Digital twin uses information-physical feedback, data fusion analysis and iterative decision optimization to simulate the behavior of physical entities based on data in virtual space. It allows for the mapping, visualization, prediction, and intelligent decision-making of physical entities. To construct a digital twin for the municipal solid waste incineration (MSWI) process and improve its safety, economic performance, and environmental sustainability, a rational construction method is required. Therefore, this paper reviews the digital twin construction of MSWI processes based on mechanical grate furnaces. First, the development of digital twin is outlined. Second, the MSWI process is described, and its digital twin framework is established. Third, the key technologies involved in constructing digital twin of MSWI are analyzed. Finally, future research directions and challenges are discussed.

The safety and security in water treatment plants (WTPs) is critically important, since they face various potential risks. Conducting real-time anomaly detection in WTPs is of great significance for risk prevention, and emergency response. This work proposes an China’s information technology-based intelligent video anomaly detection approach for WTPs. This approach employs a unidirectional information transmission framework, and conducts real-time detection through 21 alarm algorithms. The algorithms are classified into 6 types: production safety regulations, smoke and fire, environmental safety, abnormal personnel behaviors, operations of production equipments, and illegal intrusion. Multiple mechanisms are employed to increase the accuracy. Meanwhile, an interactive platform is developed to virtually view, automatically analyze, and permanently store alarm information. The whole system has been put into use in a real WTP in Beijing, experiments are conducted to substantiate the feasibility of the proposed method. After 1 year, 1572 alarms were generated successfully. Among these alarms, 11 emergencies have been detected, including 9 fall downs and 2 smokes.

In this paper, the input delay issue is investigated for a class of nonlinear systems subject to full-state constraints. A control scheme is developed based on adaptive dynamic programming (ADP), the backstepping method, and the event-triggering mechanism. The barrier Lyapunov function (BLF) is used to handle state constraints, where the optimal cost function in each step is also composed of related barrier terms. Different from the traditional Smith predictor, the input delay is compensated for using an auxiliary system. Furthermore, a relative-threshold triggering strategy is utilized to adjust the updating frequency of the controller to save communication resources. By incorporating ADP and BLF into the recursive design of the backstepping method, the closed-loop system is ensured to be semi-globally uniformly ultimately bounded. Finally, a simulation example involving a one-link manipulator is provided to verify the effectiveness of the control method.

This article focuses on the attack-tolerant and fault-tolerant cooperative fuzzy security control problem for nonlinear delayed Markov jump distributed parameter systems (DPSs) under the stochastic sampling that randomly switches between two sampling periods. First, a Takagi-Sugeno (T-S) fuzzy partial differential equation (PDE) model is presented to accurately describe the nonlinear delayed Markov jump DPSs. Subsequently, in consideration of possible random deception attacks and actuator faults, an attack-tolerant and fault-tolerant cooperative fuzzy security control approach with stochastic sampling is proposed under spatially point measurements (SPMs). Then, linear matrix inequality (LMI)-based sufficient conditions that guarantee the stochastically mean-square exponential stability of closed-loop nonlinear delayed Markov jump DPSs are obtained by employing a mode-dependent Lyapunov functional (LF). Lastly, two examples are given to illustrate the effectiveness of the presented control scheme. Note to Practitioners-Cyber-attacks, actuator faults, and time-delays have always been of great concern in security control due to their frequent occurrence in practical Markov jump DPSs. Consequently, when Markov jump DPSs are affected by these factors, system performance deterioration and instability may occur. In particular, existing security control strategies for Markov jump DPSs are currently based almost on infinite dimensional SD control input, and the sampling period is generally assumed to be a fixed constant. However, in practical applications, the sampling period often changes due to transmission distance and unpredictable network-induced changes. To overcome this challenge, an attack-tolerant and fault-tolerant cooperative stochastic SD fuzzy security (AFCSSDFS) control method is proposed in this paper for nonlinear delayed Markov jump DPSs under SPMs. This method offers the advantage of improving the upper bound of the sampling period and can be implemented by few actuators and sensors. By utilizing appropriate LF and inequality techniques, the designed approach in this article novelty transforms the problem of AFCSSDFS controller design into a standard LMI feasibility problem. Two numerical examples are provided to demonstrate the effective of the proposed control strategy.

ABSTRACT Atmospheric pollutant concentration prediction is a critical component of environmental governance. Deep learning‐based methods now dominate this prediction field, yet current deep learning models still face several challenges: inefficient hyperparameter optimization leading to high training costs, static network structures struggling to adapt to the nonlinear spatiotemporal coupling between pollutant concentrations and meteorological factors, and these limitations consequently causing performance fluctuations in cross‐timescale predictions. To address these challenges, this paper proposes a Bayesian‐optimized self‐organizing gated recurrent unit (BO‐SOGRU) model, achieving dual synergy between Bayesian optimization and dynamic self‐organizing mechanisms. First, a hyperparameter self‐configuration mechanism is constructed within the Bayesian optimization framework, where a Gaussian process surrogate model performs global optimal search for key parameters, significantly improving hyperparameter optimization efficiency. Second, a dynamic self‐organizing mechanism is designed, employing Weighted Minkowski Distance (WMD) to measure neuron similarity and integrating the Extended Fourier Amplitude Sensitivity Test (EFAST) to dynamically evaluate neuron contributions, thereby enabling adaptive network structure optimization. Convergence analysis verifies the algorithm's stability and feasibility. Experiments were conducted on two benchmark datasets and real‐world pollutant concentration prediction tasks across multiple time scales, and the prediction accuracy was calculated. Compared with other models of the same type, BO‐SOGRU has significant advantages in prediction accuracy and model compactness.

To support the low-carbon operational goals, it is essential to accurately measure the total phosphorus (TP) removal rate and energy consumption (EC) in wastewater treatment process (WWTP), which is influenced by multirate phenomenon and dynamic characteristics. To solve this issue, the multirate feature interpolation echo state network (MFIESN) is proposed. Initially, the multirate feature interpolation algorithm based on variational autoencoders (VAE) is proposed to discard irrelevant information and unify multirate features. Secondly, the multitask architecture is constructed by interconnecting two reservoirs in parallel to enable information sharing. Subsequently, the online output weights updating algorithm based on recursive least squares (RLS) is proposed, which employs the ℓ0 norm to control sparsity constraints. Concurrently, to accommodate the dynamic changes in WWTP, the regularization parameter is updated by the designed adaptive rule, whose superiority is theoretically proven. Finally, the experimental results on actual WWTP datasets show that the proposed MFIESN exhibits superior modelling accuracy compared to other models.

In order to ensure effluent quality compliance with standards and enhance operational efficiency, the optimal control of wastewater treatment process (WWTP) has been regarded as crucial. However, the excellent control performance and low operational energy consumption are often maintained through the continuous adjustment of controller parameters, which results in an increased computational burden. Therefore, the dual-triggered adaptive fuzzy robust optimal tracking control method is proposed in WWTP. First, based on the optimal framework, the designed adaptive collaborative controller not only can track and control the dissolved oxygen and nitrate nitrogen concentrations simultaneously, but also can ensure a balance between the effluent quality and energy consumption. Second, based on the designed dual dynamic trigger thresholds, the proposed event-triggering mechanism is used to selectively update network parameters, which is able to reduce unnecessary computational resource consumption. Then, the technique of actuator saturation processing is introduced to improve the robustness for adaptive optimal controller. Finally, the stability of the proposed controller of WWTP is rigorously analyzed using Lyapunov theory and the effectiveness is validated through simulation experiments.

A Hilbert Transform-based Preprocessing Method (HTPM) is proposed for training neural networks. Through rigorous mathematical analysis, we establish its effectiveness in data preprocessing, highlight the extra information it introduces to processed data, and demonstrate its potential to reduce network complexity to some extent. Furthermore, we analyze the properties of the Hilbert Transform and show that it can be directly applied to feedforward neural networks as a preprocessing technique. To evaluate its practical use, HTPM is integrated into both SCN and BP networks. Comprehensive experiments on four benchmark datasets, as well as one real-world dataset, confirm the advantages of HTPM. The results indicate that HTPM consistently improves the predictive performance of SCN and BP networks in time-series forecasting tasks.

The accurate prediction of water quality is of critical importance for the effective management of wastewater treatment plants (WWTPs). However, water quality variables exhibit high multi-periodicity, making accurate prediction of water quality more challenging. To address this issue, this study proposes a hierarchical decomposition-ensemble neural network (HDENet) for multi-periodic water quality prediction. HDENet employs a hierarchical decomposition-ensemble framework. Specifically, a two-level hierarchical decomposition algorithm is first proposed: the first level adopts the seasonal-trend decomposition using loess (STL) to decompose the wastewater series into trend, seasonalities, and remainder; the second level employs the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to further separate the seasonal component into distinct intrinsic mode functions (IMFs) representing multiple seasonalities. Then, an ensembled attention-gated recurrent unit (EA-GRU) predictor is designed to predict the multiple seasonalities, with the training efficiency improved by removing redundant components and the prediction accuracy enhanced by weighting for effective components. Experimental results using a real dataset from a WWTP show that HDENet achieves better prediction performance for the multi-periodic water quality, improved by an average of 13.1%, 14.4% and 7.8% for RMSE, MAE and R2 than other comparative models, and offers a more favorable balance between computational complexity and prediction accuracy. The improvement has been demonstrated to be attributed to the decomposition of multi-periodic components by hierarchical decomposition as well as the prediction for the seasonalities with EA-GRU.

This paper investigates the fully distributed consensus control problem for nonlinear multiagent systems (MASs) subject to denial-of-service (DoS) attacks, external disturbances, and unmodeled nonlinearities. To mitigate the adverse effects of such uncertainties, a radial basis function neural network (RBFNN)-based adaptive control law is developed, combined with sign-function-based update rules to ensure robust approximation and compensation. In addressing the communication constraints induced by DoS attacks, a dynamic event-triggered switching control strategy is further proposed to reduce communication load while maintaining resilience against intermittent network failures. To eliminate the reliance on any global information, a fully distributed implementation is achieved, enhancing the scalability and practicality of the control scheme. With the assistance of Lyapunov stability theory, some bounded consensus conditions have been established. Finally, two simulation studies are conducted to demonstrate the effectiveness and robustness of the proposed control approach.

The whole process simulation model of municipal solid waste incineration (MSWI) with fixed parameter setting values makes it difficult to accurately map the combustion state under different working conditions. As a result, the error between the simulation and the actual results is large when the model simulates different working conditions. Therefore, this paper proposes a correction method for the whole process simulation model parameter based on the MSWI plant actual data under digital twin platform. First, the MSWI whole process numerical simulation model is constructed, and a parameter setting values correction strategy is established. Next, by analyzing the reasons for the errors between the simulation results and the actual results in the model, the parameters that need to be corrected are determined. Then, the parameter correction problem is converted into an optimization problem, and the parameter setting values are optimized by an adaptive genetic algorithm (AGA) combined with actual data. Finally, the accuracy of the whole process simulation model and the effectiveness of the proposed parameter value setting correction method are verified by the actual data of the plant. Under the support of the MSWI digital twin platform, the results show that when the grate speed increases, the temperature correction value generally shows a first decreasing and then increasing trend. The fluctuation range of parameter adjustment is −7.55%∼10.11%. When the proportion of primary air gradually increases, the temperature correction value shows a trend of first decreasing and then increasing. The fluctuation range of parameter adjustment is −16.86%∼3.43%.

To guarantee strict pollutant emission compliance and high incineration efficiency in municipal solid waste incineration (MSWI), furnace temperature (FT) must be controlled under strong nonlinearities, thermal inertia, and multisource disturbances. Existing static controllers often lack sufficient temporal memory, while many robust control schemes rely on accurate mathematical models or explicit disturbance observers. To address these limitations, this article proposes a self-organizing (SO) recurrent fuzzy neural network (RFNN) with robust compensation (RC), termed SORFNN-RC, for precise FT control. First, an MSWI-oriented RFNN is developed with a dual-feedback mechanism to exploit historical information. Second, a dual-criteria offline SO mechanism is designed to obtain a compact and interpretable rule base. Third, a lightweight adaptive RC module is introduced to suppress unmodeled dynamics and external disturbances without requiring an accurate process model or an explicit disturbance observer. The semiglobally uniformly ultimately bounded stability of the closed-loop system is proven using Lyapunov theory. Industrial experiments based on real data from an MSWI plant demonstrate that the proposed SORFNN-RC achieves superior tracking accuracy and stability compared with other advanced controllers.

Online learning is widely used for policy optimization in reinforcement learning. However, traditional online learning methods commonly suffer from inefficient exploration in high-dimensional spaces, skewed data distributions, and catastrophic forgetting. To address these limitations, evolutionary critic designs are developed by integrating evolutionary strategies with an explicit diversity-preserving mechanism for policy exploration. A population of policies is maintained and evolved within the framework. Each policy is evaluated using a composite fitness function, in which cost minimization is balanced against a diversity penalty. The best policies are retained across generations through elite preservation. The dual-buffer experience replay is used to mitigate catastrophic forgetting and improve data utilization for policy training. Simulations on a robotic manipulator and a Quanser helicopter are conducted to evaluate the proposed framework. The results indicate improvements in learning reliability, exploration efficiency, and state-distribution uniformity.

Multimodal multiobjective optimization aims to provide diversified acceptable decisions (ADs), including GOS with consistent objective evaluations and local optimal solutions (LOSs) with acceptable objective evaluations. However, the discrimination of LOSs highly depends on the distribution of candidate solutions, which may result in the catastrophic elimination of LOSs to damage the diversity in the decision space. To address this problem, a local regularity model (LRM) method is proposed to improve the distribution of candidate solutions. There are three novelties of LRM. First, a HPCA is developed to extract principal components for different nondominated sets. Then, the distribution features of different nondominated sets are described in segments by a small number of candidate solutions to construct LRM. Second, a self-organization strategy, based on the feature correlation and neighborhood violation analysis, is proposed to improve local fitting ability. Then, LRM are efficiently constructed to estimate the manifold of ADs. Third, a probability reproduction strategy is developed to reconstruct the population by LRM. Then, the population is reconstructed to enhance the distribution of candidate solutions in the decision space. Finally, the proposed optimization method is integrated into the popular multimodal MMOA to demonstrate its effectiveness in terms of the benchmark multimodal MMOP test suite.

Wastewater treatment yields significant social benefits in resource recovery, economic development, and public health. During wastewater treatment, dissolved oxygen (DO) concentration serves as a critical indicator for assessing effluent quality. To achieve precise tracking of DO concentration, this paper proposes an optimal tracking control method based on the event-triggered (ET) mechanism. The ET mechanism is incorporated into adaptive dynamic programming (ADP) to enable precise tracking of DO concentration. By integrating the ET mechanism with ADP, the limitation associated with fixed-cycle triggering is overcome. Simultaneously, this integrated approach substantially reduces the computational load, thereby achieving synergistic optimization of control performance and resource consumption. Finally, simulation experiments are conducted using the benchmark simulation model no. 1. The results demonstrate that the proposed method achieves precise tracking of DO concentration, validating the effectiveness of the algorithm.

Fire and smoke segmentation in industrial combustion scenarios is crucial for precise combustion monitoring and safety assessment. To advance research in this domain, we propose the first large-scale benchmark dataset specifically designed for semantic segmentation of fire and smoke in industrial combustion environments. The dataset comprises one hundred thousand finely annotated images extracted from real combustion videos captured across multiple factory settings, demonstrating significant environmental diversity including diverse weather conditions, varying illumination intensities, and different combustion levels. Each image is provided with pixel-level annotations. We systematically evaluate various representative deep learning segmentation architectures, presenting comprehensive performance benchmarks and comparative analysis. Experimental results not only validate the dataset's effectiveness but also establish reliable baselines for future research. This dataset is expected to make significant contributions to the advancement of fire and smoke segmentation research and its practical applications in industrial combustion scenarios.