面向双碳目标的自动化和智能化理论与技术

基于国家自然科学基金委员会第324期双清论坛,本文针对面向双碳目标高耗能工业低碳运行与多介质能源协同减碳调控的国家重大需求,围绕低碳工业智能化和多能互补协同智能调控的自动化与智能化系统理论、关键技术及面向应用层面的基础性问题,分析了面向双碳目标的自动化和智能化的现状与发展趋势。在低碳工业智能化方面,聚焦工业生产全流程碳排放智能建模方法,低碳工业生产全流程数字化网络化智能化,流程工业低碳绿色制造,制造业异质能源综合利用与优化调控;在多能互补协同智能调控方面,聚焦研究多介质能源转化,多介质能源供给协同调控,多能互补与源储荷调控,能源“源-网-荷-储”一体化决策与综合安全,零碳智慧能源系统的结构化变革,城市智慧能源管控。围绕上述内容,讨论了面临的挑战,给出了凝练的科学问题与主要研究方向,提出了相关的政策建议。

An Intelligent Control Method for the Low-Carbon Operation of Energy-Intensive Equipment

Based on an analysis of the operational control behavior of operation experts on energy-intensive equipment,this paper proposes an intelligent control method for low-carbon operation by combining mechanism analysis with deep learning,linking control and optimization with prediction,and integrating decision-making with control.This method,which consists of setpoint control,self-optimized tuning,and tracking control,ensures that the energy consumption per tonne is as low as possible,while remaining within the target range.An intelligent control system for low-carbon operation is developed by adopting the end-edge-cloud collaboration technology of the Industrial Internet.The system is successfully applied to a fused magnesium furnace and achieves remarkable results in reducing carbon emissions.

Intelligent Manufacturing for the Process Industry Driven by Industrial Artificial Intelligence

Based on the analysis of the characteristics and operation status of the process industry,as well as the development of the global intelligent manufacturing industry,a new mode of intelligent manufacturing for the process industry,namely,deep integration of industrial artificial intelligence and the Industrial Internet with the process industry,is proposed.This paper analyzes the development status of the existing three-tier structure of the process industry,which consists of the enterprise resource planning,the manufacturing execution system,and the process control system,and examines the decision-making,control,and operation management adopted by process enterprises.Based on this analysis,it then describes the meaning of an intelligent manufacturing framework and presents a vision of an intelligent optimal decision-making system based on human–machine cooperation and an intelligent autonomous control system.Finally,this paper analyzes the scientific challenges and key technologies that are crucial for the successful deployment of intelligent manufacturing in the process industry.

Nonlinear Decoupling PID Control Using Neural Networks and Multiple Models

For a class of complex industrial processes with strong nonlinearity, serious coupling and uncertainty, a nonlinear decoupling proportional-integral-differential （PID） controller is proposed, which consists of a traditional PID controller, a decoupling compensator and a feedforward compensator for the unmodeled dynamics. The parameters of such controller is selected based on the generalized minimum variance control law. The unmodeled dynamics is estimated and compensated by neural networks, a switching mechanism is introduced to improve tracking performance, then a nonlinear decoupling PID control algorithm is proposed. All signals in such switching system are globally bounded and the tracking error is convergent. Simulations show effectiveness of the algorithm.

A Primal-Dual SGD Algorithm for Distributed Nonconvex Optimization

The distributed nonconvex optimization problem of minimizing a global cost function formed by a sum of n local cost functions by using local information exchange is considered.This problem is an important component of many machine learning techniques with data parallelism,such as deep learning and federated learning.We propose a distributed primal-dual stochastic gradient descent(SGD)algorithm,suitable for arbitrarily connected communication networks and any smooth(possibly nonconvex)cost functions.We show that the proposed algorithm achieves the linear speedup convergence rate O(1/(√nT))for general nonconvex cost functions and the linear speedup convergence rate O(1/(nT)) when the global cost function satisfies the Polyak-Lojasiewicz(P-L)condition,where T is the total number of iterations.We also show that the output of the proposed algorithm with constant parameters linearly converges to a neighborhood of a global optimum.We demonstrate through numerical experiments the efficiency of our algorithm in comparison with the baseline centralized SGD and recently proposed distributed SGD algorithms.

Neural-network-based two-loop control of robotic manipulators including actuator dynamics in task space

A neural-network-based motion controller in task space is presented in this paper. The proposed controller is addressed as a two-loop cascade control scheme. The outer loop is given by kinematic control in the task space. It provides a joint velocity reference signal to the inner one. The inner loop implements a velocity servo loop at the robot joint level. A radial basis function network （RBFN） is integrated with proportional-integral （PI） control to construct a velocity tracking control scheme for the inner loop. Finally, a prototype technology based control system is designed for a robotic manipulator. The proposed control scheme is applied to the robotic manipulator. Experimental results confirm the validity of the proposed control scheme by comparing it with other control strategies.

Hybrid intelligent control of combustion process for ore-roasting furnace

Because of its synthetic and complex characteristics, the combustion process of the shaft ore-roasting furnace is very difficult to control stably. A hybrid intelligent control approach is developed which consists of two systems： one is a cascade fuzzy control system with a temperature soft-sensor, and the other is a ratio control system for air flow with a compensation model for heating gas flow and air-fuel ratio. This approach combined intelligent control, soft-sensing and fault diagnosis with conventional control. It can adjust both the heating gas flow and the air-fuel ratio in real time. By this way, the difficulty of online measurement of the furnace temperature is solved, the fault ratios during combustion process is decreased, the steady control of the furnace temperature is achieved, and the gas consumption is reduced. The successful application in shaft furnaces of a mineral processing plant in China indicates its effectiveness.

Multiple-model-and-neural-network-based nonlinear multivariable adaptive control

A multivariable adaptive controller feasible for implementation on distributed computer systems （DCS） is presented for a class of uncertain nonlinear multivariable discrete time systems. The adaptive controller is composed of a linear adaptive controller, a neural network nonlinear adaptive controller and a switching mechanism. The linear controller can provide boundedness of the input and output signals, and the nonlinear controller can improve the performance of the system. The purpose of using the switching mechanism is to obtain the improved system performance and stability simultaneously. Theory analysis and simulation results are presented to show the effectiveness of the proposed method.

Modeling and Hybrid Optimization of Batching Planning System for Steelmaking-continuous Casting Process

This paper investigates the batching problem for steelmaking and continuous casting production in an iron and steel enterprise. The tasks of this problem are to decide how to select slabs and determine their width, how to group the selected slabs into charges and then group the charges into tundishes, how to determine the sequence of charges in each tundish, and how to group tundishes into casts and determine the sequence of tundishes in each cast. The effective decision on the batching problem can help balance the requirements of the sequential process after steelmaking and continuous casting,reduce production cost, and improve slab quality. We first give the mathematical description of the original problem. Based on the analysis of width, we present a decomposition strategy to divide the model into three sub-models, i.e., charge design model,tundish design model and cast design model, while adding relevant objectives and constraints. According to the characteristics of each sub-model, we present hybrid optimization algorithms separately. Computational experiments show the strategy, models and algorithms can generate satisfactory solutions.

DeceFL:a principled fully decentralized federated learning framework

Traditional machine learning relies on a centralized data pipeline for model training in various applications;however,data are inherently fragmented.Such a decentralized nature of databases presents the serious challenge for collaboration:sending all decentralized datasets to a central server raises serious privacy concerns.Although there has been a joint effort in tackling such a critical issue by proposing privacy-preserving machine learning frameworks,such as federated learning,most state-of-the-art frameworks are built still in a centralized way,in which a central client is needed for collecting and distributing model information(instead of data itself)from every other client,leading to high communication burden and high vulnerability when there exists a failure at or an attack on the central client.Here we propose a principled decentralized federated learning algorithm(DeceFL),which does not require a central client and relies only on local information transmission between clients and their neighbors,representing a fully decentralized learning framework.It has been further proven that every client reaches the global minimum with zero performance gap and achieves the same convergence rate O(1=T)(where T is the number of iterations in gradient descent)as centralized federated learning when the loss function is smooth and strongly convex.Finally,the proposed algorithm has been applied to a number of applications to illustrate its effectiveness for both convex and nonconvex loss functions,time-invariant and time-varying topologies,as well as IID and Non-IID of datasets,demonstrating its applicability to a wide range of real-world medical and industrial applications.

Intelligent optimal control system for ball mill grinding process

Operation aim of ball mill grinding process is to control grinding particle size and circulation load to ball mill into their objective limits respectively, while guaranteeing producing safely and stably. The grinding process is essentially a multi-input multi-output system （MIMO） with large inertia, strong coupling and uncertainty characteristics. Furthermore, being unable to monitor the particle size online in most of concentrator plants, it is difficult to realize the optimal control by adopting traditional control methods based on mathematical models. In this paper, an intelligent optimal control method with two-layer hierarchical construction is presented. Based on fuzzy and rule-based reasoning （RBR） algorithms, the intelligent optimal setting layer generates the loops setpoints of the basic control layer, and the latter can track their setpoints with decentralized PID algorithms. With the distributed control system （DCS） platform, the proposed control method has been built and implemented in a concentration plant in Gansu province, China. The industrial application indicates the validation and effectiveness of the proposed method.

Prediction method for energy con sumption per ton of fused magnesium furnaces using data driven and mechanism model

The electric energy consumed in every ton of acceptable product, namely energy consumption per ton (ECT), is an important overall index for the production process of a fused magnesium furnace. The furnace is the equipment for producing the fused magnesia. The ECT value depends on the current in the smelting process. The optimal operation for a fused magnesium furnace is supposed to have the ECT as low as possible, where the key is to predict ECT accurately. By introducing an unknown high-order non linear term, this paper builds a dynamic ECT model for differe nt production batches based on the static ECT model for one batch. The average current is taken as the input of the dynamic ECT model, which is composed of the unknown high-order nonlinear term and a nonlinear model with unknown parameters. The order of the nonlinear term is determined by the distance correlatio n and the nonlinear term is estimated by the stochastic con figuration n etwork, while the parameters of the non linear model is ide ntified by the least square method. The estimation of the nonli near term alter nates with the parameter identification. This paper proposes a prediction method for ECT, which is composed of the order identification of the non linear term, the alternating identification of the model and the ECT prediction model. The simulation experiments are conducted by the on-site data, and the results verify the effectiveness of the proposed prediction method.