Networked scheduling for decentralized load frequency control

This paper investigates the scheduling process for multi-area interconnected power systems under the shared but band-limited network and decentralized load frequency controllers. To cope with sub-area information and avoid node collision of large-scale power systems, round-robin and try-once-discard scheduling are used to schedule sampling data among different sub-grids. Different from existing decentralized load frequency control methods, this paper studies multi-packet transmission schemes and introduces scheduling protocols to deal with the multi-node collision. Considering the scheduling process and decentralized load frequency controllers, an impulsive power system closed-loop model is well established. Furthermore, sufficient stabilization criteria are derived to obtain decentralized 𝐻 1 output feedback controller gains and scheduling protocol parameters. Under the designed decentralized output feedback controllers, the prescribed system performances are achieved. Finally, a three-area power system example is used to verify the effectiveness of the proposed scheduling method.


INTRODUCTION
With the vigorous development of power markets, the degree of interconnection among power grid regions is increasing. At present, power grids have developed into multi-area interconnected power systems [1,2] . Generally, there are two communication schemes to transmit data between neighborhood areas, that is, the dedicated channel and the open infrastructure. Since the latter has outstanding advantages over the former, such as higher flexibility and lower cost, it has been widely used in the communication of multi-area interconnected power systems [3][4][5] .
An important indicator to measure power system operation quality is the fluctuation in frequency, and load frequency control (LFC) is widely used to maintain frequency interchanges at scheduled values and stifle frequency fluctuation caused by load disturbances [2,6,7] . Under the open communication infrastructure, the sensor in each area collects data information. Then, data is transmitted to the decentralized controller under the shared but band-limited network. The corresponding control commands are issued to actuators in each area respectively. However, due to the introduction of the network, the design and operation of LFC face some new challenges, such as node collision, data loss and network-induced delay [8][9][10] .
Due to different locations of sub-region power grids and the wide distribution of system components, multichannel transmission is inevitable in multi-area interconnected power systems, while most of the existing results [10][11][12][13] take the general assumption that sampled data is packaged into a single packet to transmit, which is not applied in large-scale multi-area interconnected power systems. On the other hand, the shared network has limited bandwidth, where the simultaneous transmission under multiple channels may cause node congestion. To solve this problem, scheduling protocols have been presented to decide which node to gain access to the communication network [15? ] .
Generally, multi-channel scheduling includes Round-Robin (RR) scheduling [16,17] , try-once-discard (TOD) scheduling [18,19] , and stochastic scheduling [20] . Under the RR scheduling protocol, each node is transmitted periodically with a fixed period whose value is the total number of transmission channels. Under the TOD scheduling, the sensor node with the largest scheduling error has access to the channel. Recently, a time-delay analysis method has been discussed to derive stability criteria for networked control systems (NCSs) which are scheduled by the above three communication protocols [14,21] . Besides, a hybrid system method has been employed to analyze NCSs with variable delays under the TOD communication protocol [19] , where a partial exponential stabilization criterion has been derived. The ∞ filtering of NCSs with multiple nodes has been investigated [22] , in which TOD protocol is used to schedule sampled data. This paper studies the ∞ LFC for multi-area interconnected power systems with decentralized controllers under the shared but band-limited network. RR and TOD protocols are used to schedule the sampling information of different sub-grids, which could greatly improve communication efficiency. Through linear matrix inequality (LMI) technology and Lyapunov analysis methods, sufficient conditions that guarantee the prescribed ∞ performance of the studied system are derived. Decentralized controller gains and protocol parameters are obtained simultaneously. The main contributions are summarized as follows.
(1) RR and TOD protocols are used to deal with the multi-node collision of large-scale power systems, which improve communication efficiency greatly. Compared with the existing LFC methods [10,23,24] , the scheduling process under multi-area transmission schemes is investigated.
(2) An networked power system impulsive closed-loop model is well constructed, which covers the multichannel scheduling, packet dropout, disturbance, and network-induced delays in a unified framework. Compared with the system without disturbance [19] , this paper studies the anti-disturbance performance of the studied system. An ∞ LFC method is presented to obtain decentralized controller gains and scheduling protocol parameters simultaneously.

Notations
Throughout this paper, R stands for the n-dimensional Euclidean space with vector norm · .
{· · · } and {· · · } denote the block-diagonal matrix and block-column vector, respectively. The superscript stands for the transpose of a matrix or a vector. is an identity matrix with an appropriate dimension. Matrix > 0( ≥ 0) means that is a positive definite (positive semi-definite) symmetric matrix. The symbol * is the symmetric term in a matrix. [−ℎ, 0] denotes the Banach type of absolutely continuous functions :

PROBLEM FORMULATION
Considering single-packet size constraints, the dynamic model of multi-area interconnected power systems with decentralized load frequency controllers is constructed in this section. An impulsive system model under TOD or RR protocol is established.

Multiarea decentralized LFC model
Diagram of the multi-area decentralized LFC under scheduling protocols is shown in Figure 1, where data transmission from sensors to decentralize controllers is scheduled by RR or TOD scheduling protocol.
The system model is represented as [23,25] : where ( ) ∈ R is the state vector, ( ) ∈ R is the measurement and ( ) ∈ R is the disturbance, ( ) ∈ R is the control input vector. The multi-area power systems consist of generators, turbines, and governors, where Δ , Δ , Δ , and Δ denote the deviation of valve position, generator mechanical output, load, and the frequency of the ℎ sub-area, respectively. The area control error ( ) is The state and measured output signals are The multi-area power system includes controlled areas, decentralized controllers, which are connected via the network. Then, measurement signals are given by ( ) = ( ) ∈ R , = 1, · · · , , =1 = . Based on the phasor measurement unit, sampling instants of sensors are synchronized in different areas of power systems. Denote sampling instant by , satisfying that 0 = 0 < 1 < · · · < < · · · , +1 − ≤ , lim →+∞ = +∞, where is the maximum allowable transfer interval.
In this paper, we take imperfect network conditions into account, such as data loss and network-induced delay. Denote the sequence after packet loss by { } ⊆ { }; that is, only at sampling instants , the input of the controller can be updated. For = 1, · · · , , an uncertain, time-varying delay ∈ [0, ] is assumed to occur, where is delay upper bound of the ℎ channel and max Letˆ( ) = [ˆ1 ( ) · · ·ˆ( )] ∈ R denote the output signal transmitted to the scheduler. At instant , only one node can be active. Let ∈ {1, · · · , } be the active output node at , which is dependent on the scheduling rule. Then, we haveˆ( Consider the scheduling error between output ( ) and the last available measurementˆ( −1 ): whereˆ( −1 ) ≜ 0, = 1, · · · , . In this paper, controllers and actuators in the ℎ area are event-driven. Let > 0, = 1, · · · , be node weighting matrices. Under TOD scheduling, Under RR scheduling protocol, the active node is selected periodically: In the following, we will design the decentralized LFC law under the communication network. Similar to [23] , a PI controller is used in this paper: Therefore, dynamic model of the scheduled power systems under the decentralized LFC Equation (8) and imperfect network environments can be formalized as

Impulsive model and study objective
From Equation (5), one can obtain that Define an artificial delay ( ) = − , from which one arrives at From Equation (5) and Equation (9), the impulsive power system model can be For system Equation (12) Using scheduling scheme Equation (6) and Equation (7), this paper is to design the decentralized controller Equation (8) such that system Equation (12) is exponentially stable with a prescribed ∞ performance .

MAIN RESULTS
In the following, we first derive sufficient criteria under scheduling scheme Equation (6) and Equation (7) to ensure the exponential stability of system Equation (12) with a prescribed ∞ performance. Then, criteria are proposed to design decentralized controllers under multi-channel transmission.
We use the reciprocally convex approach [27] to deal with the cross item in Equation (18). By using Schur' s complement, one can get In the following, we will prove that (i) with ( ) = 0, the impulsive system Equation (12) is exponentially stable.
It follows from where Then taking Equation (17) Hence, one can conclude that From Equation (24), we have which implies that It follows from Ω < 0 that for ≥ 0 For < 0 , the system takes the form = ( ) + ( ), then its corresponding solution is given by Clearly, there exists the maximum of | ( )| for Therefore, the exponential stability of the system Equation (12) with ( ) = 0 is guaranteed.

Remark 2
The feasibility of the linear martix inequalities has been sufficiently explained in [23,28] . Due to page limitations, this part is omitted here.

Proof:
The detailed derivation process can refer to [29] and the proof of Theorem 1, which is omitted here due to the limited pages. □

Proof:
The detailed derivation process can refer to [29] and the proof of Theorem 3, which is omitted here. □

AN ILLUSTRATIVE EXAMPLE
In the following, we use a three-area power system [23,24,28] interconnected by the shared communication network to demonstrate the effectiveness of main results. Parameters are listed in Table 1.
Case 1: Stability of the studied system under TOD scheduling scheme Equation (6) and ( ) = 0. Correspondingly, state responses and the switching behavior of active nodes are illustrated by Figure 2 and 3, respectively. Clearly, the designed dynamics output feedback controllers can stabilize the three-area power system under the TOD scheduling scheme Equation (6) in the absence of disturbance.
Case 2: Stability of the studied system under RR scheduling scheme Equation (7) and ( ) = 0.
To compare with the TOD protocol Equation (6) Figure 4 and 5. From Figure 2-5, one can obtain that the system can be stable under both scheduling protocols Equation (6), Equation (7). Compared with RR protocol, TOD protocol can achieve dynamic scheduling, which makes the control process more efficient.
According to Theorem 1-4, TOD protocol can degrade into RR protocol under certain conditions. In this case, we take the TOD protocol Equation (6) as an example to verify the anti-disturbance performance of the studied system. Under the disturbance ( ) shown in Figure 6,

CONCLUSIONS
This paper has designed ∞ output feedback decentralized load frequency controllers for multi-area power systems under the shared but band-limited network. TOD and RR protocols have been used to improve communication efficiency. Since transmission priority has been well designed under scheduling protocols, a higher control performance has been obtained. Then, decentralized output feedback controllers with the prescribed  performance have been designed in different areas. Finally, an example has verified the effectiveness of main results.