For deep learning-based soft sensors, the lack of interpretability and the consequent unreliability have become one of the most important problems. In this study, a neural network scheme called the deep multiple attention soft sensor (DMASS), which consists solely of attention mechanisms, is proposed to develop a self-interpretable soft sensor. DMASS was established to ensure the self-interpretability of data selection and sensor modeling and try to integrate these originally independent phases into the single scheme. First, the existing attention mechanisms’ core implementation steps are summarized as a unified form, and then the variable attention mechanism and time lag attention mechanism are proposed. When DMASS's training is completed, the obtained attention weights provide the self-interpretable data selection results. Then, a self-attention activation structure (SAAS) is proposed to extract the nonlinear spatio-temporal features of data. The mathematical expression for the extracted feature, the SAAS's attention matrix, the information path diagram for DMASS's training, and the uncertainty-aware interval prediction show the self-interpretability of sensor modeling. Finally, DMASS was applied to predict the thermal deformation of the air preheater rotor, and the validity of DMASS's self-interpretability is verified by the known mechanism analysis and information bottleneck theory. Meanwhile, DMASS's great sensing performance was confirmed through comparison with other novel soft sensors.

Deep learning-based soft sensors (DLSSs) have been demonstrated to exhibit significantly improved sensing accuracy; however, their vulnerability to adversarial attacks affects their reliability, thus hindering their widespread application. To improve the reliability of DLSSs, in this article, we conducted a systematic investigation of the adversarial attack and defense of DLSSs. By considering the task requirements of DLSSs and the actual scenarios that attackers may encounter, a framework based on black-box attack and proactive defense was proposed to realize the adversarial attack and defense of soft sensors. The adversarial attack was implemented through the proposed knowledge-guided adversarial attack (KGAA) method. By reconstructing the optimization model and introducing the mechanism knowledge into the objective function, the KGAA method could overcome the ill-posed problem of adversarial attack optimization when attacking a regression model. Moreover, based on the KGAA, a corresponding KGAA adversarial training defense method was proposed to achieve proactive defense. The attack and defense methods were verified in terms of the thermal deformation sensing of an air preheater rotor. Compared to other attacks, the KGAA exhibited higher imperceptibility, rationality, and stability; it can thus be considered a practical attack. The implementation of KGAA adversarial training enhances the adversarial robustness of DLSSs, thus aiding the defense of DLSSs to various attacks and improving their reliability.

Model-based control methods do not produce satisfactory control results with the batch process control of Czochralski (CZ) silicon monocrystalline with complex nonlinearity, large delay, and time-varying dynamics. Therefore, this paper proposes a data-driven model-free adaptive iterative learning control method (MFAILC) to achieve precision control of the batch process. Firstly, to improve the accuracy of the data-driven model, a novel deep learning model for crystal growth process is established by combining a stacked autoencoder (SAE) and a long short-term memory network (LSTM) to extract the working condition information and the dynamic timing features in process data. Traditional model-based control methods are limited by the difficulties in modeling and by the unmodeled dynamics. So, to overcome this problem, the heater controller and the crystal puller controller are designed, based on the iterative dynamic linearization technology, to ensure that the silicon monocrystalline batch manufacturing process always maintains precision control. Also, the discrete-time extended state observer (ESO) is introduced to compensate for the influence of disturbances, to improve the robustness of the control system. Finally, the efficacy of the proposed method is verified by applying it for the predictive modeling and batch control of thermal field temperature and crystal diameter during crystal growth.

The economic dispatch problem (EDP) is the solid rock to guarantee the stable operation of power systems, which is essentially an optimization problem, i.e., dispatching generators to meet the total demands with the minimum generation cost under the constraint of generation capacity. With the increasing complexity of power systems, higher requirements for the security and robustness of EDP algorithm have been put forward. On the one hand, the explicit information communication between the generators may bring serious security risks; on the other hand, in practical applications, there may be communication delays and communication link failures due to network instability. However, many existing EDP algorithms can hardly meet the requirements of both security and robustness. Thus, this article develops a fully distributed privacy preserving algorithm for EDP with line losses over time-varying and directed communication and extends it to non-ideal communication environments. In particular, the proposed algorithm only relies on the row-stochastic weight matrix, where each generator can allocate the information weight of its neighbor nodes locally, making the approach easier to implement. Moreover, the conditional noise is added to the auxiliary variable to achieve the generator's privacy preservation. The rigorous mathematical analysis reveals that the proposed approach can effectively find the optimal dispatch while achieving privacy preservation of generators under the assumption that the communication delays are arbitrarily significant but bounded. Finally, several cases are presented to testify the convergence and robustness of the proposed approach.