Abnormal air pollution emission prediction

Patent metadata

Field	Value
Publication number	US-11836644-B2
Application number	US-201916532543-A
Country	US
Kind code	B2
Filing date	Aug 6, 2019
Priority date	Aug 6, 2019
Publication date	Dec 5, 2023
Grant date	Dec 5, 2023

Abstract

A method, a device and a computer program product for abnormal air pollution emission prediction are proposed. In the method, a first set of features characterizing air condition in a zone is obtained. Whether the zone is subject to abnormal air pollution emission in a future first time period is determined based on the first set of features and using a first prediction classifier. In response to determining that the zone is subject to abnormal air pollution emission in the first time period, a second set of features characterizing air condition in the zone is obtained. A future second time period in which the zone is subject to abnormal air pollution emission is determined based on the second set of features and using a second prediction classifier. The second time period is included in the first time period. In this way, the abnormal air pollution emission in the zone can be accurately and efficiently predicted.

First claim

What is claimed is:

1. A method for abnormal air pollution emission prediction using prediction classifiers based on trained neural networks, comprising:
processing each zone of a plurality of zones based on a first set of features and a first time period by:
obtaining the first set of features, the first set of features characterizing air condition in a respective zone; and
predicting, based on the first set of features and using a first prediction classifier of the prediction classifiers, whether the respective zone is subject to abnormal air pollution emission in the first time period;

upon predicting a zone of the plurality of zones to be subject to the abnormal air pollution emission in the first time period, processing the zone based on a second set of features and a subset of the first time period by:
obtaining the second set of features, the second set of features characterizing air condition in the zone, wherein the second set of features is different than the first set of features and includes a number of times the zone was previously subject to the abnormal air pollution; and
predicting, based on the second set of features and using a second prediction classifier of the prediction classifiers, the subset of the first time period and in which the zone is subject to the abnormal air pollution emission; and

outputting an indication of at least one of the zone or the subset of the first time period, wherein the indication causes an air pollution emission of the zone to be reduced back to within normal limits;
wherein the processing based on the second set of features is performed only on any zone predicted to be subject to the abnormal air pollution emission in the first time period, wherein the subset of the first time period is shorter in duration than the first time period, and wherein each of the first and second prediction classifiers is based on a respective one of the trained neural networks.

2. The method of claim 1 , wherein the subset of the first time period consists of a second time period that is shorter in duration than the first time period, and wherein obtaining the first set of features comprises obtaining at least one of:
historical cumulative concentration information characterizing air pollutant concentration accumulated in the zone in a third time period that has passed, the zone being the predicted zone;
historical wind speed in the zone in the third time period;
historical wind direction in the zone in the third time period;
historical temperature in the zone in the third time period;
a number of times the zone being previously predicted to be subject to the abnormal air pollution emission in a fourth time period that has passed;
whether the zone was predicted to be subject to the abnormal air pollution emission in a fifth time period that has passed; or
time since the zone was last predicted to be subject to the abnormal air pollution emission.

3. The method of claim 2 , wherein obtaining the historical cumulative concentration information comprises:
dividing the third time period into a plurality of time intervals;
for each time interval of the plurality of time intervals,
determining a concentration difference between a first measure of air pollutant concentration in the zone in a respective time interval and a second measure of air pollutant concentration surrounding the zone in the respective time interval;
determining a weight for the concentration difference based on a time distance between a predetermined time point in the respective time interval and a most recent time point in the third time period, the weight decreasing as the time distance increases; and

determining the historical cumulative concentration information based on the concentration difference and the weight for the concentration difference.

4. The method of claim 1 , further comprising, prior to obtaining the first set of features:
obtaining a third set of features characterizing air condition in a zone candidate;
determining, based on the third set of features, whether the zone candidate includes a potential pollution source;
in response to determining the zone candidate includes the potential pollution source, selecting the zone candidate as the zone rather than as being merely a candidate; and
obtaining the first set of features for the zone;
wherein the first, second, and third sets of features each includes a respective, distinct feature.

5. The method of claim 4 , wherein obtaining the third set of features comprises obtaining at least one of:
an area ratio occupied by a non-pollution source in the zone candidate;
a type of the potential pollution source;
a size of the potential pollution source;
a density of the potential pollution source;
a predetermined air pollutant emission scale of the potential pollution source; or
time since the zone candidate was last predicted to be subject to the abnormal air pollution emission.

6. The method of claim 1 , wherein predicting whether the zone is subject to the abnormal air pollution emission in the first time period comprises:
determining, based on the first set of features and using the first prediction classifier, a likelihood of the zone being subject to the abnormal air pollution emission in the first time period;
comparing the likelihood, of the zone being subject to the abnormal air pollution emission in the first time period, with a first threshold likelihood of the zone being subject to the abnormal air pollution emission; and
in response to determining the likelihood exceeds the first threshold likelihood, predicting the zone is subject to the abnormal air pollution emission in the first time period.

7. The method of claim 1 , wherein the subset of the first time period consists of a second time period that is shorter in duration than the first time period, and wherein obtaining the second set of features comprises obtaining at least one of:
a number of times the zone being previously predicted to be subject to the abnormal air pollution emission in the second time period;
historical air pollutant concentration in the zone in the second time period; or
historical air pollutant concentration surrounding the zone in the second time period.

8. The method of claim 1 , wherein predicting the subset of the first time period comprises:
determining, based on the second set of features and using the second prediction classifier, a likelihood of the zone being subject to the abnormal air pollution emission in a candidate time period within the first time period;
comparing the likelihood, of the zone being subject to the abnormal air pollution emission in a candidate time period within the first time period, with a second threshold likelihood of the zone being subject to the abnormal air pollution emission; and
in response to determining the likelihood exceeds the second threshold likelihood, determining the candidate time period to be the subset of the first time period.

9. The method of claim 1 , wherein the subset of the first time period consists of a second time period that is shorter in duration than the first time period, wherein the zone comprises a first zone of the plurality of zones, and wherein processing of a second zone, different than the first zone, of the plurality of zones based on the second set of features is avoided after the second zone is processed based on the first set of features;
wherein to improve processing efficiency in performing the air pollution emission prediction, the processing based on the first set of features is performed only on zones determined to be pollution candidates, and the processing based on the second set of features is performed only on the zones predicted to be subject to the abn

Assignees

• IBM

Inventors

• Wang Lingyun
• Qu Junmei
• Xia Xi
• Bai Xin Xin
• Shao Jin Yan

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N5/048 Primary

  Fuzzy inferencing · CPC title

• G01W1/10

  Devices for predicting weather conditions (computers per se G06; display devices G09) · CPC title

• G16Z99/00

  Subject matter not provided for in other main groups of this subclass · CPC title

• G01N33/0004

  Gaseous mixtures, e.g. polluted air · CPC title