CN102710515A - Deadband scheduling method applicable to networked control systems - Google Patents

Abstract

The invention relates to a deadband scheduling method applicable to network control systems and belongs to the technical field of networked control systems (NCS). The deadband scheduling method is characterized in that the deadband scheduling method aims to ensure that the output of a plurality of NCSs sharing a same network is satisfied to enter plus or minus 5 percent or plus or minus 2 percent of fluctuation range of a steady-state value of each system, a threshold value delta of each NCS deadband is selected as 0.05 or 0.02, a deviation rate ec(k) threshold value gamma are selected as 0.025, and whether a controller node needs to transmit a data pocket to an actuator node is determined through taking e(k) and ec(k) as double constraints; when the |e(k)| is less than Delta, and the |ec(k)| is more than Gamma, each system is in a stable state, and the controller node is not required to transmit a new data packet to the actuator node through the shared network; and when the |e(k)| is more than or equal to delta or the |ec(k)| is more than or equal to gamma, each system is in a transition process, the controller node is required to transmit a new data pocket to the actuator node through the shared network. With the adoption of the deadband scheduling method, network bandwidth resources can be saved, the bandwidth utilization rate is improved, and meanwhile, the systems not only can stabilize, but also can satisfy the steady-state quality requirements.

Description

A kind of dead band dispatching method that is applicable to network control system

Technical field

The present invention relates to network control system, relate in particular in the limited a plurality of network control systems of bandwidth resources, the collaborative design method of control and scheduling belongs to the network control system technical field.

Background technology

In dcs, transducer and controller, the close-loop feedback control system that constitutes through the real time communication network between controller and the actuator is called network control system (Networked control systems, note by abridging be NCS).

NCS compares with the control system of traditional point-to-point structure, have cost low, be easy to information sharing, be easy to advantages such as expansion is safeguarded, flexibility is big, be widely used in recent years in the industrial process control.

Along with deepening continuously of NCS research; Scheduling causes gradually that with the thought of control collaborative design researcher's concern is not only relevant with control strategy with the overall performance of paying attention to .NCS; It is also closely related with the reasonable use and the dispatching algorithm of network bandwidth resources. although the theoretical method of NCS has been obtained bigger development with application; But the resource-constrained problem of available network bandwidth extensively exists in NCS; And do not solved preferably. from the angle of the communication technology, the bus network bandwidth is quite limited often in typical N CS uses; Angle from application demand; NCS of today often works under the dynamic environment, and offered load demonstrates more time-varying characteristics. the network bandwidth limited with the variable direct result of load be exactly the uncertain of available resources, the time then show as uncertain communication delay, packet loss and shake on the step response; And finally influence control of quality (Quality ofControl; QoC) and network service quality (Quality ofService QoS), even causes system unstable.

Network scheduling is meant that the node in the network sends data in shared network resource; When bumping; The specified data bag is with what kind of priority (in proper order) and the problem of when sending packet. and its objective is network resources demand is as far as possible reasonably distributed; Make whole NCS can reach the performance requirement of our expectation. the network scheduling of research concentrates on application layer mostly at present; The application layer scheduling is the sending order of the application program on upper strata according to demand active distribute data, occurs in the process of transmission data between sensor node, controller node and the actuator node.

The structure of the multiloop NCS of the shared consolidated network resource that the present invention studied is as shown in Figure 1; And among Fig. 1, the typical structure of the single NCS of formation multiloop NCS is as shown in Figure 2.

The scheduling of information among the NCS; Dynamic characteristic by algorithm can be divided three classes: static scheduling strategy, dynamic dispatching strategy and sound attitude mixed scheduling strategy. and dead band (Deadband) scheduling belongs to dynamic dispatching; Be when satisfying the control system performance requirement, through the node in the network being provided with the transmission dead band, the data volume of control accesses network; Save Internet resources; Initiatively abandoning the network packet of some, to alleviate the size of offered load, is one of method that guarantees the control system stable performance.

Its basic skills is:

1) as | X-X _Sent| during>=δ, send X, and the value of X is composed to X _Sent

2) as | X-X _Sent| during＜δ, do not send X.

In the formula: X is that node is ready for sending the data in the network; X _SentFor the node last time sends to the data in the network; δ is the dead band threshold values.

Usually after weighing interpolation dead band dispatching algorithm, systematic function has two indexs: integration IAE of Error Absolute Value (Integral of Absolute Error) and network saving rate N _s.

In continuous control system, IAE may be defined as:

$\mathrm{IAE}={\∫}_{t_0}^{t_f}|y_{\mathrm{des}}\left(t\right)-y_{\mathrm{act}}\left(t\right)|\mathrm{dt}---\left(1\right)$

In the formula: y _Des(t) be system's desired output, y _Act(t) be the actual output of system.

Because NCS is a continuous and discrete control system that mixes, sampled signal disperses. therefore, can formula (1) be rewritten into:

$\mathrm{IAE}=\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{h}_i[y_{\mathrm{des}}\left(t_k\right)-y_{\mathrm{act}}\left(t_k\right)]---\left(2\right)$

In the formula: h _iBe systematic sampling step-length, t _kBe sampling instant.

Control performance quality Q oC of system (Quality of Control) and accumulated error absolute value IAE have following relation:

$\mathrm{QoC}=\frac{1}{\mathrm{IAE}}={\left[\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{h}_i\right[y_{\mathrm{des}}\left(k_t\right)-y_{\mathrm{act}}\left(t_k\right)\left]\right]}^{-1}---\left(3\right)$

The value of IAE is more little, and promptly the Error Absolute Value of accumulation is more little, explains that the control performance in loop is good more.

Network saving rate N _sBe defined as:

$N_s=\frac{P_{\mathrm{total}}-P_{\mathrm{act}}}{P_{\mathrm{total}}}---\left(4\right)$

In the formula: P _TotalShould send data packet number during for no dead band; P _ActThe data packet number of actual transmission when the dead band is arranged.

Increase along with dead band δ; The data volume that node sends in the network will reduce; Accumulated error absolute value IAE will increase; The performance of system will variation .NCS performance optimization require the minimizing of equilibrium criterion transmission quantity, through increasing dead band δ but the design limit that need guarantee to be no more than IAE is selected the optimal value of δ. the quality of systematic function is not only relevant with the size in dead band, but also relevant with the position of The dead time.

Based on above-mentioned dead band dispatching method, Chinese scholars has been done number of research projects, and has obtained certain achievement, but still exists following problem not solve well as yet:

1) though the dead band dispatching method has defined the decision rule of data packet transmission; But still concrete and clear and definite definition are not made in the selection of δ; Especially not from guaranteeing system same network, be the selection problem that target is confirmed δ directly to satisfy a plurality of NCS performance quality indexs of sharing this network.

Utilization dead band dispatching algorithm; The selection of dead band threshold values generally is to confirm through emulation repeatedly. still, confirms the size in dead band, is difficult to find the optimal value of dead band δ usually through emulation; And after the optimal value of confirming dead band δ; In case can not change on the structure principle of system. the increase and decrease of node is arranged on the network, and network traffics have greatly changed, and just need off-line to seek the optimal value of dead band δ again; Thereby cause and can't adapt among the current NCS, the situation that bandwidth constraints and live load node change.

2) along with the increase of dead band δ, the data volume that node sends in the network will reduce, and the IAE index will increase; Yet the performance of system will variation., IAE is increased to any degree actually, and the performance of system will variation; And performance can variation arrive any degree, and is still unknown, all its variation tendency can only be described; Being that IAE is a qualitative index, is a cumulative index.

3) the dead band dispatching algorithm is arranged in the sensor node and realizes; Sampling is with the collection of data in the sensor node and send two processes and separately carry out; The data that only collect do not drop within the dead zone range, just are allowed to be sent to controller node. and can reduce the packet number of whole transmission through network like this, reduce some unnecessary data transmission; Reduce the busy extent of network; Alleviate the load of network. still,, thereby cause the controlling performance of each NCS control loop can receive influence to a certain degree inevitably because the controlled device state information that controller node is received not is complete information.

4) in order to improve the problem of using the dead band sampling may lose a part of controlled device state effective information in the sensor node; The big I that adopts dead band feedback network dispatching method dynamically to adjust dead band δ is improved the controlling performance of NCS. still; Because the feedback network scheduler is as a separate network node; Need with system in controller node and the sensor node of each NCS carry out real time communication: one side takies node resource, also can extraly take network bandwidth resources when communicating by letter between other node; Also possibly directly influence real-time and validity that dispatching algorithm is implemented on the other hand owing to the propagation delay time that reasons such as offered load fluctuation cause; And the fault of feedback network scheduler will cause whole NCS performance sharply to descend; Even paralysis. simultaneously; The feedback network dispatching method the adjustment dead band δ big or small the time, still use the IAE qualitative index.

Therefore, seek the straightforward procedure of asking for dead band δ optimal value, directly the new quantitative target of characterization control system performance quality and the relation of δ have become the key technical problem that the dead band Study of Scheduling need solve among many NCS.

Summary of the invention

In order to solve above-mentioned key technical problem, the present invention proposes the dead band dispatching method of a kind of NCS of being applicable to.

The object of the invention:

In the dispatching algorithm of NCS dead band; The problem that dead band threshold values δ optimal value is difficult to ask for; The present invention proposes a kind ofly to guarantee that system satisfies in same network; A plurality of NCS performance quality indexs of sharing this network are the new method that target is confirmed δ, in order to solve the control and the collaborative design problem of scheduling of NCS in the network environment that becomes when the information. the line dynamic that realizes Internet resources is distributed, the ability of raising NCS response environment variation.

The method that the present invention adopts is:

The first step: in controller node, definition deviation e (k) is: e (k)=y _Des(k)-y _Act(k); Simultaneously, definition deviation variation rate ec (k) is: ec (k)=e (k)-e (k-1).

Second step: the rate of change ec (k) of selected deviation e (k) and deviation comes in the common decision controller node as double constraints; The condition that packet sends to the actuator node. and the threshold values of deviation e (k) (being the dead band) is defined as parameter δ, and the threshold values of deviation variation rate ec (k) is defined as parameter γ simultaneously.

The 3rd step: with the output of each NCS get into its steady-state value ± 5% (or ± 2%) fluctuation range is a target, the value of directly setting the dead band threshold values δ of each NCS is 0.05 (or 0.02), the threshold values γ of deviation variation rate ec (k) is 0.025.

The 4th step: as | e (k) |＜δ and | ec (k) | during＜γ, controller node does not send packet.

This decision rule shows: when system gets into stable state (its its steady-state value of output entering ± 5% (or ± 2%) fluctuation range), controller node need not to send new packet through shared network to the actuator node.

The 5th step: as | e (k) | >=δ or | ec (k) | during >=γ, controller node sends packet.

This decision rule shows: system is in the transient process state, and controller node need send new packet to the actuator node through shared network, to strengthen control action, guarantees that system finishes the transient process state as soon as possible.

The scope of application of the present invention:

The present invention is applicable to a plurality of NCS that share same network, in controller node, implements the scheduling of dynamic dead band.

The invention is characterized in that this method may further comprise the steps:

1, when sensor node during, will adopt mode A to carry out work by the periodic sampling signal triggering;

2, when sensor node with Be Controlled object real output signal y _Act(k) through shared network when controller node transmits, will adopt mode B to carry out work;

3, when controller node by y _Act(k) during signal triggering, will adopt mode C to carry out work;

4, when controller node with control signal u (k) through shared network when actuator transmits, will adopt mode D to carry out work;

5, when the actuator node is triggered by u (k), will adopt mode E to carry out work;

The step of mode A comprises:

A1: sensor node works in the time type of drive, and its triggering signal is the periodic sampling signal;

A2: after sensor node is triggered, to the output signal y of controlled device G (s) _Act(k) sample;

The step of mode B comprises:

B1: the signal y that sensor node obtains sampling _Act(k), transmit to controller node through shared network;

B2: signal y _Act(k) will experience network transfer delay τ _ScAfter, could arrive controller node.

The step of mode C comprises:

C1: controller node works in event driven manner;

C2: controller node is by signal y _Act(k) trigger;

C3: in controller node, the given signal y of employing system _Des(k) deduct Be Controlled object real output signal y _Act(k), obtain the deviation e (k) of system, i.e. e (k)=y _Des(k)-y _Act(k), also obtain the rate of change ec (k) of system deviation simultaneously, i.e. ec (k)=e (k)-e (k-1);

C4: to system deviation e (k) and deviation variation rate ec (k), controller C (s) implements its control computing according to selected in advance control strategy (conventional control or Based Intelligent Control), and controlled signal u (k);

C5: if system satisfies dead band dispatching algorithm decision rule 1: promptly work as | e (k) |＜0.05 and | ec (k) | in the time of＜0.025 (its steady-state value of output entering in corresponding each NCS loop ± 5% fluctuation range); Or as | e (k) |＜0.02 and | ec (k) | in the time of＜0.025 (output in corresponding each NCS loop gets into its steady-state value ± 2% fluctuation range); Be that system is when having got into stable state; Controller node need not to send new packet to the actuator node again through shared network. and controller node will be in the situation that sky closes this moment, wait for next sampled data y _Act(k+1) arrival, system carries out work with echo plex mode A. and at this moment, can save network bandwidth resources, simultaneity factor was not only stable but also satisfy the steady-state behaviour quality requirement;

C6: if system satisfies dead band dispatching algorithm decision rule 2: promptly as | e (k) | >=0.05 or | ec (k) | in the time of >=0.025; Or as | e (k) | >=0.02 or | ec (k) | in the time of >=0.025; Be that system is in the transient process state; Controller node need send new packet to the actuator node through shared network; To strengthen control action, guarantee that system finishes the transient process state as soon as possible, system carries out work with mode of entrance D;

The step of mode D comprises:

D1: controller node transmits through shared network control signal u (k) to the actuator node;

D2: control signal u (k) will experience network transfer delay τ _CaAfter, could arrive the actuator node;

The step of mode E comprises:

E1: the actuator node works in event driven manner;

E2: actuator node controlled signal u (k) triggers;

E3: u (k) as carrying out drive signal, is implemented networking control to controlled device G (s).

The present invention has following advantage:

1, the dynamic dispatching method that proposes of the present invention is to be final goal with the control performance quality (QoC) that satisfies NCS; With the output (controlled variable) of each NCS get into its steady-state value ± 5% (or ± 2%) fluctuation range; And the dynamic change rate of deviation is as judging among each NCS; Whether controller node need transmit the judgment basis of data through shared network to the actuator node; And then size that can be through dynamic adjustment offered load is to guarantee the control performance quality of each NCS. the controller node among each NCS has alone according to its control performance quality and whole network condition autonomous and the ability of adjusting the load size adaptively; To guarantee effective utilization of the whole network bandwidth, strengthened the collaborative design of control with scheduling.

2, the dynamic dispatching method that proposes of the present invention is a kind of real-time, online and dynamic dispatching method. need not to the operation conditions of network predict, estimation or identification (because parameters such as the size that wants estimation network load exactly at present and network condition are very difficult); With the static scheduling algorithm (be that off-line is realized mostly; The time variation that is difficult to adapt to information flow in the network is with uncertain) or dynamic dispatching algorithm (complex algorithm; Usually expense that need be bigger) or through changing the sampling period (the change meeting of sample rate causes shake in system; The control algolithm parameter need be adjusted again; Expend the additional calculation time) or feedback network dispatching method (need independently be provided with feedback network scheduler node), one side takies network node resource, also can extraly take network bandwidth resources when communicating by letter between other node; The propagation delay time that also causes owing to reasons such as offered load fluctuations on the other hand directly influences real-time and the validity that dispatching algorithm is implemented; And the fault of feedback network scheduler will cause whole NCS performance sharply to descend; Even cause methods such as systemic breakdown to compare; New dispatching method is more simple, and its physical significance is clear and definite more, and realizes more easily.

3, the selection of the use of the dynamic dispatching method of the present invention's proposition and procotol is irrelevant: both can be used for deterministic network, and also can be used for the uncertainty network; Both can be used for cable network, also can be used for wireless network; Or the complex network of wired (wireless) isomery (mixing).

4, in the use and controller node of the dynamic dispatching method of the present invention's proposition, the selection of control strategy is irrelevant: both can be used for adopting the NCS of conventional PID controller formation, and also can be used for the NCS. that adopts various advanced persons (intelligence) controller to constitute

5, the dynamic dispatching method of the present invention's proposition need not extra increase network dispatcher, and the controller node in each NCS just is enough to implement, and has both saved network bandwidth resources, saves node resource again, thereby has more the engineering actual application value.

Description of drawings

Fig. 1 is for sharing the multiloop NCS structured flowchart of consolidated network resource.

Fig. 2 is the typical structure of single loop NCS.

In the multiloop NCS of shared consolidated network resource shown in Figure 1 structured flowchart; Have n separate NCS to share the consolidated network resource, the control loop of each NCS all is made up of a controlled device, a sensor node, a controller node and an actuator node.

Wherein: sensor node adopts the time type of drive to carry out work, and the triggered time is the sampling period, and to the sampling of controlled device implementation cycle, then sample information is sent to controller node through shared network;

Controller node adopts event driven manner to carry out work, and calculates control signal, then control signal is sent to the actuator node through shared network;

The actuator node adopts event driven manner to carry out work, and controlled signal triggers, and its output signal node changes the controlled device state, realizes the control action to controlled device.

In the typical structure of single loop NCS shown in Figure 2, system comprises given signal y _Des, controlled device real output signal y _Act, controlled device G (s), feedback network network delay

Controller C (s) and through path network delay

The unit, what need to prove that here through path network and feedback network network and other n-1 separate NCS share is same network bandwidth resources.

Embodiment

To make the clearer above-mentioned feature and advantage of the present invention of those of ordinary skill in the art through describing exemplary embodiment of the present invention in detail below with reference to accompanying drawing 1 and accompanying drawing 2.

Need to prove that the dead band dispatching method that is applicable to network control system proposed by the invention is to place the controller node of each NCS control loop of sharing same Internet resources to realize.

The practical implementation step is described below:

The first step: the sensor node of each NCS control loop that works in the time type of drive is in its sampling time separately; Output valve to its corresponding controlled device is carried out periodic sampling; And with the controlled device real output signal of its sampling gained through shared network, to the controller node transmission of its each corresponding NCS control loop;

Second step: the controller node that works in each NCS control loop of event driven manner; The controlled device real output signal that is transmitted through the feedback network path triggers; And the set-point of its each NCS control loop deducted its actual output signal value, obtaining its corresponding system deviation value e (k) (is e (k)=y _Des(k)-y _Act(k)), simultaneously this deviate and last deviate are subtracted each other, obtain the rate of change ec (k) (being ec (k)=e (k)-e (k-1)) of system deviation; The controller node of each NCS control loop is to its system deviation e (k) and its deviation variation rate ec (k); Controller C (s) is according to selected in advance control strategy (conventional control or Based Intelligent Control); Implement its control computing, and obtain the control signal u (k) of each NCS control loop;

The 3rd step: the controller node of each NCS control loop; To the rate of change ec (k) of its deviate e (k) and deviation be judged; To confirm that controller node whether need be through shared network to its corresponding actuator node u (k) that transmits control signal, its dead band dispatching algorithm decision rule is:

As | e (k) |＜0.05 and | ec (k) | in the time of＜0.025 (output of corresponding each NCS control loop gets into its steady-state value ± 5% fluctuation range); Or as | e (k) |＜0.02 and | ec (k) | in the time of＜0.025 (output of corresponding each NCS control loop gets into its steady-state value ± 2% fluctuation range); Be that system is when having got into stable state; Controller node need not to send new packet through shared network to its corresponding actuator node; Controller node will be in the situation that sky closes this moment, wait for next sampled data y _Act(k+1) arrival, system will return the first step and carry out work. and at this moment, can save network bandwidth resources, simultaneity factor was not only stable but also satisfy the steady-state behaviour quality index;

As | e (k) | >=0.05 or | ec (k) | in the time of >=0.025; Or as | e (k) | >=0.02 or | ec (k) | in the time of >=0.025; Be that system is in the transient process state. this moment, each controller node need be through shared network to its corresponding actuator node u (k) that transmits control signal; To strengthen control action; Guarantee that system finishes the transient process state as soon as possible, system will get into for the 4th step and carry out work;

The 4th step: work in the actuator node of each NCS control loop of event driven manner, each the control signal u (k) that is transmitted through the feedback network path triggers, and u (k) as carrying out drive signal, is implemented networking control to its controlled device.

The 5th step: return the first step.

The above is merely preferred embodiment of the present invention, and is in order to restriction the present invention, not all within spirit of the present invention and principle, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.

The content of not doing in this specification to describe in detail belongs to this area professional and technical personnel's known prior art.

Claims (1)

1. dead band dispatching method that is applicable to network control system; It is characterized in that this method with the output of satisfying each network control system get into its steady-state value ± 5% (or ± 2%) fluctuation range is a target; Directly in the controller node of each network control system, the value of setting its dead band threshold values δ is 0.05 (or 0.02), and the threshold values γ of deviation variation rate ec (k) is 0.025; As | e (k) |＜δ and | ec (k) | during＜γ; The output that shows each network control system has got into its steady-state value ± 5% (or ± 2%) fluctuation range, and controller node need not to send new packet through shared network to its actuator node again, can save network bandwidth resources; Improve network bandwidth utilance, system was not only stable but also satisfy the static control performance quality requirement; As | e (k) | >=δ or | ec (k) | during >=γ; The output that shows each network control system is in the transient process; Controller node need send new packet to its actuator node through shared network, to strengthen control action, guarantees that system finishes the transient process state as soon as possible.

Images