Some features of the systems for monitoring and diagnostic hydro units technical condition with considering smart grid technology

Recently, hydro units are used in maneuverable modes, which leads to an increase in the load on all elements of hydro units. It can lead to unpredictable failure of their components and even failures in the operation of the hydroelectric power plant. Therefore, the use of diagnostic systems is appropriate for solving such problems. The most promising non-destructive diagnostic systems are vibration diagnosis systems. However, the effectiveness of their use depends on the diagnosis algorithms, which depend on the features and functioning of the diagnosed unit of the power unit. In this work, some models of vibration signals of hydraulic units and features of the construction of vibration diagnostics systems of hydrogenerator units are considered


Introduction
Using the machine in maneuverable modes, which modes can lead to premature or unpredictable failure of their components and even to failures in the operation of the hydroelectric power plant.The ISO 19283:2020 standard [1] provides a comprehensive standardized description of test procedures that allow for a comprehensive approach to the diagnosis of the condition of hydraulic units based on the monitoring and diagnosis of the structural elements of the of the structural elements of the hydroelectric power plant.The requirements for automated monitoring and diagnostic systems are defined in the ISO 61850-7-410 standard [2] and other standards [3][4][5][6][7].In addition, when building a system for monitoring and diagnosing hydro generators, it is necessary to take into account that a significant number of hydro generators are low-speed machines with a low frequency of rotation of the hydro generator up to 150 rpm.Such features of the machine should be taken into account when choosing a diagnostic method.A significant part of the problems related to control and diagnostics of power equipment can be considered as problems of pattern recognition in the diagnostic space.However, on the other side of the problems that arise during an operation and repair of power equipment, the methodologies used during the implementation of smart control and diagnostic systems should be divided into two separate groups: • detection of a change in the technical condition caused by defects at an early stage of their development, when damages have not yet appeared in a clear form; • search for defects based on their already identified signs.
Based on the assessment and with the knowledge of probable processes, it is possible to carry out remaining useful life prediction of the equipment and plan possible maintenance actions based on the actual technical condition of the equipment.The application of forecasts in practice allows flexible changes to be made to plans for reorganizing the use of generating power plant equipment to cover demand both on the electricity market and to meet one's own needs.

The structure of power equipment maintenance system
The main prerequisite for forecasting the remaining useful life prediction of hydro unit is the availability of information about the technical condition of the equipment or its components [8,9].To obtain primary information, modern monitoring and diagnostic systems are equipped with smart sensors [10,11].Smart sensors provide the possibility of self-monitoring, which ensures obtaining data without abnormal values, which can be used without preprocessing to determine the state of the equipment or its units.
The obtained data are intended for use as input data for analysis systems of means of diagnosing the technical state of the component.The technical condition of the component depends on the actual conditions in which the generator operates.Actual conditions are determined with the help of additional sensors that are part of Smart sensors.As a rule, the assessment is performed by comparing the measured or calculated state with the permissible values under which the correct operation of the equipment occurs.
This significantly allows you to expand the possibilities of existing methods to ensure the solution of the previously mentioned maintenance tasks according to failure prediction.
The structure of the maintenance system according to the actual technical condition of the equipment is represented on figure 1 [6].

Physical and mathematical models
For the solution of problem of the remaining useful life prediction of hydro units, it is necessary to build models based on individual quantitative data describing the state of a specific unit of a specific machine, taking into account the history of its use.If there are no physical diagnostic models, the analytical or computer modeling methods are used, which based on the appropriate algorithms [12][13][14][15][16][17].
Therefore, to elaborate algorithms for the functioning of monitoring and diagnostics systems, it is advisable to use the processes of autoregression (AR process), which have become widely used during the implementation of mathematical models of various types information signals and allow to elaborate algorithms for diagnosing technical objects that have a low frequency of shaft revolutions [18,19].A theory of the same processes has been considered in the papers [20][21][22][23][24][25][26][27][28].
By definition, a stationary linear autoregressive (AR) process is defined as where {a j , a j = 0, j = 1, p} are autoregressive parameters; p is the order of autoregressive; {ζ t , t ∈ Z} is a stationary random process with infinitely divisible distribution, it has independent values, P {ζ 0 = 0} = 1.This process is often called as the generating process.It is assumed that the process is stationary in the narrow sense and ergodic theorems are fulfilled [19].Linear stationary autoregressive processes can be given by where φ AR (τ ) is a kernel of the linear AR process [19].
It is assumed that Kernel φ AR (τ ) is connected with parameters of autoregression [19]: where: φ AR (τ ) -is a kernel of linear stationary autoregressive process.
The process ξ t is assumed to be strictly stationary random process, and ergodic theorems are carried out [13].The process ξ t has a Kolmogorov representation one-dimensional form logarithm of characteristic function (CF): where parameter m ξ and spectral functions of jumps K ξ (x) define unequivocally the characteristic function of the random process.
It is necessary to remark that special installations are constructed for experimental research [29][30][31][32][33][34][35][36].A special experimental stand was created at the Institute of Electrodynamics of the National Academy of Sciences of Ukraine.It was used for carry out practical research on vibration analysis.Figure 2 shows a general view of the experimental stand.It was used for research of the influence of the operation of bearing assemblies of electric machines (EM) in different speed modes of its operation on the components of vibration processes that accompany the operation of the machine.The main goal of the research is to experimentally identify the connections between typical defects such as misalignment, eccentricity, lubricant quality, damage to the outer or inner ring of the bearing as a result of metal pitting (pitting) and the parameters of diagnostic signs that clearly allow the defect to be detected.Experimental setup consists of three parts: an electric drive, a massive shaft, a mounting unit and vibration measurement instruments of the tested bearing.Rotation of the tested bearing, installed in the attachment and vibration measurement unit, is provided by a direct current electric motor of the P-51 type through a massive shaft.Motor 11 kW ensures rotation of the tested bearing at speed is in the range 10 to 1000 rpm.Usage of a special coupling with rubber fingers makes it possible to reduce the vibrations caused by the operation of the electric motor.The reduction of the vibrations of the shaft of the experimental installation is facilitated by usage of sliding bearings.The bearings are made of PTFE.Three units of the experimental installation were placed on a massive plate.The main purpose of the bearing assembly and vibration measurement is the ability to artificially reproduce the main defects of the bearing and the placement of primary vibration-transducing equipment (accelerometers).To measure the vibration accelerations of the bearing under study, an accelerometer type ABC-017 was used, which made it possible to measure the acceleration of vibrations of the bearing in the frequency band of 20 Hz . . .30 kHz.It is installed in the radial direction in relation to the tested bearing.Research was carried out on rolling bearings of type 309 ESH2 with various types of defects (skew, lack of lubrication, defects of inner and outer rings, rolling bodies caused by metal spalling due to its fatigue).When carrying out experiments on the influence of the angular velocity of EM shaft rotation on quantitative assessments of the main diagnostic features, the speed mode of shaft rotation varied directly with the values v rot ∈ (250, 500, 750, 1000) rpm.Autoregressive parameters were used as diagnostic features.A bearing misalignment type defect was modeled.Educational groups were formed in the diagnostic space a 1 , a 2 .Two-dimensional histograms corresponding to the selected test conditions are presented in figure 3 and figure 4.   The learning algorithm, first, an evaluation of diagnostic signs is carried out.Then, on the basis of these populations, a vector of estimates of their mathematical expectations, an estimate of variances, and the construction of covariance matrices of correlations between diagnostic parameters and an assessment of the characteristics of the distribution of diagnostic signs were built.If necessary, linear transformations of covariance matrices and normalization of covariance matrices are performed to construct effective decision rules [19].
At the stage of planning the experiment (figure 6), the threshold C and the required number of observations N are calculated.
In the diagnostic mode, diagnostic signs assessed and a decision made on the technical condition of the diagnostic object based on the constructed solving rules (figure 7).The proposed forms of presentation of educational aggregates in the relevant training blocks of the EM diagnostic system allow organizing the functioning of such systems using the Smart concept.

Conclusions
Solving the problems of diagnosing the technical condition connected with the solutions of the problems, the identification of the processes of genesis of defects and the search for defects of nodes.For the evaluation of diagnostic features, the use of autoregressive processes is proposed.According to the proposed mathematical models, experimental results and with taking into account the standards ISO 19283:2020 and ISO 61850-7-410 algorithms of vibration system for monitoring and diagnosing hydro generators has been developed which enables the functioning of the diagnostic system of the hydro power station by the Smart technology.The possibilities of application showed on the experimental stand.Application results and algorithms used in the system are given.The use of the proposed structures and algorithms make it possible to organize the operation of such systems using the concept of Smart systems.

Figure 1 .
Figure 1.The principle functional structure of fault diagnosis system with predictive maintenance.

Figure 2 .
Figure 2. Installation for vibration tests of bearings.

Figure 3 .
Figure 3. Histogram of diagnostic parameters of the serviceable bearing.

Figure 4 .
Figure 4. Histogram of diagnostic parameters the misaligned bearing.

4. Algorithms of Figure 5
illustrates algorithm in training mode.