Stacked Long Short Term Memory Model Research Articles

Evaluation of Deep Learning Models for Predicting the Concentration of Air Pollutants in Urban Environments

Air pollution is an issue of great concern globally due to the risks to the health of humanity, animals, and ecosystems. On the one hand, air quality monitoring systems allow for determining the concentration level of air pollutants and health risks through an air quality index (AQI). On the other hand, accurate future predictions of air pollutant concentration levels can provide valuable information for data-driven decision-making to reduce health risks from short- and long-term exposure when indicators exceed permissible limits. In this paper, five deep learning architectures are evaluated to predict the concentration of particulate matter pollutants (in their fractions PM2.5 and PM10) and carbon monoxide (CO) in consecutive hours. The proposed prediction models are based on recurrent neural networks (RNNs), long short-term memory (LSTM), vanilla LSTM, Stacked LSTM, Bi-LSTM, and encoder–decoder LSTM networks. Moreover, a methodology is presented to guide the construction of the prediction model, encompassing raw data processing, model design and optimization, and neural network training, testing, and evaluation. The results underscore the precision and reliability of the Stacked LSTM model in predicting the hourly concentration level for PM2.5, with an RMSE of 3.4538 μg/m3. Similarly, the encoder–decoder LSTM model accurately predicts the concentration level for PM10 and CO, with an RMSE of 3.2606 μg/m3 and 2.1510 ppm, respectively. These evaluations, with their minimal differences in error metrics and coefficient of determination, validate the effectiveness and superiority of the deep learning models over other reference models, instilling confidence in their potential.

Machine Learning Algorithms for Power System Sign Classification and a Multivariate Stacked LSTM Model for Predicting the Electricity Imbalance Volume

The energy transition to a cleaner environment has been a concern for many researchers and policy makers, as well as communities and non-governmental organizations. The effects of climate change are evident, temperatures everywhere in the world are getting higher and violent weather phenomena are more frequent, requiring clear and firm pro-environmental measures. Thus, we will discuss the energy transition and the support provided by artificial intelligence (AI) applications to achieve a cleaner and healthier environment. The focus will be on applications driving the energy transition, the significant role of AI, and collective efforts to improve societal interactions and living standards. The price of electricity is included in almost all goods and services and should be affordable for the sustainable development of economies. Therefore, it is important to model, anticipate and understand the trend of electricity markets. The electricity price includes an imbalance component which is the difference between notifications and real-time operation. Ideally it is zero, but in real operation such differences are normal due to load variation, lack of renewable energy sources (RES) accurate prediction, unplanted outages, etc. Therefore, additional energy has to be produced or some generating units are required to reduce generation to balance the power system. Usually, this activity is performed on the balancing market (BM) by the transmission system operator (TSO) that gathers offers from generators to gradually reduce or increase the output. Therefore, the prediction of the imbalance volume along with the prices for deficit and surplus is of paramount importance for producers’ decision makers to create offers on the BM. The main goal is to predict the imbalance volume and minimize the costs that such imbalance may cause. In this chapter, we propose a method to predict the imbalance volume based on the classification of the imbalance sign that is inserted into the dataset for predicting the imbalance volume. The imbalance sign is predicted using several classifiers and the output of the classification is added to the input dataset. The rest of the exogenous variables are shifted to the values from previous day d − 1. Therefore, the input variables are either predicted (like the imbalance sign) or are known from d − 1. Several metrics, such as mean average percentage error (MAPE), determination coefficient R2 and mean average error (MAE) are calculated to assess the proposed method of combining classification machine learning (ML) algorithms and recurrent neural networks (RNN) that memorize variations, namely long short-term memory (LSTM) model.

Enhancing Stock Price Prediction Using Stacked Long Short-Term Memory

This research explores the Stacked Long Short-Term Memory (LSTM) model for stock price prediction using a dataset obtained from Yahoo Finance. The main objective is to assess the effectiveness of the model in capturing stock price patterns and making accurate predictions. The dataset consists of stock prices for the top 10 companies listed in the Indonesia Stock Exchange from July 6, 2015, to October 14, 2021. The model is trained and evaluated using metrics such as RMSE, MAE, MAPE, and R2. The average values of these metrics for the predictions indicate promising results, with an average RMSE of 0.00885, average MAE of 0.00800, average MAPE of 0.02496, and an average R2 of 0.9597. These findings suggest that the Stacked LSTM model can effectively capture stock price patterns and make accurate predictions. The research contributes to the field of stock price prediction and highlights the potential of deep learning techniques in financial forecasting.

Combinational Framework for Classification of Bearing Faults in Rotating Machines

Abstract In rotating machines, roller bearings are important and prone to frequent faults. Hence, accurate classification of bearing faults is significant in the maintenance of machines. Toward this, a framework using the combination of signal processing, machine learning, and deep learning algorithms has been proposed in contrast to traditional approaches for the accurate identification of bearing faults. The benefits of each algorithm have been reaped in the proposed framework to overcome challenges met in fault identification. In this, ensemble empirical mode decomposition is applied on bearing vibration signals to reduce nonstationarity and noise. The 12 intrinsic mode function (IMF) signals of 24k length obtained for three bearing conditions at four different speeds constituted feature space of dimension [36*8*24,000]. IMFs that have the highest correlation coefficient with raw vibration signals are selected as features [3*8*24,000], and intelligent algorithms are applied. Application of principal component analysis on selected IMF feature space resulted in extraction of significant feature space retaining temporal characteristics along two major components [3*2*24,000]. Considering the temporal dependence of faults in signals, a stacked long short-term memory (LSTM) deep network is chosen and trained with extracted features to improve fault classification. The performance of this developed framework has been evaluated for different metrics of the stacked LSTM model. The proposed framework also satisfactorily surpassed the performance of the stacked LSTM model trained with raw data, capable of auto-feature learning. The comparative results inclusive of models in relevant literature illustrate the efficacy of developed combinational framework in handling dynamic vibration data for precise classification of bearing faults.

A Stacked Long Short-Term Memory Approach for Predictive Blood Glucose Monitoring in Women with Gestational Diabetes Mellitus.

Gestational diabetes mellitus (GDM) is a subtype of diabetes that develops during pregnancy. Managing blood glucose (BG) within the healthy physiological range can reduce clinical complications for women with gestational diabetes. The objectives of this study are to (1) develop benchmark glucose prediction models with long short-term memory (LSTM) recurrent neural network models using time-series data collected from the GDm-Health platform, (2) compare the prediction accuracy with published results, and (3) suggest an optimized clinical review schedule with the potential to reduce the overall number of blood tests for mothers with stable and within-range glucose measurements. A total of 190,396 BG readings from 1110 patients were used for model development, validation and testing under three different prediction schemes: 7 days of BG readings to predict the next 7 or 14 days and 14 days to predict 14 days. Our results show that the optimized BG schedule based on a 7-day observational window to predict the BG of the next 14 days achieved the accuracies of the root mean square error (RMSE) = 0.958 ± 0.007, 0.876 ± 0.003, 0.898 ± 0.003, 0.622 ± 0.003, 0.814 ± 0.009 and 0.845 ± 0.005 for the after-breakfast, after-lunch, after-dinner, before-breakfast, before-lunch and before-dinner predictions, respectively. This is the first machine learning study that suggested an optimized blood glucose monitoring frequency, which is 7 days to monitor the next 14 days based on the accuracy of blood glucose prediction. Moreover, the accuracy of our proposed model based on the fingerstick blood glucose test is on par with the prediction accuracies compared with the benchmark performance of one-hour prediction models using continuous glucose monitoring (CGM) readings. In conclusion, the stacked LSTM model is a promising approach for capturing the patterns in time-series data, resulting in accurate predictions of BG levels. Using a deep learning model with routine fingerstick glucose collection is a promising, predictable and low-cost solution for BG monitoring for women with gestational diabetes.

Air-traffic Flow Prediction with Deep Learning: A Case Study for Diyarbakır Airport

Aviation industry develops rapidly. So the continuous growth of the aviation, accurate predictions play a crucial role in managing air traffic and optimizing airport operations. The prediction process involves various factors such as weather conditions, airport traffic, flight schedules, and historical data. Advanced techniques like machine learning contribute to enhancing the accuracy of predictions. In this context, air traffic data belonging to Diyarbakır province were utilized to predict the number of arrival aircraft to the airport using both traditional Autoregressive (AR) model and deep learning architecture, specifically the stacked Long Short-Term Memory (LSTM) model. The results indicate that the stacked LSTM model outperformed the AR model in terms of air traffic estimation. The AR model had a quite poorly MSE value of 48043.35 and an RMSE value of 219.18, while the stacked LSTM model achieved a significantly higher MSE value of 0.03 and an RMSE value of 0.17. The lower MSE values obtained by the stacked LSTM model indicate its ability to make more accurate predictions compared to the AR model. The stacked LSTM model's predictions were closer to the actual values, resulting in a more realistic estimation of air traffic. Accurate predictions enable efficient resource management, passenger planning, and airport security measures. Continuous efforts in predicting aircraft landings are necessary for the effective functioning of the aviation industry. In this study highlights the importance of predicting the number of aircraft landings at airports.

Predicting the Karachi Stock Price index with an Enhanced multi-layered Sequential Stacked Long-Short-Term Memory Model

The study proposes the use of a stacked Long-Short-Term Memory (LSTM) model to predict the KSE-100 stock exchange trend and provides a comprehensive review of the literature on deep learning models and time series forecasting in the stock market. The study's findings suggest that the stacked LSTM model outperforms other models in terms of prediction accuracy. The study's contribution lies in its approach to improving the accuracy of stock price prediction using deep learning models. The stacked LSTM model architecture is a novel approach that provides better results than other traditional time series forecasting models. Furthermore, the study's use of hyper-parameter optimization techniques demonstrates the importance of model tuning for improving performance intended for accurate time series forecasting in the financial market. The study's results have practical implications for investors, who can use the stacked LSTM model to make informed decisions about buying or selling stocks in the KSE-100. The model's ability to predict stock prices accurately can help investors maximize their profits and minimize their losses. Hence, the proposed stacked LSTM model can effectively predict stock prices in the KSE-100 and can assist investors in making informed decisions in the stock market.

Trend Analysis and Prediction of Rainfall Using Deep Learning Models in Three Sub-Divisions of Karnataka

Precise estimation of rainfall is a crucial and challenging task in environmental science. It involves the use of advanced and powerful models to forecast non-linear and dynamic changes in rainfall. Deep learning, a recently developed method for handling vast amounts of data and resolving complex problems, has proven to be an effective tool for rainfall forecasting. In this study, we applied various deep learning models such as Multilayer Perceptron (MLP), Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), Stacked LSTM, Gated Recurrent Units (GRUs), and a traditional model called Autoregressive Integrated Moving Average (ARIMA), to forecast monthly rainfall data (mm) for three regions of Karnataka: Coastal Karnataka, North Interior Karnataka (NIK), and South Interior Karnataka (SIK). Trend analysis was conducted using the Mann-Kendall trend test (MK test) and the Seasonal Mann-Kendall trend test, along with Sen's Slope Estimator, to determine trends and slope magnitudes. The results showed that deep learning models perform better than traditional methods in forecasting rainfall. The performance of different models was evaluated using forecasting evaluation criteria and found that the LSTM model performed best for Coastal Karnataka, with an RMSE value of 149.45, while the Bi-LSTM model performed best for NIK, with an RMSE value of 32.57, and the Stacked LSTM model performed best for SIK, with an RMSE value of 45.33. Therefore, deep learning models can be effectively used to predict rainfall data with greater accuracy.

Long Short-Term Memory-based Models for Sleep Quality Prediction from Wearable Device Time Series Data

Several studies suggest that sleep quality is associated with physical activities. Moreover, deep sleep time can be used to determine the sleep quality of an individual. In this work, we aim to find the association between physical activities and deep sleep time by modeling the time series data such as heart rate and a number of steps captured from a commercial wearable device. Our previous study demonstrates that deep learning-based time series modeling is well suited for our problem since the temporal patterns in the two physical parameters need to be captured to obtain more accurate results. We first preprocess our series data to have a time-step size of 10 minutes. To improve our previous effort in this modeling, we compare four different variants of Long Short-Term Memory (LSTM)-based models, ranging from single input to dual input models. Our result shows that the simple stacked LSTM model performs better for our data because the remaining models suffer from overfitting due to a larger number of the trained parameters.

Regional metal pollution risk assessment based on a long short-term memory model: A case study of the South Altai Mountain mining area, China

The discharge of wastewater and waste rock in mining production activities is a significant hidden cause of soil heavy metal pollution. The accumulation of heavy metals in soil occurs through a variety of processes, and exposure to these metals can permanently damage the human body. Due to multiple factors, such as the formation causes, sources, and distribution trends of heavy metals in mineral resources, reasonably applying different machine learning methods to monitor and evaluate heavy metal pollution remains challenging. In this paper, we choose the copper mining area in the southern Altai Mountains of China as the study area, and 19 different types of spatial data are uniformly managed in a distributed database to improve monitoring efficiency. Furthermore, we propose a heavy metal pollution evaluation framework based on a stacked long short-term memory (LSTM) model, which considers spatial data correlations and extracts spatial clustering features. Information is screened through state updating in the framework, the short-term memory features of long sequences are extracted, and an effective prediction model is established. The results show that 26 of the 31 mining occurrences in the study area are in moderate- and high-pollution-risk grids, suggesting that the spatial distribution of copper mines is consistent with the predicted spatial distribution of pollution risk. Overall, these results show that using the optimized stacked LSTM model to integrate multisource geological features and mine the internal rules of feature information has a positive effect on improving the risk assessment of heavy metal pollution.

Application of Three-Layer Stacked LSTM Model Assisted by Educational Robots in Children’s Learning

To enable educational robots to better accompany children’s learning and growth, first, the encoder part of the image description model is established based on the convolutional neural network (CNN), and the decoder part of the image description model is built by the long short-term memory (LSTM). The attention mechanism is combined with the input stage, LSTM, and the output stage of the model. Second, three-layer stacked LSTM is used to improve and optimize the model. Artificial intelligence (AI) challenger data set is used to test the constructed model, and bilingual evaluation understudy (BLEU) and metric for evaluation of translation with explicit ordering (METEOR) are used to evaluate the model. Finally, 48 children are used as the research objects to study the application effect of the model. The results show that the BLEU-1, BLEU-2, BLEU-3, and BLEU-4 values of the constructed model are higher than those of the neural image caption (NIC) model during the entire iteration process. This model begins to converge when the number of iterations is 20,000. However, the NIC model starts to converge when the number of iterations is 45,000. The constructed model has a faster convergence speed. The scores of BLEU-1, BLEU-2, BLEU-3, and BLEU-4 of the constructed model are 0.687, 0.54, 0.432, and 0.348, respectively, and the score of METEOR is 0.187, which is higher than other models. In the actual application, the learning efficiency and effect of children who use this model for learning are better than the traditional teaching effect. The research results can help to improve the auxiliary role of educational robots in children’s learning, which has very important practical significance.

Predicting machine's performance record using the stacked long short-term memory (LSTM) neural networks.

PurposeThe record of daily quality control (QC) items shows machine performance patterns and potentially provides warning messages for preventive actions. This study developed a neural network model that could predict the record and trend of data variations quantitively.Methods and materialsThe record of 24 QC items for a radiotherapy machine was investigated in our institute. The QC records were collected daily for 3 years. The stacked long short‐term memory (LSTM) model was used to develop the neural network model. A total of 867 records were collected to predict the record for the next 5 days. To compare the stacked LSTM, the autoregressive integrated moving average model (ARIMA) was developed on the same data set. The accuracy of the model was quantified by the mean absolute error (MAE), root‐mean‐square error (RMSE), and coefficient of determination (R 2). To validate the robustness of the model, the record of four QC items was collected for another radiotherapy machine, which was input into the stacked LSTM model without changing any hyperparameters and ARIMA model.ResultsThe mean MAE, RMSE, andR2 with 24 QC items were 0.013, 0.020, and 0.853 in LSTM, while 0.021, 0.030, and 0.618 in ARIMA, respectively. The results showed that the stacked LSTM outperforms the ARIMA. Moreover, the mean MAE, RMSE, andR2 with four QC items were 0.102, 0.151, and 0.770 in LSTM, while 0.162, 0.375, and 0.550 in ARIMA, respectively.ConclusionsIn this study, the stacked LSTM model can accurately predict the record and trend of QC items. Moreover, the stacked LSTM model is robust when applied to another radiotherapy machine. Predicting future performance record will foresee possible machine failure, allowing early machine maintenance and reducing unscheduled machine downtime.

DGSLSTM: Deep Gated Stacked Long Short-Term Memory Neural Network for Traffic Flow Forecasting of Transportation Networks on Big Data Environment.

Deep learning and big data techniques have become increasingly popular in traffic flow forecasting. Deep neural networks have also been applied to traffic flow forecasting. Furthermore, it is difficult to determine whether neural networks can be used for accurate traffic flow prediction. Moreover, since the network model is poorly structured and the parameter optimization technique is inappropriate, the traffic flow prediction is inaccurate because of the lack of certainty. The proposed system overcomes these problems by combining multiple simple recurrent long short-term memory (LSTM) neural networks with time traits to predict traffic flow using a deep gated stacked neural network. To deepen the model, the hidden layers have been trained using an unsupervised layer-by-layer approach. This approach provides a systematic representation of the time series data. A systematic representation of hidden layers improves the accuracy of time series forecasting by capturing information at multiple levels. Furthermore, it emphasizes the importance of model structure, random weight initialization, and hyperparameters used in stacked LSTM to enhance predictive performance. The prediction efficacy of the deep gated stacked LSTM model is compared with that of the gated recurrent unit model and the stacked autoencoder model.

A Novel Stacked Long Short-Term Memory Approach of Deep Learning for Streamflow Simulation

Rainfall-Runoff simulation is the backbone of all hydrological and climate change studies. This study proposes a novel stochastic model for daily rainfall-runoff simulation called Stacked Long Short-Term Memory (SLSTM) relying on machine learning technology. The SLSTM model utilizes only the rainfall-runoff data in its modelling approach and the hydrology system is deemed a blackbox. Conversely, the distributed and physically-based hydrological models, e.g., SWAT (Soil and Water Assessment Tool) preserve the physical aspect of hydrological variables and their inter-relations while taking a wide range of data. The two model types provide specific applications that interest modelers, who can apply them according to their project specification and objectives. However, sparse distribution of point-data may hinder physical models’ performance, which may not be the case in data-driven models. This study proposes a specific SLSTM model and investigates the SLSTM and SWAT models’ data dependency in terms of their spatial distribution. The study was conducted in the two distinct river basins of Samarahan and Trusan, Malaysia, with over 20 years of hydro-climate data. The Trusan basin’s rain gauges are scattered downstream of the basin outlet and Samarahan’s are located around the basin, with one station within each basin’s limits. The SWAT was developed and calibrated following its general modelling approach, however, the SLSTM performance was also tested using data preprocessing with principal component analysis (PCA). Results showed that the SWAT performance for daily streamflow simulation at Samarahan has been superior to that of Trusan. Both the SLSTM and PCA-SLSTM models, however, showed better performance at Trusan with PCA-SLSTM outperforming the SLSTM. This demonstrates that the SWAT model is greatly affected by the spatial distribution of its input data, while data-driven models, irrespective of the spatial distribution of their entry data, can perform well if the data adequacy condition is met. However, considering the structural difference between the two models, each has its specific application in a water resources context. The study of catchments’ response to changes in the hydrology cycle requires a physically-based model like SWAT with proper spatial and temporal distribution of its entry data. However, the study of a specific phenomenon without considering the underlying processes can be done using data-driven models like SLSTM, where improper spatial distribution of data cannot be a restricting factor.

Application of stacked and bidirectional long short-term memory deep learning models for wind speed forecasting at an offshore site

ABSTRACT Very short-term offshore wind speed forecasting by application of Stacked long short-term memory (LSTM) and Bidirectional LSTM deep learning models is done in this work. Wind speed data of two different offshore sites located in two different continents are used for testing the models. Performance is measured on the basis of accuracy of forecasting and computational time. The effectiveness of Stacked LSTM and Bidirectional LSTM models is also validated by comparing their performance with convolutional neural network, convolutional neural network-long short-term memory network, multi-layer perceptron, and rolling forecasting auto regressive integrated moving average models. Results of forecasting error confirm that Stacked LSTM model is better than other compared models in forecasting very short-term offshore wind speed. Mean absolute percentage error (MAPE) of wind speed forecasting by Stacked LSTM model is 4.59% at Anholt (Denmark) and 3.62% at Dhanushkodi (India) sites. From comparison of MAPE of Stacked LSTM model with that of eight other latest existing models in literature, it can be concluded that Stacked LSTM model is superior to many other existing models.

An Effective Hybrid NARX-LSTM Model for Point and Interval PV Power Forecasting

This paper proposes an effective Photovoltaic (PV) Power Forecasting (PVPF) technique based on hierarchical learning combining Nonlinear Auto-Regressive Neural Networks with exogenous input (NARXNN) with Long Short-Term Memory (LSTM) model. First, the NARXNN model acquires the data to generate a residual error vector. Then, the stacked LSTM model, optimized by Tabu search algorithm, uses the residual error correction associated with the original data to produce a point and interval PVPF. The performance of the proposed PVPF technique was investigated using two real datasets with different scales and locations. The comparative analysis of the NARX-LSTM with twelve existing benchmarks confirms its superiority in terms of accuracy measures. In summary, the proposed NARX-LSTM technique has the following major achievements: 1) Improves the prediction performance of the original LSTM and NARXNN models; 2) Evaluates the uncertainties associated with point forecasts with high accuracy; 3) Provides a high generalization capability for PV systems with different scales. Numerical results of the comparison of the proposed NARX-LSTM method with two real-world PV systems in Australia and USA demonstrate its improved prediction accuracy, outperforming the benchmark approaches with an overall normalized Rooted Mean Squared Error (nRMSE) of 1.98% and 1.33% respectively.

Fast Visualization of 3D Massive Data Based on Improved Hilbert R-Tree and Stacked LSTM Models

With the explosive growth of scientific data, significant challenges exist with respect to the interaction of large volumetric datasets. To solve these problems, we propose a visualization algorithm based on the Hilbert R-tree improved by the clustering algorithm using K-means (CUK) and a stacked long short-term memory (LSTM) model to quickly display massive data. First, we use the Hilbert R-tree optimized by the CUK to quickly store unevenly distributed data and build a fast index for the massive data. Then, we determine the position of the current point of view and use the stacked LSTM model to predict the next point of view. According to the location of two points, we divide the visible area. Finally, according to the preloading strategy, we import the data into the cache area of the graphics processing unit (GPU), which greatly realizes smoother rendering data and large-scale data interaction visualization. The experimental results showed that the proposed algorithm can quickly and accurately draw large volumetric data with high quality while guaranteeing rendering quality.

Long lead time drought forecasting using lagged climate variables and a stacked long short-term memory model

Drought forecasting with a long lead time is essential for early warning systems and risk management strategies. The use of machine learning algorithms has been proven to be beneficial in forecasting droughts. However, forecasting at long lead times remains a challenge due to the effects of climate change and the complexities involved in drought assessment. The rise of deep learning techniques can solve this issue, and the present work aims to use a stacked long short-term memory (LSTM) architecture to forecast a commonly used drought measure, namely, the Standard Precipitation Evaporation Index. The model was then applied to the New South Wales region of Australia, with hydrometeorological and climatic variables as predictors. The multivariate interpolated grid of the Climatic Research Unit was used to compute the index at monthly scales, with meteorological variables as predictors. The architecture was trained using data from the period of 1901–2000 and tested on data from the period of 2001–2018. The results were then forecasted at lead times ranging from 1 month to 12 months. The forecasted results were analysed in terms of drought characteristics, such as drought intensity, drought onset, spatial extent and number of drought months, to elucidate how these characteristics improve the understanding of drought forecasting. The drought intensity forecasting capability of the model used two statistical metrics, namely, the coefficient of determination (R2) and root-mean-square error. The variation in the number of drought months was examined using the threat score technique. The results of this study showed that the stacked LSTM model can forecast effectively at short-term and long-term lead times. Such findings will be essential for government agencies and can be further tested to understand the forecasting capability of the presented architecture at shorter temporal scales, which can range from days to weeks.

Ensemble Model Based on Stacked Long Short-Term Memory Model for Cycle Life Prediction of Lithium–Ion Batteries

To meet the target value of cycle life, it is necessary to accurately assess the lithium–ion capacity degradation in the battery management system. We present an ensemble model based on the stacked long short-term memory (SLSTM), which is used to predict the capacity cycle life of lithium–ion batteries. The ensemble model combines LSTM with attention and gradient boosted regression (GBR) models to improve prediction accuracy, where these individual prediction values are used as input to the SLSTM model. Among 13 cells, single and multiple cells were used as the training set to verify the performance of the proposed model. In seven single-cell experiments, 70% of the data were used for model training, and the rest of the data were used for model validation. In the second experiment, one cell or two cells were used for model training, and other cells were used as test data. The results show that the proposed method is superior to individual and traditional integrated learning models. We used Monte Carlo dropout techniques to estimate variance and obtain prediction intervals. In the second experiment, the average absolute percentage errors for GBR, LSTM with attention, and the proposed model are 28.6580, 1.7813, and 1.5789, respectively.

Stacked long short-term memory model for proton exchange membrane fuel cell systems degradation

Proton exchange membrane fuel cell (PEMFC) systems have numerous applications such as transportation, portable power generation, and military. In this study, we propose a stacked long-short term memory (S-LSTM) model for fitting the degradation of a PEMFC system. Moreover, the proposed model provides the remaining useful life (RUL) prediction. A stacked LSTM architecture with dropout parameters can improve the prediction accuracy of the fuel cell degradation. We optimize the hyper parameters of the S-LSTM model using a differential evolution algorithm. The ageing test conditions of two PEMFC systems are carried by a fixed current and a ripple current, respectively. The results indicate that the S-LSTM model outperforms the other models in the RUL prediction of the PEMFC degradation in terms of mean absolute percent error and root mean square error.