Reliability Components of the WAMS Information Network

: Power system modes can be controlled by use of a wide-area monitoring and control system (WAMS). It is based on the phasor measurement units (PMUs), connected by an information network covering a significant territory. The functioning reliability of such a network deter-mines to a big extend the correct operation of the entire monitoring system. To assess its reliability, the one can be decomposed conveniently into four components: hardware or technical reliability, traffic reliability, software reliability, resistance to external negative impact on the transmitted information, and to consider each m independently. In this paper, the main attention is paid to assessment of the reliability and availability of the first three components in the information network.

The need for a correct assessment of the power system state has led to the creation of the Wide Area Measurement Systems (WAMS). It is based on the measuring technology of phasors (phase vectors), on the Phasor Measurement Unit (PMU) by the signal of global navigation systems, which ensures the simultaneous measurement of phasors [1]. WAMS includes measuring transformers, PMU, Phasor Data Concentrators (PDCs) and equipment of the local information network. They allow to control the power system behavior by continuous observation over systemic events. The WAMS operation reliability is determined by the element reliability of the monitoring system.
The paper considers the WAMS network structure, and proposes the assessment of their reliability based on the network links. A network failure is determined by the loss of terminal communication, and this concept includes not only the absence of such a connection, but also the distortion of the transmitted information. Then the network reliability includes four components, as fellows: 1) Hardware or technical reliability associated with the failure (destruction) of transmission channel elements or the integrity of communication lines; 2) Traffic reliability, determined by the temporary data loss or distortion without failure of the transmission channel element; 3) Software reliability due to errors in the development of exchange execution programs; and 4) The opposition to the external targeted influence on the transmitted information. This paper considers first three reliability components of the WAMS information network. The paper includes the example of an application for assessing the availability of a 10-bus system network [2] from the positions under consideration. Optical fiber or the high frequency channels through power lines are adopted as the carriers in the example.
The WAMS network hardware is network communications, the PMU and PDC electronics. Since the operation of the central processor units and the PDC communication interface during duplication is similar to the operation of these elements in the PMU, we use the reliability estimate of these units in [3], obtained from the Markov system of probability equations for state, taking into account different lengths of the main and backup communication channels. Then the network link availability from the duplicated information source (PMU, PDC or, if necessary, an intermediate amplifier) and the lines of the communication channel can be defined as since PDCs are the same type, and = Here is the availability of the duplicated information source, and are the failure and recovery rates of the source, respectively. The physical availability of information carriers (twisted pair, optical fiber, high-frequency channel), each element of which is characterized by the length li, specific failure rate the main -and backuplines, average recovery time of the main -and backup -lines per km, should also be included here. Since the second component has small values, we neglect it. Therefore, the communication line availability is inversely proportional to the square of line length. Unlike duplication in electronics, where the backup device usually repeats the main one, duplication of information carrier is provided most often by elements of various reliability in-dicators. This is because in normal mode the communication is provided via the shortest line in the communication network, and in the backup mode case the information goes through the remaining in the communication network, which can be significantly longer than the main one. Moreover, the approach in solving this problem is the same as at duplicating electronic components (2), only taking into account the different values of and for the j th connection (3).
The algorithm for the calculation is as follows. After setting the initial data on the known link lengths of the information channels and the necessary reliability parameters, a table of the link participation in the formation of the main and backup channels is compiled (see the example of Table 1). Next, the link reliability characteristics ( , and for the j th link) are calculated, the same parameters are determined for the main and backup information exchange channels and the availability and characteristics of channels with redundancy are calculated. At the same time, all changes are determined in the network configuration by the source data and the link participation table. The estimated part remains unchanged.
Hardware reliability indicators were obtained for the test power system at the information network diagram in Fig. 1 and are given in [4]. Here we only note that with a complex network of information communications, it can find the backup connection from the server node to the node with the failed link, excluding the last one. To do this, we use the search algorithm first in depth and then in breadth, as proposed in [5]. It allows to take into account the failed links to find a backup route, if one exists, or to warn of its absence. When searching, the "Reserve channel" column of the Table 1 is built and the hardware found path reliability is evaluated. These backup routes are stored in Table 1 in the order of the decreasing reliability. A similar operation is performed in the process of network building. In a real mode, if necessary, a backup channel with operational connections and highest availability is used.
Traffic reliability lies in the information transmission in due time without losses and distortions associated with the loading of the exchange channel. Losses due to traffic are caused by an unacceptable delay or loss of some information due to an overload of the information channel, but are not associated with the element device failure of this channel, which is taken into account in hardware reliability. Therefore, the traffic reliability is determined by the choice of bandwidth, taking into account the delay of transmitted information.
The information frame from the generation unit or power line, formed by each PMU, combines 9 vector measurements: 3 currents and 3 voltages (magnitude and phase), 3 powers (active and reactive components); 2 analog values: generator excitation current and voltage; the PMU state and the state of the switching elements. In addition, the transmission package includes the frequency and speed of its change, the time stamp and the binding for interaction with the information network in standard C.37.118-2011. The structure of the data frame is given in Table 2. l -number of discrete information sources; m -number of analog information sources; n -synchronized vectors (magnitude and phase).
Then the amount of information from one PMU takes bin = 8 * 9 + 2 * 8 + 2 + 2 = 92 bytes. The amount of additional information per frame of one bus is bfr = 6 + 8 + 8 + 2 = 24 bytes. Depending on the PMU number -sources of measurement information and transmitted measurements per second, the packet volume often lies in the range of 100 -400 bytes. The approximate channel bandwidth in Kbit/s is given in Table 3 depending on the number of source devices and the sampling rate taking into account a margin of 10%, [6]. In this case, 1Kbit = 1024 bits.
The delay in information is related both with the type of the exchange channel and with the time of unloading its receive buffer. Packet delivery to a    (5) The signal propagation time, рс, in most communication systems is determined by the propagation time of the optical or electric signal (electromagnetic field). The pulse delay in the optical fiber is (3.5-5) * l (ns) [7], and in the copper wire 5 * l (μs) [8], where l is the channel length in km.
The packet transmission time pt depends on the data transfer rate on the communication line νtr (Kbit/s) and the volume or length of the packet Lp (Kbit) Obviously, the propagation speed depends only on the channel material; therefore, the propagation time on the channel is constant. Transmission time depends only on the packet length.
The main task in designing a data transmission network is to ensure the balance between traffic (request flow λ, in our case, measurement frequency), the amount of network resources (bandwidth) and the quality of service (service flow µ, request processing parameter). In solving this problem, two model levels of interaction for the open systems (OSI) are considered: network and channel layers.
Network layer. At the network layer traffic transit routes in the network are considered. For this, it is convenient to describe the communication network as the graph model [9] (in this case, non-oriented), in which the network nodes (routers) correspond to the graph vertices, and the communication lines to the graph arcs. The transmission time to the receiving node is the time that the packet spends on the network line. This time is random to a certain extent.
The intensity of the load on the network graph arcs ρij, determined by the ratio λi/μj of the flow request intensity from the node-information source i to the flow service intensity by the receiver node j, depends on the number of devices and the amount of information from each device. In our case, the flow request intensity is determined by the measurement frequency in the power system bases: = = 1 . The flow service intensity is the back volume of the packet delivery time: µ = 1 = 1 + ⁄ , and since this time is less than the application period, then wp = 0. Otherwise, information will be lost. On the other hand, the receiver electronics creates an additional delay Tre on average of about 5 μs. = + + ⁄ .
Channel level. At this level, it is required to evaluate the necessary bandwidth of communication lines between network nodes. In the general case, an approximate formula can be used to estimate the probability of losses [ where Nj is the number of sections in the receiver accumulator j; ρij is the load intensity of line ij. The absence of losses is defined as = 1 − .
It is clear that such an assessment corresponds to one information line connecting two nodes. Taking into account the serial connection of channel links for two nodes passing through intermediate nodes, the overall assessment of the probability of information loss is defined as (10) Let us evaluate the WAMS information channels for a 10-bus test power system, which is given in detail in [2]. A map of information links with a distance scale is presented in [4], and a diagram of information links is presented above in Fig. 1. All information communications are made by fiber with a propagation delay of Tpc = 5 ns/km. Electronics Delay Tre = 5ms. Baud rate vtr = 1 Mbps = 1048576 bps [11]. The measurement transmission frequency is 10 Hz or Tmsr = 0.1 s. Since the power system manager is defined in the system bus 4, the information routes in the normal and failure mode of one information transmission line are given in Table 1, and the last column shows the connection of the node-source to node # 4 in the failure event of one component link along the route. Note that a communication failure of 10-2 leads to a complete loss of the information exchange with node 10. The initial data for the calculations are summarized in Table 4. Here in the third Table 4 Input data on the information network The simulation results are given in Tables 5 and 6, from which it is seen that with the calculated loads the probability of information loss is very low.
We consider the dependence of the loss information probability on the load intensity ρ using the example of a 7-4 connection under the remaining conditions. Using the same example, we consider the effect of the number of drive sections N, Table 8. It is clear that with N = 0, the probability of information loss equals 1, since there is simply nowhere to take it. As N increases, q it drops abruptly, turning almost to zero already at N = 10. It is also obvious that the greater is the load intensity ρ, the greater is the probability of information loss q, and the increase is quite fast, requiring an increase in the number of sections of the receiver's drive N.  Failure of software is associated with its inconsistency with the tasks. There are many definitions of a software error. The definition [12] seems to be the most acceptable: Software reliability is probability that the program will run for a certain period without failures, taking into account the degree of their influence on the output results. The frequency of error occurrence from the statistical data, reduced to 100% errors, is given in Table 8 with the position "Incomplete or erroneous task" disclosed in more detail.
Software is not the subject to wear and tear; its reliability is determined only by development errors. Thus, this indicator should increase over time if the correction of the revealed errors does not introduce new errors.
For critical applications, which should include the WAMS software, the system delivered to the customer may contain from 4 to 15 errors per 100,000 lines of program code [13]. For clarity, we note that the number of lines of WINDOWS XP code is more than 45 million, NASA programs are 40 million,

Conclusion
The correct functioning of the WAMS local information network is ensured by four components of reliability.
Hardware reliability of the network is largely determined by the reliability of information carriers (optical fiber, radio waves, etc.) and by the devices providing their work -phasors meters, data concentrators. With the right organization of redundancy, the network hardware availability meets the requirements for the reliability of control systems.
Traffic reliability component is determined by the load intensity of each connection and the capabilities of receiving information, determined by the volume of the receiver's drive. The availability of the test network for traffic exceeds three nines after the decimal point.
In terms of software, the influence of code lines on the value of the reliability parameter is considered and its estimate is shown depending on the number of commands. For the example of a 10 million code lines of WAMS program, the mean time between failures should be 285 years.
Many works have been devoted to opposing the external negative impact on the transmitted information, for example, [15,16], and they are not considered by this paper.