State-of-charge (SOC), which indicates the remaining capacity at the current cycle, is the key to the driving range prediction of electric vehicles and optimal charge control of rechargeable batteries. In this paper, we propose a combined convolutional neural network (CNN) -long short-term memory (LSTM) network to infer battery SOC from measurable data, such as current, voltage, and temperature. The proposed network shares the merits of both CNN and LSTM networks and can extract both spatial and temporal features from input data. The proposed network is trained using data collected from different discharge profiles, including a dynamic stress test, federal urban driving schedule, and US06 test. The performance of the proposed network is evaluated using data collected from a new combined dynamic loading profile in terms of estimation accuracy and robustness against the unknown initial state. The experimental results show that the proposed CNN-LSTM network well captures the nonlinear relationships between SOC and measurable variables and presents better tracking performance than the LSTM and CNN networks. In case of unknown initial SOCs, the proposed network fast converges to true SOC and, then, presents smooth and accurate results, with maximum mean average error under 1% and maximum root mean square error under 2%. Moreover, the proposed network well learns the influence of ambient temperature and can estimate battery SOC under varying temperatures with maximum mean average error under 1.5% and maximum root mean square error under 2%.INDEX TERMS State-of-charge estimation, long short-term memory, convolutional neural network, lithium-ion batteries.

Accurate state-of-charge (SOC) estimation is critical for driving range prediction of electric vehicles and optimal charge control of batteries. In this paper, a stacked long short-term memory network is proposed to model the complex dynamics of lithium iron phosphate batteries and infer battery SOC from current, voltage, and temperature measurements. The proposed network is trained and tested using data collected from the dynamic stress test, US06 test, and federal urban driving schedule. The performance on SOC estimation is evaluated regarding tracking accuracy, computation time, robustness against unknown initial states, and compared with results from the model-based filtering approach (unscented Kalman filter). Moreover, different training and testing data sets are constructed to test its robustness against varying loading profiles. The experimental results show that the proposed network well captures the nonlinear correlation between SOC and measurable signals and provides better tracking performance than the unscented Kalman filter. In case of inaccurate initial SOCs, the proposed network presents quick convergence to the true SOC, with root mean square errors within 2% and mean average errors within 1%. Moreover, the estimation time at each time step is sub-millisecond, making it appropriate for real-time applications. INDEX TERMS State-of-charge estimation, lithium iron phosphate batteries, long short-term memory, recurrent neural network, unscented Kalman filter.

For most deep learning practitioners, recurrent networks are often used for sequence modeling. However, recent researches indicate that convolutional architectures may be used to optimize recurrent networks on some machine translation tasks. Problems here are which architecture we should use for a new sequence modeling. By integrating and systematically evaluating the general convolution and recurrent architecture used for sequence modeling, a convolution gated recurrent unit (CNN-GRU) network is proposed for the state-of-charge (SOC) estimation of lithium-ion batteries in this paper. Deep-learning models are well suited for SOC estimation because a battery management system is time-varying and nonlinear. The CNN-GRU model is trained using data collected from the battery-discharging processes, such as the dynamic stress test and the federal urban driving schedule. The experimental results show that the proposed method can achieve higher estimation accuracy than two commonly used deep learning models (recurrent neural network and gated recurrent unit) and two traditional machine learning approaches (support vector machine and extreme learning machine) for SOC estimation of lithium-ion batteries.INDEX TERMS State-of-charge estimation, convolutional gated recurrent unit, lithium-ion battery.

No abstract

As key components in a rotating machinery system, bearings affect the safety of the entire mechanical system. Hence, early-stage monitor of bearing degradation is critical to avoid abrupt mechanical system failure. In this paper, a novel bearing performance assessment model is constructed based on ensemble empirical mode decomposition (EEMD) and affinity propagation (AP) clustering. Unlike most clustering methods, AP clustering, which automatically finds the center of all available clusters, can determine the bearing degradation status without an experience-based selection of the number of degradation states. The original bearing vibration signal is first decomposed by EEMD and its degradation fault features are extracted from the singular-value decomposition of intrinsic mode functions. Then, the degradation features are selected as the input of AP clustering to find the cluster centers of different bearing health statuses: ''normal'', ''slight'', and ''severe''. Last, a health evaluation indicator, referred to as the confidence value, which is obtained from the dissimilarity between actual samples and the various cluster centers, is used to evaluate the bearing health status. To prove the superiority of the approach, the proposed model is compared to various popular clustering methods, including, k-means, k-medoids, fuzzy c-means, Gustafson-Kessel, and Gath-Geva, and commonly used time-domain indicators such as root mean square and kurtosis. The experimental results show that the proposed method outperforms the above time-domain indicators and clustering methods in monitoring early-stage degradation, without presetting the number of clusters.INDEX TERMS Affinity propagation clustering, bearings, ensemble empirical mode decomposition, performance degradation assessment.The associate editor coordinating the review of this manuscript and approving it for publication was Mariela Cerrada. Many signal-processing methods, including various time and frequency domain indices [2]-[5], wavelet transformation (WT) [6]-[9], empirical mode decomposition (EMD) [10]-[12], and ensemble empirical mode decomposition (EEMD) [13]-[15], have been proposed. Theodoros et al. used time-frequency indicators with a wavelet transform to assess the roller bearings' diagnostic performance [16]. Rodney et al. proposed a data-driven approach that relies on time-frequency domain features, including root mean square (RMS), to describe the evolution of bearing faults [17]. Shen et al. used various time-frequency