormal events, and in the process, an abnormal feature set is set, and is defined as f: f=[I(t-1)-I(t),I(t),I(t)-I(t+1),d min ,d min -T(t)] The elements in f comprise a difference value between a previous measured value I (T-1) of the sensor and a current measured value I (T) of the sensor, a next measured value I (t+1) of the sensor, a minimum distance d min between the measured value of the sensor and a measured value trend model, and a distance between a minimum distance d min and a distance threshold T (T); the event construction flow groups the detected abnormal phenomena into events by adopting a condensation hierarchical clustering method according to the time information, and the grouping rule is as follows: a. successive anomalies will be combined into a single event; b. When the time interval between the new anomaly and the previous anomaly is smaller than a preset fixed value, the new anomaly value is combined into the same event, otherwise, a new event is created; The event feature extraction flow adopts a K-means clustering method; The event clustering flow is specifically as follows: Let the known event set E= { E 1 ,...,E num1 }, num1 represent the number of events, and the mapping function is Let input set x=e, mapping function Setting a threshold epsilon, wherein N=0 and N max ; and calling an online robust aggregation clustering method to obtain an event packet set { y 1 ,...,y num2 }, wherein a subscript num2 represents the number of event packets.
2. The intelligent sensing system for environmental monitoring based on machine learning techniques of claim 1, wherein the workflow of the data collection layer comprises: step 1.1, predefining a data format and a communication method of each in-situ sensor in a data collection layer interface in advance; Step 1.2, the data collection layer is connected with a plurality of data sources of different observation sites and acquires data input; step 1.3 the data collection layer stores the acquired data in a centralized data repository in a predefined communication method.
3. The intelligent sensing system for environmental monitoring based on machine learning technology according to claim 1, wherein in the abnormal event detection flow, if the new measurement value I (T) is assigned to 0 and classified as a normal value, the steps of updating the measurement value trend model B (T) and updating the distance threshold T (T) are further included: the manner of updating the measured value trend model B (t) is: Step a, randomly selecting B i (t) ∈b (t) step B, letting B i (t) =i (t); The distance threshold T (T) is updated in the following manner: Wherein T inc/dec is a static value for controlling the update rate of the threshold value, T scale is a fixed value; the average value of the first N d min (t) is shown, and d min (t)＝min(dist(I(t),B i (t)));T lower 、T upper is the upper and lower bounds of the threshold, respectively.
4. The intelligent sensing system for environmental monitoring based on machine learning techniques of claim 1 wherein the graphical user interface provided by the information layer for the end user comprises a sensor reading display unit, a parameter setting unit, an abnormal event display unit.

Description

Intelligent sensing system framework for environment monitoring based on machine learning technology Technical field: The invention relates to the field of environment monitoring and machine learning, and designs an intelligent sensing system framework for environment monitoring based on a machine learning technology. The background technology is as follows: Environmental monitoring refers to long-term, systematic, scientific, qualitative and quantitative monitoring of various elements (such as air, water, soil, noise, electromagnetic radiation and the like) in the environment by using modern technological means such as chemistry, physics, biology, medicine, telemetry, remote sensing, computers and the like, so as to know the change condition, change reason and change trend of environmental quality, and provide scientific basis for environmental management, environmental protection and environmental planning. Machine learning is one of the branches of artificial intelligence, which refers to the use of computer algorithms and models that allow a computer system to automatically increase its ability to perform a particular task based on given data and goals, without explicit programming. In short, machine learning is the ability of a computer system to improve its performance and accuracy by constantly learning and optimizing. In modern society, machine learning has been widely used in various fields such as internet of things, automatic driving, medical diagnosis, financial risk assessment, voice recognition, etc., and has become an important driving force for technological development and social progress. Defects in the prior art: (a) It is difficult to widely deploy and collect large spatial data. Due to the problems of cost, energy, maintenance, data transmission capacity and the like, particularly in water environments, deployment is generally limited by quantity, and large-space data are difficult to widely deploy and collect. (B) The reliability is low. Modern environmental sensing techniques cause sensor unreliability due to environmental factors such as sensor drift and bio-agglomeration, and further result in insufficient coverage of space in most waters. The invention comprises the following steps: In order to solve the problems, the invention provides an intelligent sensing system framework for environmental monitoring based on a machine learning technology, which comprises a data collection layer, a data processing layer and an information layer; The system comprises a data collection layer, a data processing layer, a data storage library and a data storage library, wherein the front end of the data collection layer is in communication connection with an observation site for environment monitoring, and the rear end of the data collection layer is in communication connection with the data processing layer; the data processing layer extracts required data segments from the data storage library, and converts the in-situ sensor measured values into organized information which is easy to understand and access by a user through machine learning, abnormal event detection and clustering; the information layer is used for providing a graphical user interface for an end user, displaying abnormal events detected by the in-situ sensor, readings of the in-situ sensor and inquiring by the end user. Further, the workflow of the data collection layer includes: step 1.1, predefining a data format and a communication method of each in-situ sensor in a data collection layer interface in advance; Step 1.2, the data collection layer is connected with a plurality of data sources of different observation sites and acquires data input; step 1.3 the data collection layer stores the acquired data in a centralized data repository in a predefined communication method. Furthermore, the data processing layer realizes abnormal event detection and clustering by executing an abnormal event detection flow, an abnormal feature extraction flow, an event construction flow, an event feature extraction flow and an event clustering flow. Further, the abnormal event detection process is used for monitoring abnormal sensor values, and the process comprises the following steps: step 2.1, establishing a measured value trend model and abnormal classification; Defining a measured value trend model as B (t), wherein B (t) is an array consisting of N measured values B i (t) acquired from a data storage in a recent preset time period, namely: B(t)={B1(t),...,Bi(t),...,BN(t)} Obtaining a new measurement value I (t) from the data storage library, wherein the new measurement value I (t) is classified according to the following steps: Wherein dist (I (T), B i (T)) represents the distance between the new measurement value I (T) and the known measurement value B i (T), min (dist (I (T), B i (T))) represents the minimum value of the distances between all elements in the new measurement value I (T) and B (T), T (T) represents the distance threshold, I