Multistep-ahead forecasting of chlorophyll a using a wavelet nonlinear autoregressive network

Enhanced forecasting of chlorophyll-a concentration in coastal waters through integration of Fourier analysis and Transformer networks

The accurate prediction of chlorophyll-a (chl-a) concentration in coastal waters is essential to coastal economies and ecosystems as it serves as the key indicator of harmful algal blooms. Although powerful machine learning methods have made strides in forecasting chl-a concentrations, there remains a gap in effectively modeling the dynamic temporal patterns and dealing with data noise and unreliability. To wiggle out of quagmires, we introduce an innovative deep learning prediction model (termed ChloroFormer) by integrating Transformer networks with Fourier analysis within a decomposition architecture, utilizing coastal in-situ data from two distinct study areas. Our proposed model exhibits superior capabilities in capturing both short-term and middle-term dependency patterns in chl-a concentrations, surpassing the performance of six other deep learning models in multistep-ahead predictive accuracy. Particularly in scenarios involving extreme and frequent blooms, our proposed model shows exceptional predictive performance, especially in accurately forecasting peak chl-a concentrations. Further validation through Kolmogorov-Smirnov tests attests that our model not only replicates the actual dynamics of chl-a concentrations but also preserves the distribution of observation data, showcasing its robustness and reliability. The presented deep learning model addresses the critical need for accurate prediction on chl-a concentrations, facilitating the exploration of marine observations with complex dynamic temporal patterns, thereby supporting marine conservation and policy-making in coastal areas.

Real-time chlorophyll-a forecasting using machine learning framework with dimension reduction and hyperspectral data

Since water is an essential resource in various fields, it requires constant monitoring. Chlorophyll-a concentration is a crucial indicator of water quality and can be used to monitor water quality. In this study, we developed methods to forecast chlorophyll-a concentrations in real-time using hyperspectral data on IoT platform and various machine learning algorithms. Compared to regular cameras that record information only in the three broad color bands of red, green, and blue, the hyperspectral images of drinking water sources record the data in dozens or even hundreds of distinct small wavelength bands, providing each pixel in an image with a full spectrum. Different machine learning algorithms have been developed using hyperspectral data and field observations of water quality and weather conditions. Previous studies have predicted chlorophyll concentrations using either partial least squares (PLS), which is a dimensionality reduction method, or machine learning. In contrast, our study employed the PLS technique as a preprocessing step to diminish the dimensionality of the hyperspectral data, followed by the application of the machine learning techniques with optimized hyperparameters to improve the precision of the predictions, thereby introducing a real-time mechanism for chlorophyll-a prediction. Consequently, a machine learning algorithm with R2 values of 0.9 or above and sufficiently small RMSE was developed for real-time chlorophyll-a forecasting. Real-time chlorophyll-a forecasting using LightGBM has the best performance, with a mean R2 of 0.963 and a mean RMSE of 2.679. This paper is expected to have applications in algal bloom early detection on monitoring systems.

A new framework for water quality forecasting coupling causal inference, time-frequency analysis and uncertainty quantification

Accurate forecasting of water quality variables in river systems is crucial for relevant administrators to identify potential water quality degradation issues and take countermeasures promptly. However, pure data-driven forecasting models are often insufficient to deal with the highly varying periodicity of water quality in today's more complex environment. This study presents a new holistic framework for time-series forecasting of water quality parameters by combining advanced deep learning algorithms (i.e., Long Short-Term Memory (LSTM) and Informer) with causal inference, time-frequency analysis, and uncertainty quantification. The framework was demonstrated for total nitrogen (TN) forecasting in the largest artificial lakes in Asia (i.e., the Danjiangkou Reservoir, China) with six-year monitoring data from January 2017 to June 2022. The results showed that the pre-processing techniques based on causal inference and wavelet decomposition can significantly improve the performance of deep learning algorithms. Compared to the individual LSTM and Informer models, wavelet-coupled approaches diminished well the apparent forecasting errors of TN concentrations, with 24.39%, 32.68%, and 41.26% reduction at most in the average, standard deviation, and maximum values of the errors, respectively. In addition, a post-processing algorithm based on the Copula function and Bayesian theory was designed to quantify the uncertainty of predictions. With the help of this algorithm, each deterministic prediction of our model can correspond to a range of possible outputs. The 95% forecast confidence interval covered almost all the observations, which proves a measure of the reliability and robustness of the predictions. This study provides rich scientific references for applying advanced data-driven methods in time-series forecasting tasks and a practical methodological framework for water resources management and similar projects.

Algal bloom forecasting with time-frequency analysis: A hybrid deep learning approach

The rapid emergence of deep learning long-short-term-memory (LSTM) technique presents a promising solution to algal bloom forecasting. However, the discontinuous and non-stationary processes within algal dynamics still largely limit the functions of LSTMs. To overcome this challenge, an advanced time-frequency wavelet analysis (WA) technique was introduced to enhance the prediction accuracy of LSTMs. Herein, the novel hybrid approach (named WLSTM) successfully decreased the algal forecasting inaccuracy of classic LSTMs by 41% ± 8% in Lake Mendota (Wisconsin, USA), with powerful one-step-ahead predictions at hourly, daily, and monthly time resolutions (R² = 0.976, 0.878, and 0.814, respectively). In addition, the WLSTM outperformed the other two widely used algal forecasting approaches - deep neural network (DNN), and autoregressive-integrated-moving-average (ARIMA) model, represented by average 72% and 85% decrease in root-mean-square-error, respectively. Furthermore, the WLSTM was implemented in an experimentally fertilized lake (Lake Tuesday, Michigan) for a multi-step forecasting examination. It satisfactorily forecasted the algal fluctuations involving substantial peak and extreme values (average R² > 0.900) and presented accurate judgment outcomes to their bloom levels with high accuracy > 95% on average. This work highlighted the utility of deep learning approaches in effective early-warning for algal blooms, and demonstrated an important direction for improving the adaptability of conventional deep learning approaches to the aquatic problems.

Quantitative Diagnosis of TCM Syndrome Types Based on Adaptive Resonant Neural Network

Artificial intelligence has become one of the most rapidly developing disciplines in the application field of pattern recognition. In target recognition, sometimes, there are multiple identical or similar copies of the target to be recognized in the image, and it is difficult to classify and estimate by traditional methods. In this case, it is necessary to use the SOM network to separate multiple targets and use the multiple order parameters in the improved SNN to pair the target. The change of its thickness can intuitively reflect the abnormality of its tissue. Therefore, the choroidal thickness of the central fovea can be measured to study the relationship between the choroidal structure and BRVO and arteriosclerosis. The purpose of this study is to further study the correlation between branch retinal vein occlusion and arteriosclerosis by quantitatively measuring retinal vessel diameter and choroidal thickness, to analyze the correlation between different TCM syndrome types of nonischemic BRVO and retinal arteriosclerosis, and to provide theoretical basis for clinical nonischemic BRVO TCM syndrome types and traditional Chinese medicine treatment, so as to reflect its clinical application value. In order to solve the single fixed structure of traditional SNN and poor scalability, combined with the Kohonen layer structure in the self-organizing mapping network, an improved collaborative neural network model is proposed. This paper studies the network training method and operation convergence and analyzes the converged network and the pattern classification results obtained by the network. In order to solve the single fixed structure of traditional SNN and poor scalability, combined with the Kohonen layer structure in the self-organizing mapping network, an improved collaborative neural network model is proposed. The results of our proposed improved model on the MNIST dataset can achieve the same level of current state-of-the-art machine learning classifiers in recognition accuracy with a smaller network size and network complexity.

Real-Time Regulation Model of Physical Fitness Training Intensity Based on Wavelet Recursive Fuzzy Neural Network

It has been widely used in signal processing, image processing, speech recognition and synthesis, pattern recognition, machine vision, machinery fault diagnosis and monitoring, and other scientific and technological fields and has achieved great results. The application potential in nonlinear system identification is increasing. According to the theory of “overload recovery” and “functional reserve”, the mathematical model of “load-fitness state” is established to understand the adaptation characteristics and individual characteristics of athletes to sports training. The model is used to simulate the values and time required to reach the maximum fitness state for four types of precompetition reduction plans and to provide a reference for the development of precompetition training plans. The data required for parameter estimation were the actual training data of six outstanding basketball athletes (mean age 18.2 ± 0.75, mean training years 4.6 ± 0.49). And the coaches' training plan was not intervened during the test. In order to further reduce the biaxial synchronization error of the sports platform and improve the stability of the system, the wavelet transformation capable of time-varying signal analysis and the recursive structure with dynamic capability were combined with the fuzzy neural network, and the learning ability of the neural network was used to learn and adjust the scaling and translation factors in the wavelet function, the mean and standard deviation in the fuzzy structure, and the connection weights between the layers, according to the biaxial synchronization. The simulation results show that the designed global sliding mode controller can improve the convergence speed of tracking error and ensure the single-axis tracking accuracy of the H-type motion platform compared with the traditional sliding mode controller, and the tracking accuracy and synchronization accuracy of the system can be further improved after adding the cross-coupled synchronization controller, but the improvement of synchronization control accuracy is not very satisfactory due to the fixed selection of the parameters of the cross-coupled controller. Further improvement is needed.

Blood Glucose Level Prediction of Diabetic Type 1 Patients Using Nonlinear Autoregressive Neural Networks

Diabetes type 1 is a chronic disease which is increasing at an alarming rate throughout the world. Studies reveal that the complications associated with diabetes can be reduced by proper management of the disease by continuously monitoring and forecasting the blood glucose level of patients. Objective. The prior prediction of blood glucose level is necessary to overcome the lag time for insulin absorption in diabetic type 1 patients. Method. In this research, we use continuous glucose monitoring (CGM) data to predict future blood glucose level using the previous data points. We compare two neural network techniques. We apply the optimal feedforward neural network and then propose optimal nonlinear autoregressive neural networks for blood glucose prediction 15–30 minutes earlier for diabetic type 1 patients. We validate the proposed model with 2 virtual subjects using their 24-hour blood glucose level data. These two case studies have been compiled from AIDA, i.e., the freeware mathematical diabetes simulator. Results. In the prediction horizon (PH) of 15 and 30 minutes, improved results have been shown for minimal inputs for blood glucose level of a particular subject. Root mean square error (RMSE) is used for performance calculation. For the optimal feedforward neural network, the RMSE is 0.9984 and 3.78 ml/dl, and for the optimal nonlinear autoregressive neural network, it reduces the RMSE to 0.60 and 1.12 ml/dl for 15 min and 30 min prediction horizons, respectively, for subject 1. Similarly, for subject 2 for the optimal feedforward neural network, RMSE is 1.43 and 3.51 ml/dl which is improved using the optimal autoregressive neural network to 0.7911 and 1.6756 ml/dl for 15 min and 30 min prediction horizons, respectively. Validation. We further validate our proposed model using UCI machine learning datasets (Abalone and Servo), and it shows improved results on that as well. Conclusion and Future Work. The proposed optimal nonlinear autoregressive neural network model performs better than the feedforward neural network model for these time series data. In the future, we intend to investigate a greater collection of AIDA scenarios and data that are real and influence other factors of BGLs.

Advances in forecasting harmful algal blooms using machine learning models: A case study with Planktothrix rubescens in Lake Geneva

The development of anthropic activities during the 20th century increased the nutrient fluxes in freshwater ecosystems, leading to the eutrophication phenomenon that most often promotes harmful algal blooms (HABs). Recent years have witnessed the regular and massive development of some filamentous algae or cyanobacteria in Lake Geneva. Consequently, important blooms could result in detrimental impacts on economic issues and human health. In this study, we tried to lay the foundation of an HAB forecast model to help scientists and local stakeholders with the present and future management of this peri-alpine lake. Our forecast strategy was based on pairing two machine learning models with a long-term database built over the past 34 years. We created HAB groups via a K-means model. Then, we introduced different lag times in the input of a random forest (RF) model, using a sliding window. Finally, we used a high-frequency dataset to compare the natural mechanisms with numerical interaction using individual conditional expectation plots. We demonstrate that some HAB events can be forecasted over a year scale. The information contained in the concentration data of the cyanobacteria was synthesized in the form of four intensity groups that directly depend on the P. rubescens concentration. The categorical transformation of these data allowed us to obtain a forecast with correlation coefficients that stayed above a threshold of 0.5 until one year for the counting cells and two years for the biovolume data. Moreover, we found that the RF model predicted the best P. rubescens abundance for water temperatures around 14°C. This result is consistent with the biological processes of the toxic cyanobacterium. In this study, we found that the coupling between K-means and RF models could help in forecasting the development of the bloom-forming P. rubescens in Lake Geneva. This methodology could create a numerical decision support tool, which should be a significant advantage for lake managers.

Crop/Weed Discrimination Using a Field Imaging Spectrometer System

Nowadays, sensors begin to play an essential role in smart-agriculture practices. Spectroscopy and the ground-based sensors have inspired widespread interest in the field of weed detection. Most studies focused on detection under ideal conditions, such as indoor or under artificial lighting, and more studies in the actual field environment are needed to test the applicability of this sensor technology. Meanwhile, hyperspectral image data collected by imaging spectrometer often has hundreds of channels and, thus, are large in size and highly redundant in information. Therefore, a key element in this application is to perform dimensionality reduction and feature extraction. However, the processing of highly dimensional spectral imaging data has not been given due attention in recent studies. In this study, a field imaging spectrometer system (FISS; 380-870 nm and 344 bands) was designed and used to discriminate carrot and three weed species (purslane, humifuse, and goosegrass) in the crop field. Dimensionality reduction was performed on the spectral data based on wavelet transform; the wavelet coefficients were extracted and used as the classification features in the weed detection model, and the results were compared with those obtained by using spectral bands as the classification feature. The classification features were selected using Wilks' statistic-based stepwise selection, and the results of Fisher linear discriminant analysis (LDA) and the highly dimensional data processing-oriented support vector machine (SVM) were compared. The results indicated that multiclass discrimination among weeds or between crops and weeds can be achieved using a limited number of spectral bands (8 bands) with an overall classification accuracy of greater than 85%. When the number of spectral bands increased to 15, the classification accuracy was improved to greater than 90%; further increasing the number of bands did not significantly improve the accuracy. Bands in the red edge region of plant spectra had strong discriminant capability. In terms of classification features, wavelet coefficients outperformed raw spectral bands when there were a limited number of variables. However, the difference between the two was minimal when the number of variables increased to a certain level. Among different discrimination methods, SVM, which is capable of nonlinear classification, performed better.

A Lithium-ion Battery RUL Prediction Method Considering the Capacity Regeneration Phenomenon

Prediction of Remaining Useful Life (RUL) of lithium-ion batteries plays a significant role in battery health management. Battery capacity is often chosen as the Health Indicator (HI) in research on lithium-ion battery RUL prediction. In the rest time of batteries, capacity will produce a certain degree of regeneration phenomenon, which exists in the use of each battery. Therefore, considering the capacity regeneration phenomenon in RUL prediction of lithium-ion batteries is helpful to improve the prediction performance of the model. In this paper, a novel method fusing the wavelet decomposition technology (WDT) and the Nonlinear Auto Regressive neural network (NARNN) model for predicting the RUL of a lithium-ion battery is proposed. Firstly, the multi-scale WDT is used to separate the global degradation and local regeneration of a battery capacity series. Then, the RUL prediction framework based on the NARNN model is constructed for the extracted global degradation and local regeneration. Finally, the two parts of the prediction results are combined to obtain the final RUL prediction result. Experiments show that the proposed method can not only effectively capture the capacity regeneration phenomenon, but also has high prediction accuracy and is less affected by different prediction starting points.