Long Short-term Memory Neural Network Research Articles

Using Long Short-Term Memory (LSTM) recurrent neural networks to classify unprocessed EEG for seizure prediction

ObjectiveSeizure prediction could improve quality of life for patients through removing uncertainty and providing an opportunity for acute treatments. Most seizure prediction models use feature engineering to process the EEG recordings. Long-Short Term Memory (LSTM) neural networks are a recurrent neural network architecture that can display temporal dynamics and, therefore, potentially analyze EEG signals without performing feature engineering. In this study, we tested if LSTMs could classify unprocessed EEG recordings to make seizure predictions.MethodsLong-term intracranial EEG data was used from 10 patients. 10-s segments of EEG were input to LSTM models that were trained to classify the EEG signal. The final seizure prediction was generated from 5 outputs of the LSTM model over 50 s and combined with time information to account for seizure cycles.ResultsThe LSTM models could make predictions significantly better than a random predictor. When compared to other publications using the same dataset, our model performed better than several others and was comparable to the best models published to date. Furthermore, this framework could still produce predictions significantly better than chance when the experimental paradigm design was altered, without the need to reperform feature engineering.SignificanceRemoving the need to perform feature engineering is an advancement on previously published models. This framework can be applied to many different patients’ needs and a variety of acute interventions. Also, it opens the possibility of personalized seizure predictions that can be altered to meet daily needs.

Statistical and machine learning approaches for estimating pollution of fine particulate matter (PM2.5) in Vietnam

This study aims to predict fine particulate matter (PM2.5) pollution in Ho Chi Minh City, Vietnam, using autoregressive integrated moving average (ARIMA), linear regression (LR), random forest (RF), long short-term memory (LSTM), bidirectional LSTM (Bi-LSTM), and convolutional neural network (CNN) combining Bi-LSTM (CNN+Bi-LSTM). Two experiments were set up: the first one used data from 2018–2020 and 2021 as training and test data, respectively. Data from 2018–2021 and 2022 were used as training and test data for the second experiment, respectively. Consequently, ARIMA showed the worst performance, while CNN+Bi-LSTM achieved the best accuracy, with an R² of 0.70 and MAE, MSE, RMSE, and MAPE of 5.37, 65.4, 8.08 µg/m³, and 29%, respectively. Additionally, predicted air quality indexes (AQIs) of PM2.5 were matched the observed ones up to 96%, reflecting the application of predicted concentrations for AQI computation. Our study highlights the effectiveness of machine learning model in monitoring of air pollution.

Real-time fault detection for IIoT facilities using GA-Att-LSTM based on edge-cloud collaboration

With the rapid development of Industrial Internet of Things (IIoT) technology, various IIoT devices are generating large amounts of industrial sensor data that are spatiotemporally correlated and heterogeneous from multi-source and multi-domain. This poses a challenge to current detection algorithms. Therefore, this paper proposes an improved long short-term memory (LSTM) neural network model based on the genetic algorithm, attention mechanism and edge-cloud collaboration (GA-Att-LSTM) framework is proposed to detect anomalies of IIoT facilities. Firstly, an edge-cloud collaboration framework is established to real-time process a large amount of sensor data at the edge node in real time, which reduces the time of uploading sensor data to the cloud platform. Secondly, to overcome the problem of insufficient attention to important features in the input sequence in traditional LSTM algorithms, we introduce an attention mechanism to adaptively adjust the weights of important features in the model. Meanwhile, a genetic algorithm optimized hyperparameters of the LSTM neural network is proposed to transform anomaly detection into a classification problem and effectively extract the correlation of time-series data, which improves the recognition rate of fault detection. Finally, the proposed method has been evaluated on a publicly available fault database. The results indicate an accuracy of 99.6%, an F1-score of 84.2%, a precision of 89.8%, and a recall of 77.6%, all of which exceed the performance of five traditional machine learning methods.

Integrated Method of Future Capacity and RUL Prediction for Lithium‐Ion Batteries Based on CEEMD‐Transformer‐LSTM Model

ABSTRACTAccurately predict the remaining useful life (RUL) of lithium‐ion batteries for energy storage is of critical significance to ensure the safety and reliability of electric vehicles, which can offer efficient early warning signals in a timely manner. Considering nonlinear changes in the aging trajectory of lithium‐ion batteries, a method for predicting the RUL of lithium‐ion batteries was proposed in this study based on a complementary ensemble empirical mode decomposition (CEEMD) as well as transformer and long short‐term memory (LSTM) neural network dual‐drive machine learning model. First, the CEEMD algorithm was adopted to decompose the raw aging data of lithium‐ion batteries into intrinsic mode function (IMF) sequences and residual sequence, where the number of modal layers was produced by the proposed posterior feedback entropy and relevance (PFER) method. Second, prediction models of LSTM and transformer neural networks were established to predict IMF and residual sequences. Simultaneously, the sparrow search algorithm (SSA) was used to obtain the optimal value of the hyperparameter learning rate for the RUL prediction model. Finally, the predicted IMF and residual sequences were combined to comprehensively calculate the future lifespan aging trajectory of lithium‐ion batteries. The aging data of two groups of lithium‐ion batteries were obtained from the CALCE at the University of Maryland as well as the laboratory at AQNU University to verify the proposed method. Experimental results demonstrated that the proposed method can effectively predict the RUL of lithium‐ion batteries; moreover, it exhibited better robustness and generalization ability.

Enhanced Tuberculosis Detection in Chest X-Rays Using Optimal Feature Selection with Hybrid Binary Particle Swarm Optimization (HBPSO) and a Dual Classifier Framework Combining LSTM and Spiking Neural Networks

Tuberculosis (TB) remains a significant global health challenge, particularly in developing countries, where rapid and accurate diagnosis is crucial for effective treatment. Recent advancements in machine learning and image processing have shown promise in enhancing the detection of TB from chest X-ray images. This research introduces a novel approach that integrates a hybrid feature selection method with a dual classifier framework to improve the accuracy of tuberculosis detection. Feature extraction is conducted using Multi-Scale Local Binary Patterns (MSLBP), Speeded-Up Robust Features (SURF) with Bag of Words (BoW), and Cartoon Texture Features, combining them into a comprehensive feature vector. The Hybrid Binary Particle Swarm Optimization (HBPSO) is employed for optimal feature selection, reducing redundancy and improving classification efficiency. Classification is performed using a dual framework involving a Long Short-Term Memory (LSTM) network and a Spiking Neural Network (SNN), which work in tandem to provide robust predictions. Majority voting is utilized to finalize predictions, ensuring high accuracy and adaptability across varying TB presentations. Extensive evaluation on publicly available datasets shows that this hybrid approach significantly outperforms existing methods, achieving a highest accuracy of 99.73%, along with superior precision, recall, and F1 scores.

An improved prediction method for BDS-3 SISA parameters and the preliminary performance evaluation

Abstract The ephemeris accuracy parameters broadcast in the ephemeris messages are commonly utilized as an alarm threshold for the signal-in-space ranging errors (SISRE) of the Global Navigation Satellite System (GNSS). Accurately predicting the orbit and clock errors of broadcast ephemeris, and converting the prediction errors into broadcast ephemeris accuracy parameters using a robust model, is important for the integrity service of GNSS. The BeiDou Navigation Satellite System (BDS-3) broadcasts four ephemeris accuracy parameters, namely SIS A oe , SIS A ocb , SIS A oc1 , and SIS A oc2 , to represent the standard deviations of the satellite orbit and clock prediction errors in the broadcast ephemeris. The probability that the broadcast ephemeris SISRE exceeds 4.42 times the signal-in-space accuracy (SISA) without triggering an alarm in time is expected to be less than 1 × 10−5. In addition to accurately enveloping the broadcast ephemeris SISRE, SISA should also reduce the over-envelopment margin over SISRE to accurately express the system service accuracy. Thus, an accurate prediction of orbit and clock errors in the broadcast ephemeris is crucial for calculating the SISA. An orbit and clock error prediction method based on the Long Short-Term Memory (LSTM) neural network is proposed in this study. This method uses historical satellite orbit and clock errors, derived from the difference between precise and broadcast ephemeris, to train an LSTM neural network. The trained LSTM model is then used to predict the orbit and clock errors for the next 24 h. The predicted errors are used to calculate the corresponding SISA for the broadcast ephemeris. The results indicate that the 24 h accuracies of the predicted orbit and clock errors are less than 0.06 m and 0.25 m, respectively. The calculated SISA can completely envelop the actual broadcast ephemeris SISRE, achieving 100% coverage while also decreasing the over-envelopment margin by 80%. Furthermore, the horizontal protection level and vertical protection level calculated using the corresponding SISA are reduced by approximately 14%, meeting the integrity requirements set by the International Civil Aviation Organization for Approach with Vertical Guidance I. The proposed method enhances the timeliness of SISA parameter calculation and the integrity and availability of the BDS-3 service, serving as a crucial reference value for safety critical applications such as civil aviation services.

Physics-guided fault diagnosis method for proton exchange membrane fuel cells based on LSTM neural network

In this paper, we propose a novel hybrid method for fault diagnosis of Proton Exchange Membrane Fuel Cells (PEMFCs) based on the combination of a physics-based model and a long short-term memory (LSTM) neural network. By incorporating the physics-based model in the fault diagnosis algorithm, we can access to several process variables not directly measured through sensors but related to the state of the PEMFC stack. The model estimates are subsequently combined with signals measured on the PEMFC stack and inputted to a LSTM neural network. The performance of the physics-guided LSTM is evaluated on an extensive dataset comprising a thousand hours of operation of a PEMFC stack under dynamic load profiles, proving the enhanced capability of the proposed fault diagnosis method in capturing complex fault patterns. Furthermore, the effectiveness of the proposed method in dealing with PEMFC aging is tested by using data from the initial phase of stack operation for algorithm training and reserving the data from the aged stack operation for the testing phase. The experimental results reveal that the proposed physics-guided LSTM method allows for a significant amelioration over purely data-driven LSTM.

Core temperature estimation of lithium-ion battery based on numerical model fusion deep learning

Temperature has a critical impact on the lifespan and safety of lithium batteries. This paper proposes a battery core temperature estimation method based on numerical model fused with long short-term memory (LSTM) neural network. The proposed technique extracts features from the numerical model, estimates the volume-averaged temperature by electrochemical impedance spectroscopy (EIS), uses an LSTM neural network to learn thermodynamic parameters and complex calculations, which takes advantage of the strengths of each method, and achieves accurate core temperature estimation. The effects of state of charge (SOC) and temperature on EIS are explored, impedance properties are selected on the criteria of robustness and rapidity, and the estimation of the volume-averaged temperature is achieved using the imaginary part of the impedance. The proposed method can achieve root mean squared error (RMSE) of less than 0.28 °C and mean absolute error (MAE) of less than 0.23 °C. The proposed method has advantages of high estimation accuracy and does not require an electrothermal model. It also considers the effect of ambient temperature and has a good generalization capability.

A River Water Quality Prediction Method Based on Dual Signal Decomposition and Deep Learning

Traditional single prediction models struggle to address the complexity and nonlinear changes in water quality forecasting. To address this challenge, this study proposed a coupled prediction model (RF-TVSV-SCL). The model includes Random Forest (RF) feature selection, dual signal decomposition (Time-Varying Filtered Empirical Mode Decomposition, TVF-EMD, and Sparrow Search Algorithm-Optimized Variational Mode Decomposition, SSA-VMD), and a deep learning predictive model (Sparrow Search Algorithm-Convolutional Neural Network-Long Short-Term Memory, SSA-CNN-LSTM). Firstly, the RF method was used for feature selection to extract important features relevant to water quality prediction. Then, TVF-EMD was employed for preliminary decomposition of the water quality data, followed by a secondary decomposition of complex Intrinsic Mode Function (IMF) components using SSA-VMD. Finally, the SSA-CNN-LSTM model was utilized to predict the processed data. This model was evaluated for predicting total phosphorus (TP), total nitrogen (TN), ammonia nitrogen (NH3-N), dissolved oxygen (DO), permanganate index (CODMn), conductivity (EC), and turbidity (TB), across 1, 3, 5, and 7-d forecast periods. The model performed exceptionally well in short-term predictions, particularly within the 1–3 d range. For 1-, 3-, 5-, and 7-d forecasts, R2 ranged from 0.93–0.96, 0.79–0.87, 0.63–0.72, and 0.56–0.64, respectively, significantly outperforming other comparison models. The RF-TVSV-SCL model demonstrates excellent predictive capability and generalization ability, providing robust technical support for water quality forecasting and pollution prevention.

A Predictive Approach for Lithium-Ion Battery SOH using LSTM Neural Networks Enhanced by Health Matrix Optimization

The State of Health (SOH) is a critical performance metric that characterizes the condition of lithium-ion batteries, directly influencing their lifespan and operational efficiency. In order to enhance the accuracy of SOH predictions for batteries and reduce operational risks, a novel approach has been introduced. This method, based on Health Matrix Optimization for Long Short Term Memory (LSTM) neural networks, aims to optimize the prediction process. Initially, the Spearman correlation coefficient method is employed to analyze the correlation of battery state data. Through the use of a heatmap, data points with strong correlations to SOH are identified, leading to the creation of a health feature matrix. This matrix is then utilized to fine-tune the hyperparameters of the LSTM neural network, resulting in refined approximations. Subsequently, by employing this optimized LSTM neural network, accurate predictions of the SOH for lithium-ion batteries are made. The results demonstrate a notable improvement in prediction accuracy by 35.71% and a significant increase in prediction speed by 35.5% when compared to traditional methods. This innovative approach proves to be effective in enhancing battery performance and longevity.

Accurate detection of gait events using neural networks and IMU data mimicking real-world smartphone usage

Wearable technologies such as inertial measurement units (IMUs) can be used to evaluate human gait and improve mobility, but sensor fixation is still a limitation that needs to be addressed. Therefore, aim of this study was to create a machine learning algorithm to predict gait events using a single IMU mimicking the carrying of a smartphone. Fifty-two healthy adults (35 males/17 females) walked on a treadmill at various speeds while carrying a surrogate smartphone in the right hand, front right trouser pocket, and right jacket pocket. Ground-truth gait events (e.g. heel strikes and toe-offs) were determined bilaterally using a gold standard optical motion capture system. The tri-dimensional accelerometer and gyroscope data were segmented in 20-ms windows, which were labelled as containing or not the gait events. A long-short term memory neural network (LSTM-NN) was used to classify the 20-ms windows as containing the heel strike or toe-off for the right or left legs, using 80% of the data for training and 20% of the data for testing. The results demonstrated an overall accuracy of 92% across all phone positions and walking speeds, with a slightly higher accuracy for the right-side predictions (∼94%) when compared to the left side (∼91%). Moreover, we found a median time error <3% of the gait cycle duration across all speeds and positions (∼77 ms). Our results represent a promising first step towards using smartphones for remote gait analysis without requiring IMU fixation, but further research is needed to enhance generalizability and explore real-world deployment.

Prediction of fuel cell degradation trends using long short term memory optimization algorithm based on four-module experimental reactor validation

To solve the problem of life prediction of proton exchange membrane fuel cells (PEMFCs), a novel stack with four modules was used to conduct experiments. With the consistent conditions of the experimental stack, the stack voltage data was used as a life indicator to predict the remaining life of PEMFCs and the trend of performance degradation, which was advantageous for early detection of stack operation problems and timely implementation of maintenance measures. Due to the difficulty of obtaining optimal hyperparameter combinations for traditional Long-Short Term Memory (LSTM) neural networks through limited experiments, which affects the prediction accuracy, the Grey Wolf Optimization (GWO) algorithm is introduced. This improved the accuracy of the test set for Module 1 by 10.154 %, and reduced the prediction error for remaining service life by 11.3 %. Modules 2 and 3 were validated using the optimization algorithm, the accuracy of the test set was improved by 12.289 % and 11.044 %, The prediction error for the remaining service life has been reduced by 21.17 and 28.21 h, respectively. The four-module experimental fuel cell stack can provide multiple operating conditions simultaneously to verify the accuracy and effectiveness of the hybrid prediction model proposed in this paper.

Water-to-air unmanned aerial vehicle (UAV) based rolling shutter optical camera communication (OCC) system with gated recurrent unit neural network (GRU-NN)

Underwater sensors and autonomous underwater vehicles (AUVs) are widely adopted in oceanic research activities. As the number of underwater sensors and AUVs is growing quickly, the bandwidth requirements are increasing accordingly. In this work, we put forward and demonstrate a large field-of-view (FOV) water-to-air unmanned aerial vehicle (UAV) based optical camera communication (OCC) system with gated recurrent unit neural network (GRU-NN) for the first time to the best of our knowledge. As the UAVs are embedded with complementary-metal-oxide-semiconductor (CMOS) cameras, there is no need to install OCC receivers (Rxs), reducing the deployment cost. Moreover, the large photo-sensitive area of the CMOS camera can support a large FOV OCC transmission without the need for precise optical alignment. Here, by utilizing the column matrix identification during the rolling shutter pattern decoding in the CMOS image sensor, the scintillation caused by water turbulence can be reduced. Besides, in the outdoor and windy environment, the UAV will experience significant movement caused by the wind making it very difficult to capture stable OCC frames in the CMOS image sensor. Here, we propose and demonstrate utilizing GRU-NN, which is a special realization of the recurrent neural network (RNN) with memory cells capable of learning the time-domain dependent signals. It is shown that the GRU-NN can learn effectively from successive image frames in time-domain and produce correct prediction even under the windy and unstable UAV flying environment. Experimental results reveal that the proposed GRU-NN can outperform the previous pixel-per-symbol labeling neural network (PPS-NN), and also can significantly reduce the computation time when compared with long-short-term-memory-neural-network (LSTM-NN). The proposed system can decode 4-level pulse-amplitude-modulation (PAM4) rolling shutter OCC patterns at data rates of 5.4 kbit/s and 3.0 kbit/s under clear and cloudy water, respectively, fulfilling the pre-forward error correction bit-error rate (pre-FEC BER = 3.8 × 10−3). We also demonstrate that the UAV based OCC system can support data rates of 5.4 kbit/s, 4.2 kbit/s, and 3.0 kbit/s at distances of 2.2 m, 3.2 m, and 4.2 m, respectively, at outdoor and windy environments.

Electrocardiograph analysis for risk assessment of heart failure with preserved ejection fraction: A deep learning model.

Heart failure with preserved ejection fraction (HFpEF) requires an efficient screening method. We developed a deep learning model (DLM) to screen HFpEF risk using electrocardiograms (ECGs). A cohort study was conducted utilising data from Cohorts A and B. A convolutional neural network-long short-term memory (CNN-LSTM) DLM was employed. HFpEF risk was determined by left ventricular end-diastolic pressure (LVEDP) and clinical symptoms. The DLM was trained by ECGs. LVEDP for each patient was collected through invasive left ventricular catheterisation. Cohort A and B comprised data from individuals at high risk for HFpEF (LVEDP>12mmHg) and low risk for HFpEF (LVEDP≤12mmHg). The model was trained on Cohort A and prospectively validated on Cohort B. A total of 238 patients underwent ECG and left ventricular catheterisation for model training in Cohort A, and 117 patients for validation in Cohort B. The DLM achieved 78% accuracy in assessing HFpEF risk in Cohort A, while in Cohort B, it demonstrated 78% accuracy, 71.9% specificity, and 71.7% sensitivity. In the validation Cohort B, the DLM-identified high-risk HFpEF group exhibited significantly higher prevalence of diabetes (22.03%-11.86%, P<0.01), higher BMI indices (25.92-24.22kg/cm2, P<0.01), and lower usage history of calcium channel blockers (CCB) (11.76%-28.81%, P<0.01) compared with the DLM-identified low-risk HFpEF group. Traditional HFpEF indicators, including B-type natriuretic peptide (BNP) (22-20pg/mL, P=0.71) and E/E' (8.25-8.5, P=0.66), did not exhibit significant differences between the two groups. The DLM offers an accurate, cost-effective tool for HFpEF risk assessment, potentially facilitating early detection and improved clinical management.

Photometric redshift estimation for CSST survey with LSTM neural networks

ABSTRACT Accurate estimation of photometric redshifts (photo-zs) is crucial for cosmological surveys. Various methods have been developed for this purpose, such as template fitting methods and machine learning techniques, each with its own applications, advantages, and limitations. In this study, we propose a new approach that utilizes a deep learning model based on recurrent neural networks with long short-term memory (LSTM) to predict photo-z. Unlike many existing machine learning models, our method requires only flux measurements from different observed filters as input. The model can automatically learn the complex relationships between the flux data across different wavelengths, eliminating the need for manually extracted or derived input features, thereby providing precise photo-z estimates. The effectiveness of our proposed model is evaluated using simulated data from the Chinese Space Station Telescope (CSST) sourced from the Hubble Space Telescope Advanced Camera for Surveys (HST-ACS) and the COSMOS catalogue, considering anticipated instrument effects of the future CSST. Results from experiments demonstrate that our LSTM model, compared to commonly used template fitting and machine learning approaches, requires minimal input parameters and achieves high precision in photo-z estimation. For instance, when trained on the same data set and provided only with photometric fluxes as input features, the proposed LSTM model yields one-third of the outliers $f_{\mathrm{out}}$ observed with a multi-layer perceptron neural network (MLP) model, while the normalized median absolute deviation $\rm \sigma _{NMAD}$ is only two-thirds that of the MLP model. This study presents a novel approach to accurately estimate photo-zs of galaxies using photometric data from large-scale survey projects.

Judicial decision prediction using an integrated attention based bidirectional long-short term memory and dilated skip residual convolution neural network

Judicial decision prediction using an integrated attention based bidirectional long-short term memory and dilated skip residual convolution neural network

Modeling of an inductive displacement sensor based on 1DCNN-LSTM-AT

Abstract The input-output relationship of inductive displacement sensors is typically nonlinear, which demands numerous computational resources for precise calculation. Therefore, the traditional analytical methods for the inductive displacement sensors cannot meet the real-time high-precision measurement requirements. In response to the aforementioned issues, this paper proposes an optimized Long Short-Term Memory (LSTM) neural networks based on one-dimensional convolutional neural networks (1DCNN) to modeling the input-output relationship of the inductive displacement sensor. First, the measurement principle of the inductive displacement sensors was analyzed, and the analytical model of the sensor output was derived. And the influences of the key parameters of sensors on the relationship between the induced voltage and the displacement were studied. Then, the 1DCNN-LSTM-AT network for modelling the input-output relationship of the inductive displacement sensor was studied. The spatial features of historical induced electromotive force (EMF) data generated by the induction coil were initially extracted using the 1DCNN network. And these spatial features were then utilized as input to the LSTM neural network to capture the temporal features of the historical induced EMF data. Subsequently, the spatiotemporal features of the induced EMF data were fed into the regression prediction layer to compute the displacement measurement results corresponding to the current input. Moreover, the attention mechanism was used for the 1DCNN-LSTM model to enhance the prediction accuracy and stability of the model. Finally, the experimental results demonstrate that the proposed 1DCNN-LSTM-AT model achieves an average absolute percentage error (MAPE) of 3.1%, significantly outperforming traditional models such as LSTM (29.3%), CNN (4.7%), and ANN (4.3%). This paper provides a new method for modelling the nonlinear relationships of the inductive displacement sensor, and presents a fresh perspective for research in the data processing of nonlinear sensors.

Novel state of charge estimation method of containerized Lithium–Ion battery energy storage system based on deep learning

State of charge (SOC) is a critical indicator for lithium–ion battery energy storage system. However, model-driven SOC estimation is challenging due to the coupling of internal charging and discharging processes, ion diffusion, and chemical reactions in the electrode materials. These factors result in frequent changes in their internal characteristics. Therefore, we devise a novel deep learning SOC estimation method based on the attention mechanism (A) and convolutional neural networks-long short term memory (CNN–LSTM) network model. The CNN layers are employed to extract the critical features of the battery dataset. The LSTM layers are utilized to capture the long-term dependencies in the time series data. The attention mechanism can assign attention weights to different features within the output of the convolution layer. The A–CNN–LSTM method not only amplifies the positive influence of critical knowledge but also provides a global reference. The comparative experiments of three datasets and ablation study are carried out. The root mean square error, mean absolute error, and mean absolute percentage error obtained from the A–CNN–LSTM model decreased by 32.75 %, 45.88 %, and 53.36 %, respectively, compared with the best results obtained from the existing CNN, LSTM, gate recurrent unit (GRU), CNN–LSTM, and CNN–GRU models. The proposed model demonstrates greater stability and consistency than other models, underscoring its efficacy in SOC estimation.

Fast stability enhancement of inverter‐based microgrids using NGO‐LSTM algorithm

AbstractTo improve the stability of the inverter‐based microgrid (MG), this paper employs a novel data‐driven based method to coordinately adjust control parameters of inverters in a fast local manner. During the design process, an offline eigenvalue based optimization problem that is used to calculate the optimal control parameters under various operating conditions is first constructed. In order to reduce reliance on full system information, a feature selection algorithm is utilized to extract the most relevant local measurements that influence the adjustment of each control parameter. Then, regarding local measurements as input variables and optimal control parameters as output variables, based on northern goshawk optimization (NGO) and long short‐term memory (LSTM) network, a novel deep learning algorithm is proposed to train the local parameter adjustment model (LPAM) by learning the mapping relationship between them. During the application, to guarantee the stability of MG all the time, a security region based shielding mechanism is developed, where the improper control parameter adjustment will be replaced by a safe one. The case study indicates that the proposed algorithm has better mapping accuracy than traditional LSTM neural networks and also faster calculation speed than the traditional offline optimization‐based method. The effectiveness and advantages of the proposed method are demonstrated in a modified 9‐bus MG.

Productivity forecasting of the solar still with phase change material using LSTM and GRU neural networks

Abstract Solar desalination is one of the renewable energy techniques by which freshwater can be obtained economically. Solar desalination experiments are time and resource-consuming methods; there is a need for a robust system to identify the serviceability of the solar still in a specific region. The objective of this study is to develop a forecasting model using artificial neural networks to predict freshwater productivity. The study specifically aims to compare the accuracy of Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks in forecasting the productivity of Conventional Solar Stills (CSS) and Solar Stills with Phase Change Material (SSPCM). The current investigation involved analysing the experimental outcomes of a solar still that employed phase change material (PCM) and pin fins. Palmitic acid was implemented as the energy storage material and was placed beneath the absorber plate. The neural network model was trained and validated using time-series solar still experimental data. Different statistical measures were utilised to evaluate the accuracy of Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU). The results indicate that the freshwater productivity forecasted by LSTM exhibited greater accuracy than GRU. Specifically, the coefficient of determination values for LSTM were 0.96 and 0.98 for the CSS and SSPCM, respectively, which were better than the corresponding values for GRU.