WO2020010701A1 - Pollutant anomaly monitoring method and system, computer device, and storage medium - Google Patents

Abstract

A pollutant anomaly monitoring method and system, a computer device, and a storage medium, relating to the technical field of environmental pollution data processing. The monitoring method comprises: obtaining historical environment monitoring data and setting same as a data set for storage; setting index thresholds for index items of a pollutant, screening the data set according to the index thresholds to obtain data not exceeding the index thresholds, setting said data as feature items and storing the feature items in a data set not exceeding a standard; training an isolation forest model by using the feature items; and obtaining real-time monitoring data, inputting the real-time monitoring data into the isolation forest model, and determining, according to path lengths of the real-time monitoring data from a root node to leaf nodes of the isolation forest model, whether said data are abnormal points, and summarizing the abnormal points. Abnormal point data can be monitored simply and quickly, and the situation that a pollution source exceeds the standard can be predicted in advance and output, so that the situation of exceeding the standard is prevented.

Description

Pollutant abnormality monitoring method, system, computer equipment and storage medium

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on July 11, 2018, with application number 201810757268.8, and the invention name is "Contaminant Anomaly Monitoring Method, System, Computer Equipment, and Storage Medium". Citations are incorporated in this application.

Technical field

The present application relates to the technical field of environmental pollution data processing, and in particular, to a method, a system, a computer device, and a storage medium for monitoring anomalies of pollutants.

Background technique

A pollution source is a source of pollutants that causes environmental pollution. It usually refers to a place, equipment, device, or human body that emits harmful substances to the environment or has a harmful effect on the environment. Any substance or energy that enters the environmental system at an inappropriate concentration, quantity, speed, form, and path and causes pollution or damage to the environment is collectively referred to as a pollutant. In some links in industrial production, such as production equipment or production sites used in raw material production, processing, combustion, heating and cooling, and finishing of finished products, they can become sources of industrial pollution. In the prior art, there are generally two methods for monitoring the emissions of pollution sources. The first is supervisory monitoring, which periodically checks whether the content of harmful substances in the exhaust gas emitted by pollution sources meets national regulations. The second is research-based monitoring, which monitors the types, emissions, and discharge laws of pollutants emitted by pollution sources, which helps to identify the main sources of air pollution, discuss the development trend of air pollution, formulate pollution control measures, and improve ambient air quality.

However, no matter what kind of detection method, currently, the threshold of the single pollution source is set, and compared with the detection data of the single pollution source, it is found that the pollutant discharge enterprises exceed the standard. However, there are problems in that the types of pollution sources are complicated, the setting and checking of single thresholds are cumbersome, and the setting of thresholds cannot prevent the occurrence of excessive standards.

Summary of the invention

In view of this, it is necessary to set the check threshold for the compliance threshold of a single pollution source, which is tedious, and cannot prevent the problem of exceeding the standard, and provide a method, system, computer equipment and storage medium for monitoring the abnormality of pollutants.

A method for monitoring anomalies of pollutants includes the following steps:

Acquiring historical environmental monitoring data of a pollution source monitoring point of each enterprise from a preset environmental monitoring data system, and storing the historical environmental monitoring data with one pollutant of each enterprise as a data set for storage;

An index threshold is set for each index item of each of the pollutants, and each of the data sets is filtered according to the index threshold, and data that does not exceed the index threshold is set as a feature item, and the feature is set. Items are stored in non-exceeded data sets;

Use the feature terms in the non-exceeded data set to train an isolated forest model, and establish the corresponding isolated forest model for each non-exceeded data set;

Obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, input the real-time monitoring data into the isolated forest model corresponding to the pollutant, and pass the real-time The monitoring data reaches the path length from the root node to the leaf node of the isolated forest model to determine whether it is an abnormal point, and the abnormal points are summarized.

A pollutant abnormality monitoring system includes the following units:

The acquiring data unit is configured to acquire historical environmental monitoring data of a pollution source monitoring point of each enterprise from a preset environmental monitoring data system, and set the historical environmental monitoring data to one data of one pollutant per enterprise Set for storage;

The screening unit is configured to set an index threshold for each index item of each of the pollutants, filter each of the data sets according to the index threshold value, and filter out data that does not exceed the index threshold value as a feature item , Storing the feature items in a non-exceeding data set;

A training unit configured to train an isolated forest model by using the feature terms in the non-exceeded data set, and establish the corresponding isolated forest model for each non-exceeded data set;

The abnormal point summary unit is configured to obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, and input the real-time monitoring data into the isolated forest corresponding to the pollutant In the model, it is determined whether the path length from the root node to the leaf node of the isolated forest model is an abnormal point through the real-time monitoring data, and the abnormal points are summarized.

A computer device includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor causes the processor to perform the following steps:

Acquiring historical environmental monitoring data of a pollution source monitoring point of each enterprise from a preset environmental monitoring data system, and storing the historical environmental monitoring data with one pollutant of each enterprise as a data set for storage;

An index threshold is set for each index item of each of the pollutants, and each of the data sets is filtered according to the index threshold, and data that does not exceed the index threshold is set as a feature item, and the feature is set. Items are stored in non-exceeded data sets;

Use the feature terms in the non-exceeded data set to train an isolated forest model, and establish the corresponding isolated forest model for each non-exceeded data set;

Obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, input the real-time monitoring data into the isolated forest model corresponding to the pollutant, and pass the real-time The monitoring data reaches the path length from the root node to the leaf node of the isolated forest model to determine whether it is an abnormal point, and the abnormal points are summarized.

A storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the following steps:

Acquiring historical environmental monitoring data of a pollution source monitoring point of each enterprise from a preset environmental monitoring data system, and storing the historical environmental monitoring data with one pollutant of each enterprise as a data set for storage;

An index threshold is set for each index item of each of the pollutants, and each of the data sets is filtered according to the index threshold, and data that does not exceed the index threshold is set as a feature item, and the feature is set. Items are stored in non-exceeded data sets;

Use the feature terms in the non-exceeded data set to train an isolated forest model, and establish the corresponding isolated forest model for each non-exceeded data set;

Obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, input the real-time monitoring data into the isolated forest model corresponding to the pollutant, and pass the real-time The monitoring data reaches the path length from the root node to the leaf node of the isolated forest model to determine whether it is an abnormal point, and the abnormal points are summarized.

The above-mentioned pollutant abnormality monitoring method, device, computer equipment and storage medium include obtaining historical environmental monitoring data of each enterprise's pollution source monitoring point from a preset environmental monitoring data system, and converting historical environmental monitoring data to Each pollutant is set as a data set for storage; each indicator of each pollutant is set with an index threshold, each data set is filtered according to the index threshold, and the data that does not exceed the index threshold is set as a feature. Store the feature items in the non-exceeding data set; use the feature items in the non-exceeding data set to train an isolated forest model, and establish a corresponding isolated forest model for each non-exceeding data set; obtain an enterprise-in-one The real-time monitoring data of the pollutants are measured in hours. The real-time monitoring data is input into the isolated forest model corresponding to the pollutants, and the path length of the root node to the leaf node of the isolated forest model is determined by the real-time monitoring data to determine whether it is an abnormal point. The abnormal points are summarized. This application screens historical data and selects non-exceeding pollutant data as feature items, and uses an isolated forest model to monitor abnormal points of the company's sewage. The monitoring of the abnormal point data is simple and fast, and it can predict the pollution source exceeding the standard in advance. And output, to prevent the occurrence of excessive standards.

BRIEF DESCRIPTION OF THE DRAWINGS

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the detailed description of the preferred embodiments below. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present application.

FIG. 1 is a flowchart of a pollutant abnormality monitoring method in an embodiment of the present application; FIG.

FIG. 2 is a flowchart of step S3 in FIG. 1;

3 is a structural diagram of a tree constructed in step S3;

FIG. 4 is a flowchart of step S4 in FIG. 1;

5 is a structural diagram of a pollutant abnormality monitoring system in an embodiment of the present application;

FIG. 6 is a schematic block diagram of the abnormal point summary unit in FIG. 5.

detailed description

In order to make the purpose, technical solution, and advantages of the present application clearer, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.

Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms "a", "an", "the" and "the" may include plural forms. It should be further understood that the word "comprising" used in the specification of the present application refers to the presence of the described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, Integers, steps, operations, elements, components, and / or groups thereof.

FIG. 1 is a flowchart of a method for monitoring abnormal pollutants in an embodiment of the present application. As shown in FIG. 1, the monitoring method includes the following steps:

Step S1: Obtaining data: Obtain historical environmental monitoring data of each enterprise's pollution source monitoring point from a preset environmental monitoring data system, and set historical environmental monitoring data with one pollutant of each enterprise as a data set. storage.

The abnormality monitoring of pollutants in this step is mainly for the pollutants discharged by enterprises. Therefore, the enterprises in this embodiment are the key pollutant discharge units included in the pollution source monitoring center of the Ministry of Ecology and Environment. The preset environmental monitoring data system is provided by the government environmental protection department. The automatic monitoring work scheduling platform of the pollution source monitoring center of the Ministry of Ecology and Environment, or a third-party environmental monitoring data system. The environmental monitoring data system collected historical environmental monitoring data and real-time monitoring data from all pollution source monitoring points of each key sewage unit.

The pollution source monitoring points of the enterprise are generally set at the drainage outlet and the exhaust outlet. Therefore, the pollutants of the enterprise include the drainage pollutants based on the drainage outlet monitoring and the exhaust pollutants based on the exhaust outlet monitoring. When stored, the data set is classified and stored according to the drainage pollutant data set and the exhaust pollutant data set.

Step S2, screening data: setting index thresholds for each index item of each pollutant, filtering each data set according to the index thresholds, filtering out data that does not exceed the index threshold value, setting them as feature items, and storing the feature items in the Out-of-standard data set.

Before filtering the data in the data set, a corresponding index threshold is set for each index item of the pollutant. Among them, the index items of drainage pollutants include at least one of suspended solids index, chemical oxygen demand index, pH value or ammonia nitrogen index, and corresponding index thresholds are set for the index items of drainage pollutants. The exhaust pollutant index items include at least one of nitrogen oxide index, sulfur dioxide index, soot index, or carbon monoxide index, and corresponding exhaust thresholds are set for the exhaust pollutant index items. In this step, the index items of pollutants and the corresponding index thresholds are shown in Table 1 below:

Table 1

When setting thresholds for pollutants, refer to the "Integrated Emission Standard for Air Pollutants" (GB 16297-1996) for exhaust pollutants, and refer to the "Integrated Sewage Emission Standards" (GB8978-1996) for drainage pollutants in order In order to better meet the national standards for pollutant discharge, the sieve finds out more accurate non-standard data.

In this step, each data set is filtered according to a preset index threshold, and the data that does not exceed the threshold is screened out and stored as a feature item in a non-exceeding data set. When the feature items are stored in the non-exceeded data set, the feature items are classified with one pollutant per enterprise and stored in the corresponding non-exceeded data set. That is, the feature items selected by the drainage pollutant data set of each enterprise are stored as the drainage non-standard data set, and the feature items selected by the exhaust pollutant data set of each enterprise are stored as the exhaust non-standard data set. In order to facilitate the subsequent training of the isolated forest model, it can be used as a sample of a certain pollutant in the enterprise.

Step S3, training the model: using the feature terms in the non-exceeding data set to train an isolated forest model, and establishing a corresponding isolated forest model for each non-exceeding data set.

The isolated forest model, the Isolation Forest model, is a fast anomaly detection method with linear time complexity and high accuracy, and is an algorithm that meets the requirements of big data processing. The isolated forest model is suitable for continuous data anomaly detection. The anomaly is defined as “outliers that are easy to be isolated”, which can be understood as the points that are sparsely distributed and far from the densely populated group. In terms of statistics, that is, in the data space, a sparsely distributed area indicates that the probability of data occurring in this area is very low, so the data falling in these areas can be considered abnormal.

Therefore, the isolated forest model is based on the above principles and builds a binary tree from the samples: the input training data set A, e is the current tree height, and l is the height limit of the tree. First place A in the root node, randomly select a dimension q in A, and randomly choose a value p between the maximum and minimum values on q, and flow the sample in A that is larger than p to q to the right. In the node, samples smaller than p flow to the left child node. Then repeat the above steps until: each child node has only one sample or multiple identical samples, that is, each sample is isolated, or the height of the tree reaches l. When using the above method to build a tree, the abnormal point is more easily isolated, so the path length of the leaf node where it is isolated is also shorter, that is, the number of edges experienced by the root node to the leaf node where the abnormal point is located is shorter. The normal point is not easy to be isolated, so its path length is also longer.

Step S4: Summarization of abnormal points: Obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, enter the real-time monitoring data into an isolated forest model corresponding to the pollutants, and achieve The path length from the root node to the leaf node of the isolated forest model determines whether it is an abnormal point, and summarizes the abnormal points.

This step is based on the feature that the outliers in the isolated forest model are easier to isolate, and the number of edges experienced by the outlier from the root node to the leaf node where the outlier is located is short. By inputting real-time monitoring data into the isolated forest model, Obtain short data on the side, that is, some data with short path length are set as abnormal points, and summarize these abnormal points to realize the monitoring of abnormal points of pollution sources.

In this embodiment, a more accurate isolated forest model is trained by obtaining historical environmental monitoring data from a preset environmental monitoring data system, and the isolated forest model is used to monitor real-time monitoring data, and abnormal points are monitored and summarized. The process monitoring data is simple and fast, and can accurately monitor the excessive situation of various pollution sources in the enterprise and summarize them to prevent the occurrence of excessive standards.

In one embodiment, in step S3, when the isolated forest model is trained using the feature terms in the non-exceeded data set, as shown in FIG. 2, the following method is adopted:

In step S301, a point structure feature is obtained: a feature item in every non-standard data set for every N hours is set as a point structure feature. Since there may be more data in the non-exceeded data set, in order to reduce the use of data, take the feature items in units of the non-exceeded data set for training, that is, you can select the hourly feature items or the feature items every 2 hours to set as Point construction features. These point construction features are put into the root node of the tree of the isolated forest model.

In step S302, a difference threshold is set: a difference between each point structure feature and a previous point structure feature is set to a difference threshold of X. The difference threshold can be randomly generated as the cutting point of the current node.

In step S303, the left and right child nodes of the tree are constructed: the difference between the structural features of two adjacent points is less than X and is divided into the left child node of the tree, and the difference between X and X is divided into the right child node of the tree.

Step S304, recursively construct the tree: recursively, steps S302 and S303, continuously construct left child nodes and right child nodes until the following conditions are met: the training non-exceeded data set has only one record or multiple identical records, or the height of the tree reaches a pre-set Set the height range. The height range of the tree is: in a non-exceeded data set containing n records, the minimum height of the constructed tree is log (n), and the maximum height of the constructed tree is n-1.

Specifically, as shown in FIG. 3, for example, four points a, b, c, and d are constructed into an isolated forest model. When traversing a tree, the four points a, b, c, and d are constructed first. The root node of the tree into the isolated forest model. The difference between the three point structure features a, b, and c and the previous point structure feature is less than X, and is divided into the left child node of the tree. The difference between the point d structure feature and the previous point structure feature is greater than X, and To the right child node of the tree. The above steps are recursively, and the left child node and the right child node are continuously constructed. As shown in FIG. 3, after three recursions, each leaf node has only one record. It can be seen that the structural feature at point d was isolated at the earliest, so the structural feature at point d is most likely an anomaly.

In this embodiment, when training an isolated forest model on the feature items in the non-standard data set, the data in the non-standard data set is first filtered to ensure that the trained isolated forest model is as accurate as possible to reduce the data collection amount. When setting the cut point, the difference threshold is used to effectively reflect the changes between the structural features of two adjacent points every N hours. The isolated forest model finally obtained can be more reliable as an abnormal point monitoring.

In one embodiment, in step S4, the real-time monitoring data is input into the isolated forest model corresponding to the pollutants, and the path length of the root node to the leaf node of the isolated forest model is determined by the real-time monitoring data to determine whether it is an abnormal point, such as As shown in Figure 4, the following steps are taken:

Step S401, generating a path length: real-time data in the real-time monitoring data is input into the corresponding isolated forest model one by one, and the real-time data is divided into M times according to the isolated forest model and is no longer divided, the real-time data is at the root node of the isolated forest model The path length to the leaf node is M.

Specifically, as shown in FIG. 3, if a, b, c, and d are the four real-time data in the real-time monitoring data, the path length of the structural feature at point a is 2, and the path length of the structural feature at point b and c is 3 The path length of the structural feature at point d is 1.

In step S402, the normalization process is performed on the path length M of the real-time data to obtain M '.

When normalizing, the following methods are used:

First calculate the average path of the point x to be judged on all the trees, and record it as E (h (x)), E () represents the average, and h (x) represents the path length of x. Assume that there are n points in the real-time monitoring data. If the n points are searched with a tree in an isolated forest model, the average path length of the n points is c (n) = 2H (n-1)-(2 (n-1) / n). H (k) = ln (k) + ξ, ξ = 0.5772156649 is Euler's constant, and the normalized s (x, n) is:

Where M = h (x) and M '= s (x, n);

The range of s (x, n) is [0,1].

In step S403, the abnormal point summary is preset: the abnormal point threshold value Y is preset, and when M 'is greater than Y, the real-time data is set as the abnormal point, and summarized to generate an abnormal point summary table.

Since the normalized M 'value range is [0,1], when judging an abnormal point, the closer M' is to 1, the higher the probability that this real-time data is an abnormal point, and the closer M 'is to 0, it means that the real-time data is The possibility of normal points is high. If the real-time data in the real-time monitoring data are close to 0.5, there are no obvious abnormal points in the entire real-time monitoring data. Therefore, the range of the abnormal point threshold Y should be greater than 0.5 and close to 1.

In this embodiment, when determining an abnormal point, a normalization method is introduced to normalize the path length of a certain real-time data, and change the path length into a scalar without rigidity, so as to facilitate the summary and sum of the abnormal points. Comparison of abnormal points of each subsequent pollutant.

In one embodiment, a pollutant abnormality monitoring system is proposed, as shown in FIG. 5, and includes the following units:

The acquisition data unit is configured to acquire historical environmental monitoring data of each enterprise's pollution source monitoring point from a preset environmental monitoring data system, and set the historical environmental monitoring data as one pollutant for each enterprise as a data set. storage;

The screening unit is set to set an index threshold for each index item of each pollutant, and to filter each data set according to the index threshold, to filter out data that does not exceed the index threshold as a feature item, and to store the feature item in a non-exceeding standard Data set

The training unit is set to train the isolated forest model by using the feature items in the non-exceeded data set, and establish a corresponding isolated forest model for each non-exceeded data set;

The outlier summary unit is set to obtain real-time monitoring data of a pollutant in an enterprise from the environmental monitoring data system in units of hours, and input the real-time monitoring data into an isolated forest model corresponding to the pollutants. The path length from the root node to the leaf node of the isolated forest model determines whether it is an abnormal point, and summarizes the abnormal points.

In one embodiment, the pollutants in the data acquisition unit include drainage pollutants based on drainage port monitoring and exhaust pollutants based on exhaust port monitoring. When historical environmental monitoring data is stored, each enterprise stores drainage Pollutant dataset and exhaust pollutant dataset.

In one embodiment, the index items of drainage pollutants include at least one index item of suspended solids index, chemical oxygen demand index, pH value or ammonia nitrogen index, and corresponding index thresholds are set for the index items of drainage pollutants. ; The exhaust pollutant index items include at least one of nitrogen oxide index, sulfur dioxide index, soot index or carbon monoxide index, and the exhaust pollutant index items are provided with corresponding index thresholds.

In one embodiment, the screening unit is further configured to classify and store feature items with one pollutant per enterprise in a corresponding non-exceeding data set.

In one embodiment, the training unit includes: left and right child node modules of the construction tree, which are set to take the feature items of every N hours in the non-exceeded data set and set as a point construction feature, each point construction feature and the previous point construction feature Set the difference threshold to X, then the difference between the structural features of two adjacent points is less than X is divided into the left child node of the tree, and the difference greater than or equal to X is divided into the right child node of the tree;

Recursive module, set to recursively construct left and right child nodes until the following conditions are met: the training non-standard data set has only one record or multiple identical records, or the height of the tree reaches a preset height range, and the height range of the tree For: In a non-exceeded data set containing n records, the minimum height of the constructed tree is log (n), and the maximum height of the constructed tree is n-1.

In one embodiment, as shown in FIG. 6, the abnormal point summary unit includes:

Generate a path length module and set it to input the real-time data in the real-time monitoring data one by one into the corresponding isolated forest model. When the real-time data is divided into M times according to the isolated forest model and no longer divided, the real-time data is at the root node of the isolated forest model. The path length to the leaf node is M;

The normalization processing module is configured to perform normalization processing on the path length M of the real-time data to obtain M ′;

The abnormal point summary table module is set to a preset abnormal point threshold value Y. When M ′ is greater than Y, the real-time data is set as an abnormal point, and summarized to generate an abnormal point summary table.

In one embodiment, a computer device is provided, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor causes the processor to execute the pollutant abnormality in the foregoing embodiments. Steps in a monitoring method.

In one embodiment, a storage medium storing computer-readable instructions is provided. When the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the pollutant abnormality in each of the foregoing embodiments. Steps in a monitoring method. The storage medium may be a non-volatile storage medium.

A person of ordinary skill in the art may understand that all or part of the steps in the various methods of the foregoing embodiments may be implemented by a program instructing related hardware. The program may be stored in a computer-readable storage medium. The storage medium may include: Read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical disks, etc.

The technical features of the embodiments described above can be arbitrarily combined. In order to simplify the description, all possible combinations of the technical features in the above embodiments have not been described. However, as long as there is no contradiction in the combination of these technical features, It should be considered as the scope described in this specification.

The above-mentioned embodiments only express some exemplary embodiments of the present application, and their descriptions are more specific and detailed, but cannot be understood as a limitation on the scope of the patent of the present application. It should be noted that, for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of this application patent shall be subject to the appended claims.

Claims (20)

1. An abnormality monitoring method for pollutants, including:

   Acquiring historical environmental monitoring data of a pollution source monitoring point of each enterprise from a preset environmental monitoring data system, and storing the historical environmental monitoring data with one pollutant of each enterprise as a data set for storage;

   An index threshold is set for each index item of each of the pollutants, and each of the data sets is filtered according to the index threshold, and data that does not exceed the index threshold is set as a feature item, and the feature is set. Items are stored in non-exceeded data sets;

   Use the feature terms in the non-exceeded data set to train an isolated forest model, and establish the corresponding isolated forest model for each non-exceeded data set;

   Obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, input the real-time monitoring data into the isolated forest model corresponding to the pollutant, and pass the real-time The monitoring data reaches the path length from the root node to the leaf node of the isolated forest model to determine whether it is an abnormal point, and the abnormal points are summarized.
2. The method for monitoring pollutant anomaly according to claim 1, wherein the pollutants include drainage pollutants based on drainage outlet monitoring and exhaust pollutants based on exhaust outlet monitoring, and the historical environmental monitoring data is stored At the time, each enterprise stored a drainage pollutant data set and an exhaust pollutant data set.
3. The pollutant abnormality monitoring method according to claim 2, wherein the index items of the drainage pollutants include at least one index item of a suspended matter index, a chemical oxygen demand index, a pH value, or an ammonia nitrogen index. The indicators of drainage pollutants are set with corresponding indicator thresholds;

   The index items of the exhaust pollutant include at least one of a nitrogen oxide index, a sulfur dioxide index, a soot index, or a carbon monoxide index, and a corresponding index threshold is set for each of the exhaust pollutant index items.
4. The method for monitoring anomalies of pollutants according to claim 1, wherein said storing the characteristic items in a non-exceeding data set comprises: classifying and storing the characteristic items with one pollutant per enterprise in a corresponding Of non-exceeded data sets.
5. The method for monitoring pollutant anomaly according to claim 1, wherein the training of the isolated forest model by using the feature terms in the non-exceeding data set comprises:

   Taking the feature item every N hours in the non-exceeding data set as a point structure feature, and setting the difference threshold between each of the point structure feature and the previous point structure feature to X, two adjacent point structures are constructed Differences between features less than X are assigned to the left child of the tree, and differences greater than or equal to X are assigned to the right child of the tree;

   Construct the left child node and the right child node recursively until the following conditions are met:

   The non-exceeded data set for training has only one record or multiple identical records, or the height of the tree reaches a preset height range, and the height range of the tree is: in the non-exceeded data set containing n records, construct The minimum height of the tree is log (n), and the maximum height of the constructed tree is n-1.
6. The pollutant abnormality monitoring method according to claim 1, wherein the real-time monitoring data is input into the isolated forest model corresponding to the pollutant, and the isolated forest model is reached through the real-time monitoring data. The length of the path from the root node to the leaf node determines whether it is an abnormal point, including:

   When the real-time data in the real-time monitoring data is input into the corresponding isolated forest model one by one, and the real-time data is divided into M times according to the isolated forest model and no longer divided, the real-time data is in the isolated The path length from the root node to the leaf node of the forest model is M;

   Performing normalization processing on the path length M of the real-time data to obtain M ';

   The abnormal point threshold value Y is preset, and when M 'is greater than Y, the real-time data is set as an abnormal point, and summarized to generate an abnormal point summary table.
7. A pollutant abnormality monitoring system includes:

   The acquiring data unit is configured to acquire historical environmental monitoring data of a pollution source monitoring point of each enterprise from a preset environmental monitoring data system, and set the historical environmental monitoring data to one data of one pollutant per enterprise Set for storage;

   The screening unit is configured to set an index threshold for each index item of each of the pollutants, filter each of the data sets according to the index threshold value, and filter out data that does not exceed the index threshold value as a feature item , Storing the feature items in a non-exceeding data set;

   A training unit configured to train an isolated forest model by using the feature terms in the non-exceeded data set, and establish the corresponding isolated forest model for each non-exceeded data set;

   The abnormal point summary unit is configured to obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, and input the real-time monitoring data into the isolated forest corresponding to the pollutant In the model, it is determined whether the path length from the root node to the leaf node of the isolated forest model is an abnormal point through the real-time monitoring data, and the abnormal points are summarized.
8. The pollutant abnormality monitoring system according to claim 7, wherein the pollutants in the acquisition data unit include drainage pollutants based on monitoring of a drain outlet, and exhaust pollutants based on monitoring of an exhaust outlet. When the historical environmental monitoring data is stored, each enterprise stores a drainage pollutant data set and an exhaust pollutant data set.
9. The pollutant abnormality monitoring system according to claim 8, wherein the index items of the drainage pollutants include at least one index item of a suspended matter index, a chemical oxygen demand index, a pH value, or an ammonia nitrogen index. The indicators of drainage pollutants are set with corresponding indicator thresholds;

   The index items of the exhaust pollutant include at least one of a nitrogen oxide index, a sulfur dioxide index, a soot index, or a carbon monoxide index, and a corresponding index threshold is set for each of the exhaust pollutant index items.
10. The pollutant abnormality monitoring system according to claim 7, wherein the screening unit is further configured to classify and store the characteristic items with one pollutant per enterprise in a corresponding non-exceeding data set.
11. The pollutant abnormality monitoring system according to claim 7, wherein the training unit comprises:

    The left and right child node modules of the construction tree are set to take a feature item every N hours in the non-exceeding data set as a point structure feature, and a difference threshold is set for the difference between each of the point structure features and the previous point structure feature. Is X, the difference between the structural features of two adjacent points is less than X is divided into the left child node of the tree, and the difference greater than or equal to X is divided into the right child node of the tree;

    A recursive module configured to recursively construct the left child node and the right child node until the following conditions are met: the non-exceeded data set for training has only one record or multiple identical records, or the height of the tree reaches a preset height Range, the height range of the tree is: in the non-exceeding data set containing n records, the minimum height of the constructed tree is log (n), and the maximum height of the constructed tree is n-1.
12. The pollutant abnormality monitoring system according to claim 7, wherein the abnormal point summary unit comprises:

    A path length generating module is configured to input real-time data in the real-time monitoring data one by one into the corresponding isolated forest model, and when the real-time data is divided into M times according to the isolated forest model and no longer divided, The path length of the real-time data from the root node to the leaf node of the isolated forest model is M;

    A normalization processing module configured to perform normalization processing on the path length M of the real-time data to obtain M ';

    The abnormal point summary table module is set to a preset abnormal point threshold value Y. When M 'is greater than Y, the real-time data is set as an abnormal point, and summarized to generate an abnormal point summary table.
13. A computer device includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor causes the processor to perform the following steps:

    Acquiring historical environmental monitoring data of a pollution source monitoring point of each enterprise from a preset environmental monitoring data system, and storing the historical environmental monitoring data with one pollutant of each enterprise as a data set for storage;

    An index threshold is set for each index item of each of the pollutants, and each of the data sets is filtered according to the index threshold, and data that does not exceed the index threshold is set as a feature item, and the feature is set. Items are stored in non-exceeded data sets;

    Use the feature terms in the non-exceeded data set to train an isolated forest model, and establish the corresponding isolated forest model for each non-exceeded data set;

    Obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, input the real-time monitoring data into the isolated forest model corresponding to the pollutant, and pass the real-time The monitoring data reaches the path length from the root node to the leaf node of the isolated forest model to determine whether it is an abnormal point, and the abnormal points are summarized.
14. The computer device according to claim 13, wherein the storing the feature items in a non-exceeded data set causes the processor to perform the following steps:

    The characteristic items are classified by one pollutant of each enterprise and stored in the corresponding non-exceeding data set.
15. The computer device according to claim 13, wherein the training of the isolated forest model by using the feature terms in the non-exceeded data set causes the processor to perform the following steps:

    Taking the feature item every N hours in the non-exceeding data set as a point structure feature, and setting the difference threshold between each of the point structure feature and the previous point structure feature to X, two adjacent point structures are constructed Differences between features less than X are assigned to the left child of the tree, and differences greater than or equal to X are assigned to the right child of the tree;

    Construct the left child node and the right child node recursively until the following conditions are met:

    The non-exceeded data set for training has only one record or multiple identical records, or the height of the tree reaches a preset height range, and the height range of the tree is: in the non-exceeded data set containing n records, construct The minimum height of the tree is log (n), and the maximum height of the constructed tree is n-1.
16. The computer device according to claim 13, wherein the real-time monitoring data is input into the isolated forest model corresponding to the pollutant, and the real-time monitoring data is used to reach a root node of the isolated forest model. When determining whether the path length to a leaf node is an abnormal point, the processor is caused to perform the following steps:

    When the real-time data in the real-time monitoring data is input into the corresponding isolated forest model one by one, and the real-time data is divided into M times according to the isolated forest model and no longer divided, the real-time data is in the isolated The path length from the root node to the leaf node of the forest model is M;

    Performing normalization processing on the path length M of the real-time data to obtain M ';

    The abnormal point threshold value Y is preset, and when M 'is greater than Y, the real-time data is set as an abnormal point, and summarized to generate an abnormal point summary table.
17. A storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the following steps:

    Acquiring historical environmental monitoring data of a pollution source monitoring point of each enterprise from a preset environmental monitoring data system, and storing the historical environmental monitoring data with one pollutant of each enterprise as a data set for storage;

    An index threshold is set for each index item of each of the pollutants, and each of the data sets is filtered according to the index threshold, and data that does not exceed the index threshold is set as a feature item, and the feature is set. Items are stored in non-exceeded data sets;

    An isolated forest model is trained by using the feature items in the non-exceeded data set, and the corresponding isolated forest model is established for each non-exceeded data set;

    Obtain real-time monitoring data of a pollutant in an enterprise in units of hours from the environmental monitoring data system, input the real-time monitoring data into the isolated forest model corresponding to the pollutant, and pass the real-time The monitoring data reaches the path length from the root node to the leaf node of the isolated forest model to determine whether it is an abnormal point, and the abnormal points are summarized.
18. The storage medium according to claim 17, wherein when storing the feature items in a non-exceeded data set, causes one or more of the processors to perform the following steps:

    The characteristic items are classified by one pollutant of each enterprise and stored in the corresponding non-exceeding data set.
19. The storage medium according to claim 17, wherein when training the isolated forest model by using the feature terms in the non-exceeded data set, one or more of the processors are caused to perform the following steps:

    Taking the feature item every N hours in the non-exceeding data set as a point structure feature, and setting the difference threshold between each of the point structure feature and the previous point structure feature to X, two adjacent point structures are constructed Differences between features less than X are assigned to the left child of the tree, and differences greater than or equal to X are assigned to the right child of the tree;

    Construct the left child node and the right child node recursively until the following conditions are met:

    The non-exceeded data set for training has only one record or multiple identical records, or the height of the tree reaches a preset height range, and the height range of the tree is: in the non-exceeded data set containing n records, construct The minimum height of the tree is log (n), and the maximum height of the constructed tree is n-1.
20. The storage medium according to claim 17, wherein the real-time monitoring data is input into the isolated forest model corresponding to the pollutant, and the real-time monitoring data is used to reach a root node of the isolated forest model. When determining whether the path length to a leaf node is an abnormal point, one or more of the processors are caused to perform the following steps:

    When the real-time data in the real-time monitoring data is input into the corresponding isolated forest model one by one, and the real-time data is divided into M times according to the isolated forest model and no longer divided, the real-time data is in the isolated The path length from the root node to the leaf node of the forest model is M;

    Performing normalization processing on the path length M of the real-time data to obtain M ';

    The abnormal point threshold value Y is preset, and when M 'is greater than Y, the real-time data is set as an abnormal point, and summarized to generate an abnormal point summary table.

Images