Currently, most real-world time series datasets are multivariate and are rich in dynamical information of the underlying system. Such datasets are attracting much attention; therefore, the need for accurate modelling of such high-dimensional datasets is increasing. Recently, the deep architecture of the recurrent neural network (RNN) and its variant long short-term memory (LSTM) have been proven to be more accurate than traditional statistical methods in modelling time series data. Despite the reported advantages of the deep LSTM model, its performance in modelling multivariate time series (MTS) data has not been satisfactory, particularly when attempting to process highly non-linear and long-interval MTS datasets. The reason is that the supervised learning approach initializes the neurons randomly in such recurrent networks, disabling the neurons that ultimately must properly learn the latent features of the correlated variables included in the MTS dataset. In this paper, we propose a pre-trained LSTM-based stacked autoencoder (LSTM-SAE) approach in an unsupervised learning fashion to replace the random weight initialization strategy adopted in deep LSTM recurrent networks. For evaluation purposes, two different case studies that include real-world datasets are investigated, where the performance of the proposed approach compares favourably with the deep LSTM approach. In addition, the proposed approach outperforms several reference models investigating the same case studies. Overall, the experimental results clearly show that the unsupervised pre-training approach improves the performance of deep LSTM and leads to better and faster convergence than other models.

/pdf/unsupervised-pre-training-of-a-deep-lstm-based-stacked-2mesdohlf3.pdf

Unsupervised Pre-training of a Deep LSTM-based Stacked Autoencoder for Multivariate Time Series Forecasting Problems.

Fault detection and diagnosis for rotating machinery: A model based on convolutional LSTM, Fast Fourier and continuous wavelet transforms

Wind farms are enormous and complex control systems. It is challenging and valuable to control and optimize wind farms. Their applications are widely used in various industries. Artificial intelligent algorithms are effective methods for optimization problems due to their distinctive characteristics. They have been successfully applied to wind farms. In this paper, several issues in wind farms are presented. Applications of artificial intelligent algorithms in wind farm controllers, Mach number, wind speed prediction, wind power prediction and other problems of wind farms are reviewed. Two future research directions are pointed out to develop artificial intelligent algorithms for wind farm control systems and wind speed and power prediction.

A review of applications of artificial intelligent algorithms in wind farms

At present, as the candidate services in the cloud service pool increase, the scale of the service composition increases rapidly. When the existing intelligent optimization algorithms are used to solve the large-scale cloud service composition and optimization (CSCO) problem, it is difficult to ensure the high precision and stability of the optimization results. To overcome such drawbacks, a new dynamic ant-colony genetic hybrid algorithm (DAAGA) is proposed in this paper. The best fusion evaluation strategy is used to determine the invoking timing of genetic and ant-colony algorithms, so the executive time of the two algorithms can be controlled dynamically based on the current solution quality, then the optimization ability is maximized and the overall convergence speed is accelerated. An iterative adjustment threshold is introduced to control the genetic operation and population size in later iterations, in which the effect of genetic algorithm is reduced when the population closes to optimal solution, only the mutation operation is implemented to reduce the calculation, and the population size is increased to find the optimal solution more quickly. A series of comparison experiments are carried out and the results show that the accuracy and stability of DAAGA are significantly improved for the large-scale CSCO problem, and the time consumption of the algorithm is also optimized.

A dynamic ant-colony genetic algorithm for cloud service composition optimization

Developing a novel fuzzy neutrosophic numbers based decision making analysis for prioritizing the energy storage technologies

The Active Disturbance Rejection Control (ADRC) prefers the cascaded integral system for a convenient design or better control effect and takes it as a typical form. However, the state variables of practical system do not necessarily have a cascaded integral relationship. Therefore, this paper proposes an algebraic substitution method and its structure, which can convert a noncascaded integral system of PID control into a cascaded integral form. The adjusting parameters of the ADRC controller are also demonstrated. Meanwhile, a numerical example and the oscillation control of a flexible arm are demonstrated to show the conversion, controller design, and control effect. The converted system is proved to be more suitable for a direct ADRC control. In addition, for the numerical example, its control effect for the converted system is compared with a PID controller under different disturbances. The result shows that the converted system can achieve a better control effect under the ADRC than that of a PID. The theory is a guide before practice. This converting method not only solves the ADRC control problem of some noncascaded integral systems in theory and simulation but also expands the application scope of the ADRC method.

/pdf/an-adrc-method-for-noncascaded-integral-systems-based-on-5eax7ytspq.pdf

An ADRC Method for Noncascaded Integral Systems Based on Algebraic Substitution Method and Its Structure

In order to handle the non-linear system and the complex disturbance in marine engines, a finite-time convergence active 
disturbance rejection control (ADRC) technique is developed for the control of engine speed. First, a model for the 
relationship between engine speed and fuel injection is established on the basis of the mean value engine model. Then, 
to deal with the load disturbances and model parameter perturbation of the diesel engine, this paper designs an ADRC 
approach to achieve finite-time stability. Finally, simulation experiments show that the proposed method has better control 
effect and stronger disturbance rejection ability in comparison with the standard linear ADRC

https://sciendo.com/pdf/10.2478/amns.2020.1.00004

Finite-time active disturbance rejection control for marine diesel engine

The ultra-low resolution underwater terrain maps of the Arctic region reduce the localization and navigation accuracy of the underwater vehicle relying on terrain-aided navigation. In this paper, we study the navigation ability of Autonomous Underwater Vehicles (AUVs) under the ultralow-resolution terrain map. Firstly, the low-resolution map is transformed into a continuous map by bilinear interpolation. Then, a Terrain-Aided Navigation (TAN) system based on the Particle Filter (PF) is constructed to estimate the state of AUV position by particles. Particles of a random distribution of fixed variance can effectively track targets. However, a fixed variance distribution is not well adapted to many different situations. To improve navigation accuracy and robustness, fuzzy logic is used to estimate the distribution variance of particles under the current terrain gradient dynamically. The simulation results show that our proposed Fuzzy-PF TAN system is robust under various current disturbance situations. The position error of our system is within a map resolution unit of 500 m.

/pdf/study-on-the-arctic-underwater-terrain-aided-navigation-3wyzbxxj8e.pdf

Study on the Arctic Underwater Terrain-Aided Navigation Based on Fuzzy-Particle Filter

A data-driven online ADP control method for nonlinear system based on policy iteration and nonlinear MIMO decoupling ADRC

The feedback PID method was mainly used for the navigating control of an unmanned surface vessel (USV). However, when the intelligent control era is coming now, the USV can be navigated more effectively. According to the USV character in its navigating control, this paper presents a parallel action-network ADHDP method. This method connects an adaptive controller parallel to the action network of the ADHDP. The adaptive controller adopts a RBF neural network approximation based on the Lyapunov stability analysis to ensure the system stability. The simulation results show that the parallel action-network ADHDP method has an adaptive control character and can navigate the USV more accurately and rapidly. In addition, this method can also eliminate the overshoot of the ADHDP controller when navigating the USV in various situations. Therefore, the adaptive stability design can greatly improve the navigating control and effectively overcome the ADHDP algorithm limitation. Thus, this adaptive control can be one of the intelligent ADHDP control methods. Furthermore, this method will be a foundation for the development of an intelligent USV controller.

/pdf/adaptive-navigating-control-based-on-the-parallel-action-1uok8n32wj.pdf

Guichen Zhang

Papers

An ADRC Method for Noncascaded Integral Systems Based on Algebraic Substitution Method and Its Structure

Finite-time active disturbance rejection control for marine diesel engine

Study on the Arctic Underwater Terrain-Aided Navigation Based on Fuzzy-Particle Filter

A data-driven online ADP control method for nonlinear system based on policy iteration and nonlinear MIMO decoupling ADRC

Adaptive Navigating Control Based on the Parallel Action-Network ADHDP Method for Unmanned Surface Vessel