Characteristics Of Time Series Research Articles

Macroscale intrinsic dynamics are associated with microcircuit function in focal and generalized epilepsies

Epilepsies are a group of neurological disorders characterized by abnormal spontaneous brain activity, involving multiscale changes in brain functional organizations. However, it is not clear to what extent the epilepsy-related perturbations of spontaneous brain activity affect macroscale intrinsic dynamics and microcircuit organizations, that supports their pathological relevance. We collect a sample of patients with temporal lobe epilepsy (TLE) and genetic generalized epilepsy with tonic-clonic seizure (GTCS), as well as healthy controls. We extract massive temporal features of fMRI BOLD time-series to characterize macroscale intrinsic dynamics, and simulate microcircuit neuronal dynamics used a large-scale biological model. Here we show whether macroscale intrinsic dynamics and microcircuit dysfunction are differed in epilepsies, and how these changes are linked. Differences in macroscale gradient of time-series features are prominent in the primary network and default mode network in TLE and GTCS. Biophysical simulations indicate reduced recurrent connection within somatomotor microcircuits in both subtypes, and even more reduced in GTCS. We further demonstrate strong spatial correlations between differences in the gradient of macroscale intrinsic dynamics and microcircuit dysfunction in epilepsies. These results emphasize the impact of abnormal neuronal activity on primary network and high-order networks, suggesting a systematic abnormality of brain hierarchical organization.

Weighted Fuzzy Clustering for Time Series With Trend-Based Information Granulation.

The highly dimensional characteristic of time series brings many challenges on direct mining time series, such as high cost in time and space. Granular computing provides a potential strategy for representing and dealing with time series at a higher level of abstraction. In this study, we propose an information granulation-based weighted fuzzy C -means (wFCM) method to realize time-series clustering, which could avoid high dimensionality processing and provide a concise and visible granular prototype for each cluster. In this method, each time series is first transformed into a series of information granules with trend following the principle of justifiable granularity. The formed granular time series can well capture the main features lying in the original time series and help realize dimensionality reduction. Then, the wFCM method is developed to complete time-series clustering in the granular space. Here, the dynamic time warping (DTW) is extended to capture the similarity for trend-based granular time series. Furthermore, the weighted DTW barycenter averaging is introduced to derive prototypes presented in a granular format, capturing the level, the fluctuation, and the changing trend, which are meaningful and understandable clustering results. The experiments conducted on real-world datasets coming from the UCR time-series database and Chinese stocks are presented to illustrate the effectiveness and practicality of the designed time-series clustering model.

Optimal economic dispatch of a virtual power plant based on gated recurrent unit proximal policy optimization

The intermittent renewable energy in a virtual power plant (VPP) brings generation uncertainties, which prevents the VPP from providing a reliable and user-friendly power supply. To address this issue, this paper proposes a gated recurrent unit proximal policy optimization (GRUPPO)-based optimal VPP economic dispatch method. First, electrical generation, storage, and consumption are established to form a VPP framework by considering the accessibility of VPP state information. The optimal VPP economic dispatch can then be expressed as a partially observable Markov decision process (POMDP) problem. A novel deep reinforcement learning method called GRUPPO is further developed based on VPP time series characteristics. Finally, case studies are conducted over a 24-h period based on the actual historical data. The test results illustrate that the proposed economic dispatch can achieve a maximum operation cost reduction of 6.5% and effectively smooth the supply–demand uncertainties.

Exchange rate prediction research based on LSTM-ELM hybrid model

The fluctuation of exchange rates holds paramount importance for a country's economic and trade activities. Due to the non-stationary and nonlinear structural characteristics of exchange rate time series, accurately predicting exchange rate movements is a challenging task. Single machine learning models often exhibit lower precision in exchange rate prediction compared to combined machine learning models. Hence, employing a combined model approach aims to enhance the predictive performance of exchange rate models. Both Long Short-Term Memory (LSTM) and Extreme Learning Machine (ELM) exhibit intricate structures, making their direct integration challenging. To address this issue, an innovative weighted approach is adopted in this study, combining LSTM and ELM models and further refining the combination weights using an improved Marine Predators Algorithm. This paper encompasses both univariate and multivariate prediction scenarios, employing two distinct allocation strategies for training and testing datasets. This is done to investigate the influence of different dataset allocations on exchange rate prediction. Finally, the proposed LSTM-ELM weighted combination exchange rate prediction model is compared with SVM, Random Forest, ELM, LSTM, and LSTM-ELM average combination models. Experimental results demonstrate that the LSTM-ELM weighted combination exchange rate prediction model outperforms the others in both univariate and multivariate prediction settings, yielding higher predictive accuracy and superior fitting performance. Consequently, the LSTM-ELM weighted combination prediction model proves to be effective in exchange rate forecasting.

Identifying the systemic importance and systemic vulnerability of financial institutions based on portfolio similarity correlation network

The indirect correlation among financial institutions, stemming from similarities in their portfolios, is a primary driver of systemic risk. However, most existing research overlooks the influence of portfolio similarity among various types of financial institutions on this risk. Therefore, we construct the network of portfolio similarity correlations among different types of financial institutions, based on measurements of portfolio similarity. Utilizing the expanded fire sale contagion model, we offer a comprehensive assessment of systemic risk for Chinese financial institutions. Initially, we introduce indicators for systemic risk, systemic importance, and systemic vulnerability. Subsequently, we examine the cross-sectional and time-series characteristics of these institutions’ systemic importance and vulnerability within the context of the portfolio similarity correlation network. Our empirical findings reveal a high degree of portfolio similarity between banks and insurance companies, contrasted with lower similarity between banks and securities firms. Moreover, when considering the portfolio similarity correlation network, both the systemic importance and vulnerability of Chinese banks and insurance companies surpass those of securities firms in both cross-sectional and temporal dimensions. Notably, our analysis further illustrates that a financial institution’s systemic importance and vulnerability are strongly and positively associated with the magnitude of portfolio similarity between that institution and others.

>Water quality prediction of artificial intelligence model: a case of Huaihe River Basin, China.

Accurate prediction of water quality contributes to the intelligent management of water resources. Water quality indices have time series characteristics and nonlinearity, but the existing models only focus on the forward time series when long short-term memory (LSTM) is introduced and do not consider the parallel computation on the model. Owing to this, a new neural network called LSTM-multihead attention (LMA) was constructed to predict water quality, using long short-term memory to process time series data and multihead attention for parallel computing and extracting feature information. Additionally, water quality indices have the issues of multiple data types and complex data correlations, as well as missing data and abnormal data problems in water quality data. In order to solve these problems, this study proposes a water quality prediction model called GRA-LMA-based linear interpolation, gray relational analysis and LMA. Two experiments are carried out to verify the predictive performance of the GRA-LMA with the water quality data of the Huaihe River Basin as a case study sample. The first experiment focuses on data processing, including the processing of missing data and abnormal data of water quality data, and the correlation analysis of water quality indices. Linear interpolation is adapted to process the missing data, while a combination of boxplot and histogram is adopted to analyze and eliminate the abnormal data, which is then repaired the abnormal data with linear interpolation. The gray relational analysis is adopted to calculate the correlation between different water quality indices, and water quality indices with high correlation are retained to determine the input variables of the water quality prediction model. The data processing results demonstrate that repairs can be made using linear interpolation without altering the pattern of data change and the model by using the gray relational analysis to reduce the quantity of data it needs as input. In the second experiment, the predictive capacity of GRA-LMA and existing models such as backpropagation neural network (BP), recurrent neural network (RNN), long short-term memory (LSTM), and gate recurrent unit (GRU) was evaluated and compared using different numerical and graphical performance evaluation metrics.Comparative experimental results show that the mean square error of pH, dissolved oxygen, chemical oxygen demand, ammonia nitrogen, electrical conductivity, turbidity, total phosphorus, and total nitrogen of GRA-LMA is reduced to 0.05890, 0.40196, 0.32454, 0.04368, 14.71003, 8.13252, 0.01558, and 0.14345. The results indicate that GRA-LMA has superior adaptability for predicting various water quality indices and can significantly lower the induced prediction error.

A hybrid model of CNN and LSTM autoencoder-based short-term PV power generation forecasting

AbstractSolar energy is one of the main renewable energies available to fulfill global clean energy targets. The main issue of solar energy like other renewable energies is its randomness and intermittency which affects power grids stability. As a solution for this issue, energy storage units could be used to store surplus energy and reuse it during low solar generation intervals. Also, in order to sustain stable power grid and better grid operation and energy storage management, photovoltaic (PV) power forecasting is inevitable. In this paper, new hybrid model based on deep learning techniques is proposed to predict short-term PV power generation. The proposed model incorporates convolutional neural network (CNN) and long short-term memory (LSTM) autoencoder network. The new model differentiates itself in accomplishing high prediction accuracy by extracting spatial features in time series via CNN layers and temporal features between the time series data through LSTM. The introduced model is tested on dataset of power generation from southern UK solar farm and the weather data corresponding to same location and time intervals; the forecasting performance of the suggested model is evaluated in metrics of root-mean-square error (RMSE) and mean absolute error (MAE). The used model is compared with different models from the literature either of pure type of network such as LSTM and gated recurrent unit (GRU) or hybrid combination of different networks like CNN-LSTM and CNN-GRU. The results show that proposed model provides enhanced results and reduces training time significantly compared to other competitive models, where the performance of the proposed model improved averagely by 5% to 25% in terms of RMSE and MAE performance metrics, and the execution time of training significantly reduced with almost 70% less compared to other models.

Adaptive Wavelet Domain Principal Component Analysis for Nonstationary Time Series

High-dimensional multivariate nonstationary time series, that is, data whose second order properties vary over time, are common in many scientific and industrial applications. In this article we propose a novel wavelet domain dimension reduction technique for nonstationary time series. By constructing a time-scale adaptive principal component analysis of the data, our proposed method is able to capture the salient dynamic features of the multivariate time series. We also introduce a new time and scale dependent cross-coherence measure to quantify the extent of association between a multivariate nonstationary time series and its proposed wavelet domain principal component representation. Theoretical results establish that our associated estimation scheme enjoys good bias and consistency properties when determining wavelet domain principal components of input data. The proposed method is illustrated using extensive simulations and we demonstrate its applicability on a real-world dataset arising in a neuroscience study. Supplementary materials, with proofs of theoretical results, additional simulations and code, are available online.

Detection of abnormal operation in geothermal binary plant feed pumps using time-series analytics

In this study, we investigate potential pump cavitation events in circulating feed pumps deployed in a geothermal binary power plant using feature-based time series analytics and machine learning. The feed pumps have been in operation for over a decade but have been experiencing episodes wherein the pumps were drawing in lower power even while pump speeds were maintained at the same levels.A workflow combining time-series analytics and first-principle thermodynamics was applied to characterise potential drivers of the abnormal events in the feed pumps. An approach based on systematic time-series feature engineering and semi-supervised machine learning was used to fully annotate the input data and identify twenty (20) undocumented occurrences of the target event. Time-series features extracted from the fully labelled dataset were then used to train forecasting models to identify potential drivers of the target events. Forecasting models trained using pump data outperformed those trained on times-series data from other power plant components (Matthew’s correlation coefficient score of 0.490 vs 0.255). This result indicates that the abnormal events are primarily driven by operating conditions local to the feed pump system. This finding was confirmed by Bayesian A/B testing, showing that the target events were more likely to occur by 22% when four (4) feed pumps were running compared to when only three (3) pumps were in operation. Furthermore, thermodynamic analysis of the state of the working fluid across the feed pumps shows irregularities at the pump inlet, where indications of partial vaporisation of the working fluid have been identified.

ESTSS—energy system time series suite: a declustered, application-independent, semi-artificial load profile benchmark set

This paper introduces an univariate application-independent set of load profiles or time series derived from real-world energy system data. The generation involved a two-step process: manifolding the initial dataset through signal processors to increase diversity and heterogeneity, followed by a declustering process that removes data redundancy. The study employed common feature engineering and machine learning techniques: the time series are transformed into a normalized feature space, followed by a dimensionality reduction via hierarchical clustering, and optimization. The resulting dataset is uniformly distributed across multiple feature space dimensions while retaining typical time and frequency domain characteristics inherent in energy system time series. This data serves various purposes, including algorithm testing, uncovering functional relationships between time series features and system performance, and training machine learning models. Two case studies demonstrate the claims: one focused on the suitability of hybrid energy storage systems and the other on quantifying the onsite hydrogen supply cost in green hydrogen production sites. The declustering algorithm, although a bys study, shows promise for further scientific exploration. The data and source code are openly accessible, providing a robust platform for future comparative studies. This work also offers smaller subsets for computationally intensive research. Data and source code can be found at https://github.com/s-guenther/estss and https://zenodo.org/records/10213145.

Application of a long short-term memory neural network algorithm fused with Kalman filter in UWB indoor positioning

Ultra-wideband technology has good anti-interference capabilities and development prospects in indoor positioning. Since ultra-wideband will be affected by random errors in indoor positioning, to exploit the advantages of the Kalman filter (KF) and the long short-term memory (LSTM) network, this paper proposes a long short-term memory neural network algorithm fused with the Kalman filter (KF–LSTM) to improve UWB positioning. First, the ultra-wideband data is processed through KF to weaken the noise in the data, and then the data is fed into the LSTM network for training, and the capability of the LSTM network to process time series features is employed to obtain more accurate label positions. Finally, simulation and measurement results show that the KF–LSTM algorithm achieves 71.31%, 37.28%, and 49.31% higher average positioning accuracy than the back propagation (BP) network, (back propagation network fused with the Kalman filter (KF-BP), and LSTM network algorithms, respectively, and the KF–LSTM algorithm performs more stably. Meanwhile, the more noise the data contains, the more obvious the stability contrast between the four algorithms.

A machine vision method for the evaluation of ship-to-ship collision risk

The development of ship technology and information technology has been driving the continuous improvement of ship intelligence, with safety being an inevitable requirement in the shipping industry. A machine vision-based ship collision warning method is proposed for high monitoring system cost and limited information acquisition in safety design of autonomous ship navigation. The method combines machine learning with image algorithms. Firstly, the backbone of YOLOv7 detector is replaced by EfficientFormerV2 network to achieve model lightweight while ensuring detection accuracy. Public datasets SeaShips, Flow and self-made ship pictures are combined, and the network is trained on this dataset. StrongSORT is used for target tracking. Secondly, a data fusion algorithm is introduced to determine the target point at the bow-bottom of the ship based on the time-varying attitude of the camera and the time-series features of the bounding boxes. Ship navigation trajectory estimation is performed using image algorithms. Finally, a collision evaluation model is established to calculate the collision risk index. Experimental results demonstrate that the improved YOLOv7 network maintains similar mAP.5 and Recall compared to the original model, while reducing the parameters by 31.2 % and GFLOPs by 58.4 %. The accuracy of target ship trajectory estimation is high, with MAE values below 1.5 % and RMSE values below 2 % in experiments. In ship collision warning experiments, the proposed method accurately identifies navigating vessels, estimates the trajectories, and provides timely warnings for imminent collision accidents. Compared to traditional ship collision warning methods, this paper offers a more intelligent and lightweight solution.

GAN-Enhanced Nonlinear Fusion Model for Stock Price Prediction

Stock price prediction is a significant field of finance research for both academics and practitioners. Numerous studies have proved that the stock movement can be fully reflect various internal features of stock price including non-stationary behavior, high persistence in the conditional variance. The fusion of time-series prediction model such as Auto-Regressive Integrated Moving Average (ARIMA) and neural network is an availability but difficult approach for stock price prediction. Although the orientation has been studied through some methods in different research, there are still difficulties with the poor capture ability of time-series features and insufficient effectiveness of integrating temporal feature and frequency domain information. In this paper, we propose a Generative Adversarial Network (GAN) framework with the Convolution Neural Networks (CNN) as the discriminator and a hybrid model as the generator for forecasting the stock price. The hybrid model includes Attention-based Convolution Neural Networks (ACNN), Long Short-Term Memory (LSTM), and ARIMA model. Moreover, this proposed framework uses the Generative Adversarial patten and Attention Mechanism to achieve effective analysis and feature extraction for stock price movement. The extensive experiments in different history periods of dataset demonstrate an improvement in forecasting of stock price using our model as compared to the baseline models.

A Novel Hybrid Model Combining Improved VMD and ELM with Extended Maximum Correntropy Criterion for Prediction of Dissolved Gas in Power Transformer Oil

The prediction of dissolved gas change trends in power transformer oil is very important for the diagnosis of transformer faults and ensuring its safe operation. Considering the time series and nonlinear features of the gas change trend, this paper proposes a novel robust extreme learning machine (ELM) model combining an improved data decomposition method for gas content forecasting. Firstly, the original data with nonlinear and sudden change properties will make the forecasting model unstable, and thus an improved variational modal decomposition (IPVMD) method is developed to decompose the original data to obtain the multiple modal dataset, in which the marine predators algorithm (MPA) optimization method is utilized to optimize the free parameters of the VMD. Second, the ELM as an efficient and easily implemented tool is used as the basic model for dissolved gas forecasting. However, the traditional ELM with mean square error (MSE) criterion is sensitive to the non-Gaussian measurement noise (or outliers). In addition, considering the nonlinear non-Gaussian properties of the dissolved gas, a new learning criterion, called extended maximum correntropy criterion (ExMCC), is defined by using an extended kernel function in the correntropy framework, and the ExMCC as a learning criterion is introduced into the ELM to develop a novel robust regression model (called ExMCC-ELM) to improve the ability of ELM to process mutational data. Third, a gas-in-oil prediction scheme is proposed by using the ExMCC-ELM performed on each modal obtained by the proposed IPVMD. Finally, we conducted several simulation studies on the measured data, and the results show that the proposed method has good predictive performance.

Multi-Step Prediction of Wind Power Based on Hybrid Model with Improved Variational Mode Decomposition and Sequence-to-Sequence Network

Due to the complexity of wind power, traditional prediction models are incapable of fully extracting the hidden features of multidimensional strong fluctuation data, which results in poor multi-step prediction performance. To predict continuous power effectively in the future, an improved wind power multi-step prediction model combining variational mode decomposition (VMD) with sequence-to-sequence (Seq2Seq) is proposed. Firstly, the wind power sequence is smoothed using VMD and the decomposition parameters of VMD are optimized by using the squirrel search algorithm (SSA) to effectively optimize the decomposition effect. Then, the subsequence obtained from decomposition, together with the original wind power data, is reconstructed into multivariate time series features. Finally, a Seq2Seq model is constructed, and convolutional neural networks (CNNs) with bidirectional gate recurrent units (BiGRUs) are used to learn the coupling and timing relationships of the input data and encode them. The gate recurrent unit (GRU) is decoded to achieve continuous power prediction. Based on the actual operating data of a wind farm, a case analysis is conducted. Experimental results show that SSA-VMD can effectively optimize the decomposition effect, and the subsequences obtained with its decomposition are highly accurate when applied to predictions. The Seq2Seq model has better multi-step prediction results than traditional prediction methods, and as the prediction step size increases, the advantages are more obvious.

A novel model to estimate regional differences in time-series solar and wind forecast predictability across small regions: A case study in South Korea

Forecasting techniques for solar and wind energy are essential for controlling their variability and are being heavily researched. However, regional differences in predictability in these forecast models have rarely been studied. Regional differences refer to differences in forecast performance based on location. In this study, regional differences in time-series forecast models often used for short-term forecasts were quantitatively analyzed. A three-step methodology was devised to extract significant features for forecasting performance and develop a model to estimate forecast performance. (1) Multiple forecasting models, including machine learning, were applied to 100 random sites to calculate performance. (2) The time-series features of each site were statistically extracted and compared with forecast performance. (3) A formula-based modefl was developed to estimate forecast performance based on highly correlated metrics. The standard deviations of the rolling mean and those of the rolling standard deviations were significant metrics among the rolling statistics useful for interpreting time-series data and showed a correlation of >0.8 with regional forecast performance for both solar and wind. These findings can reduce the uncertainty of forecast model applications and minimize the risk of power system operations.

Predicting China’s elderly population by using a fractional grey prediction model with time power term

Accurate prediction of the aging population can provide valuable reference and corresponding theoretical support for the adjustment of national population development policy and economic development strategy. To explore the future development trend of China’s aging population, this paper establishes a novel fractional grey prediction model with the time power term (abbreviated as FGM (1, 1, t α) model) to study China’s aging population. FGM (1, 1, t α) has the properties of fractional order accumulation operation and GM (1, 1, t α) model, which makes it good at capturing nonlinear features in time series. Furthermore, the quantum genetic algorithm is used to search for unknown parameters in the model to facilitate the solving task of the model. Data on China’s aging population from 2000 to 2009 are used to train the prediction models, and data from 2010 to 2019 are used to evaluate the models’ prediction performance. The results show that the FGM (1, 1, t α) model outperforms the other competing models, which means that it has good generalization. Finally, the FGM (1, 1, t α) model is used to forecast China’s aging population from 2020 to 2029.

Using meta-learning to recommend an appropriate time-series forecasting model

BackgroundThere are various forecasting algorithms available for univariate time series, ranging from simple to sophisticated and computational. In practice, selecting the most appropriate algorithm can be difficult, because there are too many algorithms. Although expert knowledge is required to make an informed decision, sometimes it is not feasible due to the lack of such resources as time, money, and manpower.MethodsIn this study, we used coronavirus disease 2019 (COVID-19) data, including the absolute numbers of confirmed, death and recovered cases per day in 187 countries from February 20, 2020, to May 25, 2021. Two popular forecasting models, including Auto-Regressive Integrated Moving Average (ARIMA) and exponential smoothing state-space model with Trigonometric seasonality, Box-Cox transformation, ARMA errors, Trend, and Seasonal components (TBATS) were used to forecast the data. Moreover, the data were evaluated by the root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and symmetric mean absolute percentage error (SMAPE) criteria to label time series. The various characteristics of each time series based on the univariate time series structure were extracted as meta-features. After that, three machine-learning classification algorithms, including support vector machine (SVM), decision tree (DT), random forest (RF), and artificial neural network (ANN) were used as meta-learners to recommend an appropriate forecasting model.ResultsThe finding of the study showed that the DT model had a better performance in the classification of time series. The accuracy of DT in the training and testing phases was 87.50% and 82.50%, respectively. The sensitivity of the DT algorithm in the training phase was 86.58% and its specificity was 88.46%. Moreover, the sensitivity and specificity of the DT algorithm in the testing phase were 73.33% and 88%, respectively.ConclusionIn general, the meta-learning approach was able to predict the appropriate forecasting model (ARIMA and TBATS) based on some time series features. Considering some characteristics of the desired COVID-19 time series, the ARIMA or TBATS forecasting model might be recommended to forecast the death, confirmed, and recovered trend cases of COVID-19 by the DT model.

Spatial and Temporal Variability Characteristics and Driving Factors of Extreme Precipitation in the Wei River Basin

As global climate change intensifies, the global atmospheric circulation process is undergoing significant changes, and the local water vapor pattern has also changed. This study takes the Wei River Basin as the research area. Firstly, an evaluation index system for extreme precipitation was established, and the time-series characteristics of the magnitude, frequency, and duration of extreme precipitation were analyzed. Statistical methods were used to analyze the non-consistency in time-series changes in extreme precipitation indicators. Using spatial heterogeneity analysis methods, the spatial variation differences in extreme precipitation in the Wei River Basin were identified. This study selected the El Niño-Southern Oscillation (ENSO) index, global land-ocean temperature index (LOTI), and land surface temperature (LST) index to quantitatively evaluate the impact of climate change on regional extreme precipitation and analyzed the correlation between temperature and extreme precipitation, identifying the key driving factors of extreme precipitation changes. The conclusions of this study are as follows: (1) The southern region of the Wei River Basin experiences more frequent and intense precipitation events, while the northern region experiences relatively few. (2) From 1981 to 2021, the intensity, frequency, and duration of precipitation events in the Wei River Basin gradually increased, with the most significant increase in extreme precipitation in the Guanzhong Plain. (3) Global climate change has an important impact on precipitation events in the Wei River Basin. The increase in the ENSO, LOTI, and LST indices may indicate an increase in the probability of drought and flood events in the Wei River Basin. The relationships between extreme precipitation and temperature present a peak structure. This conclusion is helpful to better understand the impact of climate change on extreme precipitation in the Wei River Basin and provides some support for the response to extreme meteorological events under the background of future climate change.

State of health estimation of lithium-ion battery based on CNN–WNN–WLSTM

Accurate and stable estimation of the state of health (SOH), which is one of the critical indicators to characterize the ability of lithium-ion (Li-ion) batteries to store and release energy, is critical in the stable driving of electric vehicles. In this paper, a novel SOH estimation method based on the aging factors of battery, which combines convolutional neural network (CNN), wavelet neural network (WNN), and wavelet long short-term memory (WLSTM) named CNN–WNN–WLSTM, is designed. The proposed CNN–WNN–WLSTM estimation scheme inherits both the fast convergence and robust stability of the WNN, as well as the ability of long short-term memory neural network (LSTM) to extract the time series features of the data; moreover, using CNN can make the proposed algorithm extract the data features from the original battery data automatically, and the WNN–WLSTM is then adopted to produce the final SOH estimation by exploiting the features from the CNN. To further speed and achieve global optimization, the RMSprop optimizer, instead of the usually used Adagrad optimizer, is chosen as the solver of the CNN–WNN–WLSTM network. Experimental results on data set from the NASA Ames Prognostics Center of Excellence show that the proposed algorithm can be commendably used for Li-ion battery health management by quantitative comparison with other commonly used machine learning methods, such as back-propagation neural network, WNN, LSTM, WLSTM, convolutional neural network–long short-term memory neural network (CNN–LSTM), and Gaussian process regression.