Residual Recurrent Neural Network Research Articles

Residual Echo State Networks: Residual recurrent neural networks with stable dynamics and fast learning

Residual connections have been established as a staple for modern deep learning architectures. Most of their applications are cast towards feedforward computing. In this paper, we study the architectural bias of residual connections in the context of recurrent neural networks (RNNs), specifically in the temporal dimension. We frame our discussion from the perspective of Reservoir Computing and dynamical system theory, focusing on important aspects of neural computation like memory capacity, long-term information processing, stability, and nonlinear computation capability. Experiments corroborate the striking advantage brought by temporal residual connections for a plethora of different time series processing tasks, comprehending memory-based, forecasting, and classification problems.

Automatic multi-label diagnosis of single-lead ECG using novel hybrid residual recurrent convolutional neural networks

Arrhythmia is a prevalent cardiovascular disease that can unveil various heart health issues. In recent times, the rise of wearable devices has garnered attention towards the advantages of portable single-lead ECG devices in heart health monitoring. Achieving end-to-end ECG automatic diagnosis through deep learning is essential due to the time-consuming and labour-intensive characteristics of manual diagnosis of ECG records. This paper presents the design of a hybrid residual recurrent convolutional neural network (HRRCNN) for automatically diagnosing single-lead ECG with multiple labels. Firstly, the convolutional layers and residual networks extract the spatial and short-term temporal features. Subsequently, a hybrid recurrent neural network (RNN) comprised of a double-layer bidirectional gated recurrent unit (GRU) and a bidirectional long short-term memory (LSTM) is employed to capture long and short-term temporal features. Finally, an output layer, comprising an attention layer and a fully connected layer, converts the extracted features into classification results. To evaluate the performance of HRRCNN, a dataset from the Shanghai First People’s Hospital Affiliated to Shanghai JiaoTong University(Shanghai General Hospital, SGH) is utilized. The experimental results demonstrate that HRRCNN achieves an accuracy of 95.08 %, F1-macro of 87.22 %, Sensitivity of 88.51 %, and a positive prediction rate of 86.13 %. Comparative analysis with existing methods underscores the superior performance of HRRCNN, thereby establishing its potential as an out-of-hospital heart health monitoring tool and an auxiliary tool in clinical settings.

Safe drug recommendation through forward data imputation and recurrent residual neural network

Drug recommendation is crucial for aiding clinical decisions and promoting the scientific and rational use of drugs. The goal of drug recommendation is to provide safe and effective drug combinations for patients based on their historical medical records. Previous research heavily relies on electronic health records (EHRs) for data analysis. However, incomplete information and the presence of abnormal data in EHRs often lead to imprecise patient characterization. And it is challenging to learn patient representations and efficiently recommend safe drug combinations through EHRs. To tackle these concerns, this paper presents FDIRNet model to improve the performance of drug recommendation. FDIRNet employs forward data imputation to deal with missing and abnormal data of single-visit and multi-visits. Meanwhile, FDIRNet utilizes recurrent neural networks and residual neural networks to capture the changing patterns of a patient’s historical information. In addition, a novel loss function is proposed, which is combined with a Drug-Drug Interaction (DDI) loss strategy to mitigate DDI, thereby improving the overall robustness of the model. FDIRNet is evaluated on the publicly available dataset and it exhibits significant improvements over previous methods. We have observed distinct increases of 3.1%, 1.6%, and 3.1% in Jaccard, F1 score, and PRAUC, respectively.

Real-time prediction of ship motion based on improved empirical mode composition and dynamic residual neural network

Ship motion attitude data have strong random and non-stationary characteristics under severe sea states. Real-time and accurate prediction of the ship's motion attitude can significantly improve the safety of navigation. Therefore, this article proposes a prediction model based on the improved empirical mode decomposition (IEMD) and the dynamic residual recurrent neural network with bidirectional structure and time pattern attention mechanism (TPA-Bi-DRRNN), and proposes a new algorithm: dynamic adaptive beetle swarm antennae search (DABSAS) algorithm to optimise the initial weight and threshold of the prediction model. The input data are adaptively composed into multiple intrinsic mode functions (IMFs) containing their frequency characteristics using IEMD, and a prediction is made for each IMF using TPA-Bi-DRRNN. The model structure can be regulated in real time according to the sliding window. The destroyer DTMB5415 is used as an example to conduct a performance test of the hybrid prediction model and DABSAS. The results show that no mode mixing occurred in the IMFs, the noise in each IMF was significantly reduced, thus dramatically reducing the difficulty of using the input data for prediction; and the optimization performance of DABSAS in applying the TPA-Bi-DRRNN is much better than that of the traditional algorithms. Compared with other models, the difference in the predictive accuracy of TPA-Bi-DRRNN under each condition is the smallest, suggestings that it has extremely high robustness and will always be able to maintain much higher accuracy than other models over a long period, thus meeting the needs for real-time accurate prediction of ship motion attitude.

A multivariate natural gas load forecasting method based on residual recurrent neural network

AbstractCurrent natural gas load forecasting encounters with the conundrum of unsatisfying accuracy and interpretability. To address the challenge, a multi‐variate forecasting method is proposed, which contains three phases: First, an integrate history‐climate‐holiday factor set is established to provide multi‐perspective for a more explainable forecast; Second, factor fusion interaction between features and instances is carried out based on hierarchical contrastive learning, which contributes to inter‐intra factors potential relationships exploration. Third, a multivariate forecasting model named ResRNN is trained using fused target dataset. Due to its innovation in structure and loss, forecasting accuracy is further improved. Finally, the authors’ method's superiority is confirmed by several groups of comparative experiments and results demonstrate that it outperforms mainstream methods.

Sound Event Detection with Perturbed Residual Recurrent Neural Network

Sound event detection (SED) is of great practical and research significance owing to its wide range of applications. However, due to the heavy reliance on dataset size for task performance, there is often a severe lack of data in real-world scenarios. In this study, an improved mean teacher model is utilized to carry out semi-supervised SED, and a perturbed residual recurrent neural network (P-RRNN) is proposed as the SED network. The residual structure is employed to alleviate the problem of network degradation, and pre-training the improved model on the ImageNet dataset enables it to learn information that is beneficial for event detection, thus improving the performance of SED. In the post-processing stage, a customized median filter group with a specific window length is designed to effectively smooth each type of event and minimize the impact of background noise on detection accuracy. Experimental results conducted on the publicly available Detection and Classification of Acoustic Scenes and Events 2019 Task 4 dataset demonstrate that the P-RRNN used for SED in this study can effectively enhance the detection capability of the model. The detection system achieves a Macro Event-based F1 score of 38.8% on the validation set and 40.5% on the evaluation set, indicating that the proposed method can adapt to complex and dynamic SED scenarios.

An intelligent radar signal classification and deinterleaving method with unified residual recurrent neural network

AbstractThe accuracy of radar emitter signal sorting nowadays deteriorates due to the high flexibility and complexity of modern radar pulse streams and the density of crowded electromagnetic environment. In modern radar signal sorting based on pulse repetition interval, conventional methods usually fail to achieve acceptable accuracy and lack stable performance for two main reasons: (1) Conventional methods require a large number of pulses in the stream, which is not practical in many applications. (2) These methods are sensitive to pulse loss and random noise pulses. These two reasons are the main problem that is addressed in this paper. Our proposed model is a machine learning architecture called Unified Residual Recurrent Neural Network (URRNN). In this architecture, residual neural network and recurrent neural network are combined and modified to alleviate the forementioned shortcomings of traditional approaches and enhance the model performance in both classification and deinterleaving tasks. This aim is achieved due to the fact that URRNN extracts both spatial and temporal features, which means more information about processed stream that is exploited to enhance model performance. Three different architectural combinations of URRNN, which show high accuracy and reasonable processing time, are built and trained. The structural and functional description are provided for each architecture. Simulation demonstrates high accuracy and reliable performance of the proposed methods in different circumstances. The results are compared with the results obtained by other conventional machine learning techniques.

Congestive Heart Failure Detection From ECG Signals Using Deep Residual Neural Network

The early and accurate detection of congestive heart failure (CHF) using an electrocardiogram (ECG) is of great significance for improving the survival rate of patients. Existing approaches show limited detection accuracy as they fail to capture the temporal ECG dynamics. Also, these methods lack model transparency and are often difficult to interpret. This article proposes a novel end-to-end diagnostic attention-based deep residual recurrent neural network (DA-DRRNet) that effectively captures the temporal dynamics and extracts high-level attentive representations for accurate CHF detection. Specifically, we first employ a recurrent neural network (RNN) layer to encode the temporal dynamics from the raw ECG beats. Then, multilayered RNNs with residual connections are incorporated to extract high-level feature representations hierarchically. The residual connections allow gradients in deep RNN to propagate effectively, thereby improving the network’s representation ability. Finally, an attention module identifies the hidden vectors corresponding to the diagnostically prominent ECG characteristics to form an attentive representation for improved CHF detection. Using ECG signals from the three publicly available datasets (BIDMC-CHF, PTBDB, and MIT-BIH NSRDB), the proposed method achieves an impressive accuracy of 98.57 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\%$</tex-math> </inline-formula> and nearly 100 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\%$</tex-math> </inline-formula> for beat-level and 24-h record-level diagnosis, respectively. Notably, the analysis of learned attention weights demonstrates that the proposed model focuses on the clinically relevant ECG features that characterize CHF. This model transparency and improved detection results advance research in this field and provide a reliable and transparent diagnostic system for CHF analysis.

Real-time ship motion prediction based on adaptive wavelet transform and dynamic neural network

Ship motion attitude is the time series data with highly nonlinear and non-stationary characteristics significantly affected by environmental factors. To accurately predict the ship motion attitude in real-time, a ship motion attitude prediction model based on the Adaptive Discrete Wavelet Transform Algorithm (ADWT) and the space-time Residual Recurrent Neural Network (RRNN) with a time-varying structure is proposed. The ADWT decomposes the original data into components easier to predict, each element in the optimal component combination is predicted using the RRNN, the structure and parameters of which are adjusted in real-time with the sliding of data window. The model performance tests are conducted based on the simulation data of the ship motion of DTMB5415. The results show that the subsequences decomposed by the ADWT are easier to predict. Compared with other prediction models, the prediction accuracy of the ADWT-RRNN is the highest under all working conditions, its prediction accuracy and stability of it do not fluctuate significantly over a long prediction period. Hence, the more severe the sea states, the more pronounced the performance advantage is over other models. Finally, a reliable and efficient tool is eventually provided for real-time and accurate prediction of ship motion.

Epidemic efficacy of Covid-19 vaccination against Omicron: An innovative approach using enhanced residual recurrent neural network.

The outbreak of COVID-19 has engulfed the entire world since the end of 2019, causing tremendous loss of lives. It has also taken a toll on the healthcare sector due to the inability to accurately predict the spread of disease as the arrangements for the essential supply of medical items largely depend on prior predictions. The objective of the study is to train a reliable model for predicting the spread of Coronavirus. The prediction capabilities of various powerful models such as the Autoregression Model (AR), Global Autoregression (GAR), Stacked-LSTM (Long Short-Term Memory), ARIMA (Autoregressive Integrated Moving Average), Facebook Prophet (FBProphet), and Residual Recurrent Neural Network (Res-RNN) were taken into consideration for predicting COVID-19 using the historical data of daily confirmed cases along with Twitter data. The COVID-19 prediction results attained from these models were not up to the mark. To enhance the prediction results, a novel model is proposed that utilizes the power of Res-RNN with some modifications. Gated Recurrent Unit (GRU) and LSTM units are also introduced in the model to handle the long-term dependencies. Neural Networks being data-hungry, a merged layer was added before the linear layer to combine tweet volume as additional features to reach data augmentation. The residual links are used to handle the overfitting problem. The proposed model RNN Convolutional Residual Network (RNNCON-Res) showcases dominating capability in country-level prediction 20 days ahead with respect to existing State-Of-The-Art (SOTA) methods. Sufficient experimentation was performed to analyze the prediction capability of different models. It was found that the proposed model RNNCON-Res has achieved 91% accuracy, which is better than all other existing models.

Improved Fruitfly Optimization with Stacked Residual Deep Learning Based Email Classification

Applied linguistics means a wide range of actions which include addressing a few language-based problems or solving some language-based concerns. Emails stay in the leading positions for business as well as personal use. This popularity grabs the interest of individuals with malevolent intentions—phishing and spam email assaults. Email filtering mechanisms were developed incessantly to follow unwanted, malicious content advancement to protect the end-users. But prevailing solutions were focused on phishing email filtering and spam and whereas email labelling and analysis were not fully advanced. Thus, this study provides a solution related to email message body text automatic classification into phishing and email spam. This paper presents an Improved Fruitfly Optimization with Stacked Residual Recurrent Neural Network (IFFO-SRRNN) based on Applied Linguistics for Email Classification. The presented IFFO-SRRNN technique examines the intrinsic features of email for the identification of spam emails. At the preliminary level, the IFFO-SRRNN model follows the email pre-processing stage to make it compatible with further computation. Next, the SRRNN method can be useful in recognizing and classifying spam emails. As hyperparameters of the SRRNN model need to be effectually tuned, the IFFO algorithm can be utilized as a hyperparameter optimizer. To investigate the effectual email classification results of the IFFO-SRDL technique, a series of simulations were taken placed on public datasets, and the comparison outcomes highlight the enhancements of the IFFO-SRDL method over other recent approaches with an accuracy of 98.86%.

Attention mechanism combined with residual recurrent neural network for sound event detection and localization

In the task of sound event detection and localization (SEDL) in a complex environment, the acoustic signals of different events usually have nonlinear superposition, so the detection and localization effect is not good. Given this, this paper is based on the Residual-spatially and channel Squeeze-Excitation (Res-scSE) model. Combined with Multiple-scale Convolutional Recurrent Neural Network (M-CRNN), the Res-scSE-CRNN model is proposed. Firstly, to solve the problem of insufficient extraction of time-frequency feature in single-size convolution kernel, multi-scale feature fusion is carried out by using the feature hierarchy of the convolutional neural network to improve the accuracy of detection. Secondly, aiming at the problem of overlapping audio event localization accuracy is not high, with Res-scSE to replace common convolution module and add residual structure to strengthen the feature extraction, and combining with an attention mechanism to enhance neural network channels and spatial relationships, to improve the network to extract the characteristics of directivity, achieve the goal of the overlapped audio localization. In this paper, experiments are carried out in the open dataset DCASE2019, and evaluation indicators are used to analyze the effectiveness of the proposed model and baseline model in the detection and localization of audio events. The results show that compared with the M-CRNN model, the detection error rate of Res-scSE-CRNN model is reduced 4%, the F1-Score is increased 3.4%, the localization error is reduced by 22.8°, and the frame recall rate is increased 3%.

A Deep Residual Recurrent Neural Network Model-Augmented Attention With Physical Characteristics: Application to Turntable Servo System

Turntable servo systems are affected by working status changing in actual work, thus, bringing some uncertain factors. Having models that have accurate predictive capabilities can be valuable for control and decision-making purposes. This article proposes a deep residual recurrent neural network (DRRNN) model-augmented attention with physical characteristics, which is used to build an accurate speed prediction model for a turntable servo system. Integrating the known physical features into the design of the residual network model can increase the interpretability of the overall model. Meanwhile, the recurrent neural network-augmented attention model is used to represent the unknown nonlinear characteristics. The addition of the attention mechanism can effectively improve the accuracy of the nonlinear model. Finally, this article proposes an iterative identification method based on the linear and nonlinear model error with alternative compensation. Through alternating error compensation, the linear model’s accuracy can be continuously improved, then the overall and the nonlinear part of the model can be established accurately by retraining the whole model. Experimental results show that the DRRNN model designed in this article can predict the rotational speed in single-step or multistep well. Compared with the traditional mechanism modeling method, the modeling accuracy can be greatly improved.

Semi-sparse Residual Recurrent Neural Network via dictionary representation for throat microphone quality enhancement

Throat microphone (TM) speech can be used for communication in noisy environment as it collects signals directly from human skin, but it should be improved in clarity and intelligibility due to the severe loss of high-frequency components. As recovery directly by neural networks is not sufficient to achieve satisfactory performance, we propose a dictionary representation based neural network to address this issue. Specifically, a magnitude spectrum dictionary of air-conducted speech is computed via sparse non-negative matrix factorization (SNMF), and then it is used to represent the transformed speech in hidden layer of the network. Meanwhile, a compensating dictionary is adopted to improve the representation accuracy. A memory efficient Semi-sparse Residual Recurrent Neural Network (SResRNN) with interactive mechanism and a special ResNet is employed to generate the coefficients on the dictionaries. Lastly, a three-layer neural network using a special initialization scheme is constructed as the recovery model. In the experiments, the model is compared with other five recovering models, and different criteria are adopted to measure the performance, the objective and subjective results can demonstrate the superiority of our proposed model.

Residual RNN Models With Pruning for Digital Predistortion of RF Power Amplifiers

The paper presents several novel residual recurrent neural network (RNN) models for digital predistortion of radio-frequency power amplifiers. Then, unstructured and structured neural network (NN) pruning algorithms are proposed to reduce the per-sample computational complexity and the required memory, without compromising their modeling performance. The mathematical equations of the complexity and the memory are derived to analyze the hardware cost and design low cost models. An experimental platform is developed to measure the modeling performance in terms of adjacent channel leakage ratio (ACLR), normalized mean square error, error vector magnitude (EVM) and bit error rate. The platform uses an envelope tracking PA excited by 80 MHz IEEE 802.11ac signals. Compared with the state-of-the-art RNN model in (Sun <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">et al.</i> , 2019), the proposed models reduce the per-sample complexity and required memory by <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$94\hbox{--}99$</tex-math></inline-formula> %, they also improve the ACLR and EVM by at least 3 dB and 1 dB at 0.45 dB and 3.15 dB gain compression, respectively. Furthermore, an observation path with time alignment is used to remove the transmit latency in the state-of-the-art approach. At 3.15 dB gain compression, the proposed models outperform the state-of-the-art feedforward and convolutional NN models by at least 1.3 dB in EVM, outperform the general memory polynomial model by at least 3.3 dB in EVM. The above evidences indicate that the proposed RNN models can be implemented with relatively low hardware cost and achieve the best EVM in high power efficiency scenario.

A Model RRNet for Spectral Information Exploitation and LAMOST Medium-resolution Spectrum Parameter Estimation

This work proposes a residual recurrent neural network (RRNet) for synthetically extracting spectral information and estimating stellar atmospheric parameters together with 15 chemical element abundances for medium-resolution spectra from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST). The RRNet consists of two fundamental modules: a residual module and a recurrent module. The residual module extracts spectral features based on the longitudinally driving power from parameters, while the recurrent module recovers spectral information and restrains the negative influences from noises based on Cross-band Belief Enhancement. RRNet is trained by the spectra from common stars between LAMOST DR7 and the APOGEE-Payne catalog. The 17 stellar parameters and their uncertainties for 2.37 million medium-resolution spectra from LAMOST DR7 are predicted. For spectra with a signal-to-noise ratio ≥ 10, the precision of estimations (T eff and log g) are 88 K and 0.13 dex, respectively, elements C, Mg, Al, Si, Ca, Fe, and Ni are 0.05–0.08 dex, and N, O, S, K, Ti, Cr, and Mn are 0.09–0.14 dex, while that of Cu is 0.19 dex. Compared with StarNet and SPCANet, RRNet shows higher accuracy and robustness. In comparison to Apache Point Observatory Galactic Evolution Experiment and Galactic Archaeology with HERMES surveys, RRNet manifests good consistency within a reasonable range of bias. Finally, this work releases a catalog of 2.37 million medium-resolution spectra from the LAMOST DR7, the source code, the trained model, and the experimental data, respectively, for astronomical science exploration and data-processing algorithm research reference.

Heart Sound Classification Based on Mel-Frequency Cepstrum Coefficient Features and Multi-Scale Residual Recurrent Neural Networks

A rapid and accurate algorithm model of extracting heart sounds plays a vital role in the early detection of cardiovascular disorders, especially for small primary health care clinics. This paper proposes a heart sound extraction and classification algorithm based on static and dynamic combination of Mel-frequency cepstrum coefficient (MFCC) feature extraction and the multi-scale residual recurrent neural network (MsRes-RNN) algorithm model. The standard MFCC parameters represent the static characteristics of the signal. In contrast, the first-order and second-order MFCC parameters represent the dynamic characteristics of the signal. They are extracted and combined to form the MFCC feature representation. Then, the MFCC-based features are fed to a MsRes-RNN algorithm model for feature learning and classification tasks. The proposed classification model can take advantage of the encoded local characteristics extracted from the multi-scale residual neural network (MsResNet) and the long-term dependencies captured by recurrent neural network (RNN). Model estimation experiments and performance comparisons with other state-of-the-art algorithms are presented in this paper. Experiments indicate that a classification accuracy of 93.9% has been achieved on 2016 PhysioNet/CinC Challenge datasets.

An FPGA-Based Residual Recurrent Neural Network for Real-Time Video Super-Resolution

In this paper, we propose a hardware-efficient residual recurrent neural network for real-time video super-resolution (VSR) based on field programmable gate array (FPGA). Although recent learning-based VSR methods have achieved remarkable performance, the large computational complexity prohibits the deployment of the sophisticated VSR models on FPGA for real-time applications. Limited by the hardware resources, state-of-the-art FPGA-based VSR methods perform single-image super-resolution over the video sequence and suffer from temporal inconsistency. In order to exploit the inter-frame temporal correlation for real-time VSR on low-complexity hardware, we introduce a hardware-efficient recurrent neural network ERVSR. Specially, the proposed ERVSR leverages the input frame and the temporal information entailed in the hidden state to reconstruct the high-resolution counterpart. To reduce the network parameters, the low-resolution input branch and the hidden state branch are convolved individually and a channel modulation coefficient is proposed to explicitly guide the network to allocate the amount of output feature channels to each branch. Additionally, in order to reduce the memory consumption, we perform a dedicated lightweight compression of the hidden state by introducing a statistical normalization scheme followed by a fixed-point quantization. Besides, we adopt group convolution and depthwise separable convolution to further compact the network. We evaluated the proposed ERVSR on multiple public datasets from different aspects. Experimental results demonstrate that ERVSR performs better than the existing state-of-the-art FPGA-based VSR methods in both image quality and data throughput.

Predicting the At-Risk Online Students Based on the Click Data Distribution Characteristics

High fail and dropout rates are the major problems in distance education. Due to a large number of online learners and limited teacher resources, it is essential to accurately identify these potential at-risk students in advance and provide timely aids, which will help to improve the educational outcome. In the online learning environment, students’ online learning behaviors can be recorded easily, with the click data being the most common one. Students’ learning behavior can reflect their learning situation and may differ among different students and periods. This paper proposed a model that uses the short-period activity characteristic and long-term changing pattern to predict the potential at-risk students. The model contains two stages: information extraction and information utilization. The first stage extracts data from the log files and organizes it in a form suitable for the model. In the second stage, according to the different characteristics of students’ short-term and long-term learning behavior, a convolution residual recurrent neural network (CRRNN) model is proposed. The convolutional neural network is used to obtain the representation of the student’s learning behavior in a certain period. Then, the residual recurrent neural network is used to get the behavior changing pattern over the periods. The experimental results indicate that the proposed model has higher performance than the three widely used baseline methods on the OULA dataset and has good practical application value for teaching and management.

Physics-informed deep learning model in wind turbine response prediction

Subjected to strong cyclic wind and wave loads, wind turbines could experience severe fatigue damages and possibly fail to function normally due to accumulated damages at certain critical locations. Therefore, fatigue damage evaluation and prediction are essential and important to be conducted, which could involve massive numerical simulations and computational costs due to dynamic analyses of the wind turbines under various environmental conditions. To reduce the calculation cost related to the time-consuming dynamic analysis, sequence models such as the recurrent neural network (RNN) and the long-short term memory model (LSTM) originated from the deep learning topic are good and promising candidates to predict structural dynamic responses at multiple critical locations under different environmental scenarios. However, the training cost and prediction accuracy of these deep learning models might not be satisfiable since these models are purely data-driven and require significant amount of training data and a large number of training parameters. To reduce the computational cost and improve the prediction accuracy, a hybrid method that integrates the physical information of the underlying wind turbine system into the data-driven model is implemented in the present study as a computationally efficient simulation model. Structural properties and linearized representations of the wind turbine system are served as the physical constraints and applied in a recently proposed deep residual recurrent neural network (DR-RNN) to form as a physics-informed deep learning model. This physics-informed model is first applied to a frame structure with four degrees of freedom as a benchmark study to show the accuracy and efficiency of this model. The applicability of this physics-informed model to a complex wind turbine system is then investigated, and the performance of the developed physics-informed model on the structural response prediction is also compared with a regular data-driven model.