Bad Odour Prediction Using Historical Environmental Sensors

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods fail to accurately predict bad odour episodes in geographical areas, which impact the environment and human health, lacking real-time and reliable prediction capabilities.

Innovation Solution

A method involving the formation of a training dataset with past sensor and observation data, configuring a predictor model using machine learning techniques, and providing real-time sensor data for near-real-time odour predictions, utilizing clustering and retrospective approaches to enhance prediction accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional monitoring methods are used, then system complexity is low, but prediction accuracy and reliability are insufficient

Engineering Contradiction:
Improveprediction accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the prediction task into multiple independent modules: data collection from multiple sensors, data preprocessing and cleaning, feature extraction, model training, and prediction generation. This modular architecture improves prediction accuracy while managing system complexity through division of labor.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary actions by collecting and preprocessing historical sensor data before actual prediction is needed. Training datasets are prepared in advance, and models are trained offline, so that when prediction is required, the system can quickly generate accurate results without complex real-time processing.

Inventive Principle:
Principle #10Preliminary action

2Loss of time

If real-time prediction is implemented, then response time to odour episodes is reduced, but data processing requirements and computational resources increase

Engineering Contradiction:
Improveresponse timeVSAvoidcomputational resources
Core Design Contradiction:
Loss of timeVSUse of energy by moving object

Solution Approach 1:

Historical sensor data is collected and preprocessed in advance to build training datasets. Machine learning models are trained offline using this preprocessed data, so that real-time prediction only requires feeding current sensor readings into the trained model, significantly reducing computational resources needed during actual prediction events.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system uses readily available sensor data from the environment without requiring extensive external computational resources. The trained model serves itself by automatically processing incoming sensor data and generating predictions, minimizing the need for complex real-time computational infrastructure.

Inventive Principle:
Principle #25Self-service

3Reliability

If comprehensive sensor data is collected, then prediction reliability is improved, but data management complexity and storage requirements increase

Engineering Contradiction:
Improveprediction reliabilityVSAvoiddata management complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system extracts only the relevant features from comprehensive sensor data that are most predictive of odour episodes. Instead of managing and processing all raw sensor data, the system identifies and extracts key features such as pollutant concentrations, meteorological conditions, and temporal patterns, significantly reducing data management complexity while maintaining prediction reliability.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Sensor data is preprocessed and cleaned in advance, with irrelevant data removed and important features extracted before model training. This preliminary data preparation reduces the complexity of data management during prediction operations while ensuring that only high-quality, relevant data is used to improve prediction reliability.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4641456A1Bad odour prediction
Publication Date: 2025.10.29 FUNDACIO EURECAT
  • EP4641456A1 patent drawingFigure 1
  • EP4641456A1 patent drawingFigure 2
  • EP4641456A1 patent drawingFigure 3

AI summary

Methods are provided of predicting bad odour in a geographical area. These methods comprise: forming a training dataset comprising data rows each having example input data including past sensor data generated in the geographical area from environmental sensors during a past time interval, and target output data including past observation data representing bad odour observations generated in the geographical area during a past time period posterior to the past time interval; configuring a computing module by training a predictor model implemented at/by the computing module using the training dataset; building a production dataset comprising production input data including present sensor data generated in the geographical area from the environmental sensors during a present time interval; and providing the production dataset to the configured computing module to cause the trained predictor model to output bad odour prediction. Computer programs, systems and computing systems suitable to perform such methods are also provided.