Self-adaptive optimization framework for water quality prediction
Inventors
Assignees
Hong Kong Applied Science and Technology Research Institute ASTRI
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12306166-B2
Publication Date
2025-05-20
Expiration Date
Abstract
The framework predicts water-quality data from observation data associated with raw features, e.g., rainfall. In the framework, an artificial neural network (ANN) computes the predicted water-quality data from feature data associated with impact features and derived from the observation data. The impact features are learnable, and are selected from impact-feature candidates comprising directional change- (DC-based) features each indicating occurrence of DC events in a corresponding raw feature. Including the DC-based features in the candidates enhances the ANN's ability of capturing significant change patterns of water quality due to extreme/unexpected events. The ANN architecture is also configurable according to model hyperparameters, which are learnable. The impact features and model hyperparameters are learnt by differential evolution for maximizing a prediction performance achieved by the ANN, thereby enabling the ANN architecture and impact features to be automatically optimized without requiring manual adjustment by domain experts in applying the ANN to different situations.
Core Innovation
The invention provides a computer-implemented method for predicting water quality from observation data associated with a plurality of raw features relevant to water-quality prediction, to generate predicted water-quality data. The method sets up an artificial neural network that computes the predicted water-quality data from feature data associated with a plurality of impact features, and the feature data associated with the plurality of impact features is computed from the observation data. The plurality of impact features is learnable, and the ANN architecture is configurable according to model hyperparameters, with ANN model parameters for configuring an input-output relationship of the ANN.
The method learns the plurality of impact features, the plurality of model hyperparameters, and the plurality of ANN model parameters. In learning the plurality of impact features, an individual impact feature is selected from a plurality of impact-feature candidates comprising a plurality of directional change-based features, each associated with a corresponding raw feature and used to indicate occurrence of any directional change event in the corresponding raw feature. Inclusion of the directional change-based feature in the plurality of impact-feature candidates enhances the ANN’s ability of capturing possible significant change patterns of water quality due to extreme or unexpected events.
A key aspect is that the impact features, model hyperparameters, and ANN model parameters are automatically optimized, including learning the plurality of impact features and the plurality of model hyperparameters by differential evolution for maximizing a water-quality prediction performance achieved by the ANN. This enables the ANN architecture and the plurality of impact features to be automatically optimized without a need for manual adjustment by domain experts in applying the ANN to different situations of water-quality prediction. The framework is also described with a self-adaptive optimization differential evolution controller and an ANN configuration, training, and testing workflow that evaluates prediction accuracy via a prediction-performance value, with iterative search and stopping when the value converges or a stopping criterion is met.
Claims Coverage
The partial content includes two independent claims. Across the independent claims, the inventive coverage centers on learnable directional-change-based impact features combined with configurable ANN hyperparameters and parameters, and automatic optimization using differential evolution, including self-adaptive differential evolution components and decoding where specified.
Learnable impact features using directional change-based features to capture extreme or unexpected events
A method that sets up an ANN to compute predicted water-quality data from feature data associated with a plurality of learnable impact features computed from observation data, where impact-feature candidates include a plurality of directional change-based features each associated with a corresponding raw feature and indicating occurrence of a directional change event, and inclusion of the directional change-based feature enhances the ANN’s ability of capturing possible significant change patterns of water quality due to extreme or unexpected events.
Automatic optimization of learnable impact features and model hyperparameters by differential evolution for prediction performance
A method in which the plurality of impact features and the plurality of model hyperparameters are learned by differential evolution for maximizing a water-quality prediction performance achieved by the ANN, thereby enabling the ANN architecture and the plurality of impact features to be automatically optimized without a need for manual adjustment by domain experts when applying the ANN to different situations of water-quality prediction.
The independent-claim coverage is grounded in learnable directional-change-based impact features associated with raw features for capturing significant change patterns under extreme or unexpected events, and automatic optimization of impact features and model hyperparameters using differential evolution to maximize water-quality prediction performance without manual domain-expert adjustment.
Stated Advantages
Captures possible significant change patterns of water quality due to extreme or unexpected events.
Automatically optimizes the ANN architecture and the plurality of impact features without a need for manual adjustment by domain experts in applying the ANN to different situations of water-quality prediction.
Documented Applications
Water-quality prediction, including pH prediction, using an ANN optimized to improve performance, particularly during extreme or unexpected events [procedural detail omitted for safety].
Interested in licensing this patent?