Malicious traffic attack can prolong the transmission delay of normal data flows by exhausting the limited bandwidth with injecting numerous invalid data packets. For the networked load frequency control (LFC) systems, the traffic attacks can cause the control performance deterioration or even instability. In this paper, an adaptive LFC scheme where the controller gains can be flexibly adjusted according to the attack intensities is proposed. Firstly, to quantitatively evaluate the impact of traffic attacks on the network transmission performance, the relationship between attack intensity and transmission delay bounds of normal data flows in multi-hop transmission process is well revealed based on deterministic network calculus. Secondly, to alleviate the adverse impacts of traffic attacks and improve the control performance under low-level attack intensities, the studied systems are described as a series of subsystems with the actual transmission delays of normal data flows as the switching decisions. Thirdly, a controller bank corresponding to all the attack intensities is elaborately constructed. The constraints and update criteria of each controller in the bank are strictly deduced in the switching control framework. Finally, the effectiveness of the proposed adaptive method is verified by four case studies.

This paper proposes a fuzzy k-means cluster based generalized predictive control method for a 1000MW ultra supercritical power plant to improve of the boiler combustion efficiency. Firstly, to fully use the statistic characteristic of the historical data, a fuzzy k-mean cluster network (FKN) is well constructed to derive the local linear models, and the nonlinear dynamic process of studied system is elaborately approximated by the fuzzy combination of the local linear models. Then, a global generalized predictive control (GPC) method is proposed to improve the control performance by using the membership of the current FKN. Different from the traditional GPC, the advantage of proposed GPC is that local GPC is fuzzily combined together to achieve the purpose of global GPC by a scheduling algorithm. Finally, an example illustrates that the proposed control strategy can achieve the satisfactory performance.

The ultra-supercritical power plant is an advanced energy conversion system with high coal utilization and low emission. It is a complex system associated with high nonlinearity, strong coupling, and subjected to various uncertainties and interferences. This article introduces a nonlinear generalized predictive control based on a composite weighted human learning optimization network to incorporate both local linear modeling and local control into a multi-layer dynamic structure. Since composite weighted human learning optimization network can be considered as a linear-in-weight network, the limitation of the non-convex is elaborately broken through incorporating composite weighted human learning optimization models into the convex quadratic programming routine of local linear generalized predictive control. Detailed analysis is given to show the validity and advantage of the proposed method in controlling the ultra-supercritical power plant.