SOLVING A TASK OF COORDINATED CONTROL OVER A SHIP AUTOMATED ELECTRIC POWER SYSTEM UNDER A CHANGING LOAD

Global practices show that the research into sophisticated technical systems and complexes (STS and C) as the control and automation objects ultimately results in the acquiring of semantic information in the form of requirements for the scope of functions and a list of control tasks. It is obvious that the more a priori information is available about objects the fuller and more accurately the properties of the object are reproduced. This is achieved by a certain redundancy of information, the existence of specific data, parameters, etc. Primary and secondary semantic information is used to describe the tasks and functions of control. Obtaining secondary semantic information is considered to be a continuation of the knowledge about an object, which involves the following: a) the identification of the most robust and characteristic features; b) the results of the analytical-synthetic and logical conversion of primary semantic information, which is reproduced by symbols. Basically, the development of such symbol-based systems (thesaurus) is in the plane of designation of descriptors within it, which carry not only a load of meaning but also facilitate the reading of model semantic information. On the other hand, devising methods for effective control over the technological processes on a vessel and operation of marine vehicles of various types is generally limited to a significant set of contradictory and, in some cases, mutually exclusive situations. Research and design of compound STS and C necessitate improving energy transmission processes based on the principles of dynamic control and stabilization of a vessel’s power system while minimizing inevitable losses. Given this, the designations of the main descriptors are supplemented with the symbols of international designation and indexing of SAEPS elements, as well as the technological process, for example, ST (start), SP (stop), STQ (start-quickly), EM (emergency), etc. To extract and generate information about membership, many auxiliary characters are used, related to the root of the descriptor ε=(N, i, j, l...), where N={NBAS} is a set of natural numbering NBAS, in which the descriptor BAS denotes the basis of numbering. For example, =1, m N m is the numbering of How to Cite: Budashko, V., Shevchenko, V. (2021). Solving a task of coordinated control over a ship auto-


Introduction
Global practices show that the research into sophisticated technical systems and complexes (STS and C) as the control and automation objects ultimately results in the acquiring of semantic information in the form of requirements for the scope of functions and a list of control tasks. It is obvious that the more a priori information is available about objects the fuller and more accurately the properties of the object are reproduced. This is achieved by a certain redundancy of information, the existence of specific data, parameters, etc. Primary and secondary semantic information is used to describe the tasks and functions of control. Obtaining secondary semantic information is considered to be a continuation of the knowledge about an object, which involves the following: a) the identification of the most robust and characteristic features; b) the results of the analytical-synthetic and logical conversion of primary semantic information, which is reproduced by symbols. Basically, the development of such symbol-based systems (thesaurus) is in the plane of desig-nation of descriptors within it, which carry not only a load of meaning but also facilitate the reading of model semantic information.
On the other hand, devising methods for effective control over the technological processes on a vessel and operation of marine vehicles of various types is generally limited to a significant set of contradictory and, in some cases, mutually exclusive situations. Research and design of compound STS and C necessitate improving energy transmission processes based on the principles of dynamic control and stabilization of a vessel's power system while minimizing inevitable losses.
Given this, the designations of the main descriptors are supplemented with the symbols of international designation and indexing of SAEPS elements, as well as the technological process, for example, ST Proposed thesaurus are constantly supplemented with descriptors as the tasks are solved, and individual descriptors are entered into the texts of requirements for automated control systems.
Thus, to summarize, the existing techniques to improve the reliability of SAEPS operation require the formation of coordinated control algorithms, which would minimize possible emergency modes when assigning false control combinations.
There are also errors in the sequence of enabling GU related to the incompleteness of the specified combination, an unspecified sequence, the indication of one GU several times in one sequence, and the coincidence of binary sets of different combinations in a sequence. Therefore, solving the tasks of coordinated control over SAEPS under a changing load is a relevant scientific and technical issue for the maritime industry.

Literature review and problem statement
Modern research is advanced towards improving the principles for synthesizing the systems that can effectively control the synchronization processes involving generator units (GUs) as part of compound technical systems and complexes (STS and C) [1]. Based on the application of resultant functions, the stages in solving the tasks of controlling the synchronization of frequency fitting in a hierarchical sequence [2] are determined. The functioning of STS and C control elements is analyzed using integrated optimization criteria and dual control principles [3].
In accordance with the requirements stated in [4], control over the structure of generator units (GUs) under a changing load should be carried out by selecting the PR-SEL reserve and forming commands to start ST(i) or stop SP(i). GU is enabled/disabled when the load reaches the upper PH(l) or lower PD(l) load thresholds for l parallel operating and backup GUs in accordance with the valid order SQ(i)∈SQ(N SQ ). Similarly, PRST start-up programs or PRSP stop programs, PRSY synchronization, PRSH distribution, or PRUNL load transfer are executed [5].
Hence, it follows that the coordinator program, after each technological cycle o T ТС , should determine the number of GUs connected to the main switchboard (MSB). That implies executing a subroutine of SBCNT counting, which generates a predicate WRK(l)=I, if l GUs are in operation, and calculating the required number of GUs according to the load. PRNRY program generates predicates NRY(l)=I, NRY(l+1)=I, NRY(l-1)=I for, respectively, l, l+1, l-1 parallel operating GUs to select a backup GU using the PRSEL program to start PRST or stop PRSP [6]. The upper loading thresholds are determined on the basis of calculating a load P L permissible under the technical conditions per a single GU and the required supply of generated power ν 1 P L , 0<ν 1 <1. The lower thresholds are determined from calculating the power reserve ν 2 P L , based on economic expediency, 0<ν 2 <1. In this case, the generation of predicates PH(l), PD(l) are written as the following rules: where P G (i) is the current load on the i-th GU. The power P G (i), depending on the specifications, would vary within: where K RCL is the power conversion coefficient when changing environmental conditions;  is the load factor; P N is the GU rated power. Expression (2) shows that an increase in the number l of GUs running in parallel leads to an increase in the load factor. Fig. 2 displays a diagram of GU structure control based on the principle of "rigid" loading thresholds at ν 1 =0.2 and ν 2 =0.4 regarding the control over a 4-unit SAEPS. It follows from the analysis of (1), (2) that one can change the loading thresholds of SAEPS by varying the parameters ν 1 , ν 2 , P L , separately or in different combinations. That is, it is possible to implement many ways to control SAEPS based on the principle of "flexible" loading thresholds for units. However, these techniques would not be exhaustive because they imply control over generated power only. If we take into consideration the possibility of load control (by disabling/enabling part of consumers and prohibiting enabling them at a certain peak of loading), as well as provide for the use of electricity storage devices, the scope of control tools could be significantly expanded. In this case, it is necessary to set a task on the time delays in starting a backup GU.
In order to filter short-term load emissions associated with the start or repeatedly short-term work of consumers, the techniques for the time distribution of emissions are used, and, to coordinate the GU overload characteristics -time delays (a method of amplitude-time division).
The random time function p=x(t) most fully describes changes in the power consumption by SAEPS. It is assigned by the mathematical expectation M x (t), variance and a correlation function K х (τ r ). As follows from some experimental data processing [7], the random process of changing the load under the running and stationary modes is characterized by a correlation relation, which decreases sharply as the τ r interval increases. Therefore, already in the region τ r >1, a random process can be considered ergodic, and a change in load is characterized by a random value .
x That is why the energy forecast , S necessary to provide for an additional, higher than x PH , load in the time intervals after t PH , requires that formulae for performing calculations should be derived.
Thus, to summarize, the existing ways to improve the reliability of SAEPS operation do not completely eliminate but only minimize possible emergency modes when assigning false control combinations. In addition, there are errors in the sequence of enabling GU if the specified combination is incomplete, the sequence is not set, one GU is included several times in the same sequence, and when the binary sets of different combinations of the sequence coincide.
The tasks related to SAEPS control based on the principle of "flexible" loading thresholds, considered in [8], necessitate a detailed elucidation of optimization tasks regarding the minimization of fuel consumption. SAEPS that are operated on modern vessels are characterized by a low load factor of installed capacity [9], and, as a result, low efficiency. That is explained not only by the existing principles of SAEPS configuration but also by the imperfect organization of its operating modes. The tasks considered are typically solved by means of approximate procedures, seeking to improve only one criterion -excessive redundancy of energy under the basic modes of vessel operation [10]. When calculating the required power and permissible load values of SAEPS, they consider the average or limit values of unmanaged variables (request for loading, temperature, pressure, humidity, etc.) not taking into consideration the possible range of their change [11]. That leads to a decrease in the technical and economic indicators of STS and C. Pay special attention to the need to account for weather conditions when calculating the upper levels of loading the diesel engines in order to avoid thermal overloads [12]. It is known that the maximum load of vessel diesel engines begins to decrease with an increase in ambient air temperature exceeding 27 °C, humidityexceeding 60 %, and with a decrease in barometric pressure below 101.3 kPa [13]. The influence of these factors in modern systems is taken into consideration selectively while they act interdependently.
The manufacturer "Wartsila-Sulzer" [14] indicates in its diesel engine specifications that the normal parameters of the environment are the atmospheric pressure of 720 mm Hg, the air temperature of 20 °C, and humidity of 70 %. It is relevant to warn that any deviation from these values, especially the temperature of the suction air and pressure, significantly impairs the characteristics of the engine. The most characteristic changes, when sailing a vessel in the tropics, is a decrease in power, an increase in the specific fuel consumption and temperature of the exhaust gases, caused by a decrease in the weight charge of the air and an increase in its initial temperature [15]. That is explained by the fact that a decrease in the weight charge of air ρ k leads to a decrease in the coefficient of excess air α=kη v ρ k g C , , where η v is the coefficient of filling the cylinder, and g C is the cycle fuel supply. In turn, a decrease in α would cause a decrease in the indicative coefficient of efficiency η i and, as a result, in the effective efficiency of the engine η e =f(α). As a result, the torque of the diesel engine would begin to decrease while the rotation frequency controller, in order to stabilize the frequency of the generated current, would begin to increase the fuel supply g C , which could lead to overload in terms of mean indicative pressure. At the same time, the temperature of the exhaust gases would start to increase. Excessive increase in the temperature of the exhaust gases forces to reduce the cycle fuel supply, which leads to a total decrease in the GU generated capacity.
A study into the operating modes of a 6-unit SAEPS on the vessels of type "Marco Polo" reported in [16] showed that typical fluctuations in external conditions were within the following ranges: t, ℃=20÷45; φ=60÷85 %; P a =750÷780 mm Hg. This leads to a change in the maximum permissible, in terms of thermal load, powerfrom 735 to 604 kW (by 18 %). Frequent fluctuations in load and environmental parameters over 24 hours make it difficult for service personnel to solve the task of determining the optimal GU configuration. Therefore, in order to avoid the risks of thermal overload, it is necessary to main-tain an excess supply of generated capacity. For example, under a running mode, instead of the possible four, there are five GUs operated at low and medium loads [17]. That leads to a decrease in mechanical efficiency, and, as a result, a decrease in the effective efficiency η e =η i η m . Analysis of the operating conditions of diesel generators (DG) [18] reveals that the average load of SAEPS is 25-30 % of the total rated power of generators. On ships, certain types of DG, during almost all working hours (90 %), are loaded by less than 50 %. There are cases of failure of butterfly and frame bearings of engines due to long-term operation under small loads [19]. For example, when a single DG is operated with a load of 270 kW instead of two DGs running in parallel with a load of less than 50 % N enom , the daily savings in fuel reach 0.8 tons and 10 liters of circulation oil. Thus, the elimination of these shortcomings makes it possible to obtain a significant technical and economic effect. Resolving this issue is associated, first, with the development of algorithms for estimation and continuous control of permissible loads for diesel engines, and, second, with the development automatic control system (ACS) of such engine which take into consideration changes in environmental parameters.
In this case, the mode of operation of the diesel engine can be adjusted by reducing fuel consumption per value ΔG T by redistributing loads among the GUs running in parallel PRSH. By connecting the drives ONACC, disabling secondary OFCNS for a given mode of the consumer or reducing the temperature of the supercharged air Т S by increasing the cooling water inflow in the cooling system [20]. It should be noted that the α C (i)>α opt zone is also undesirable since an increase in α C leads to a deterioration in the combustion process, an increase in heat losses coming with gases, and a decrease in the indicative efficiency [21]. Some studies, for example [22], have made it possible to synthesize the optimization algorithm PRORT for the diesel engine modes of operation, provided a condition is accepted that the system contains all the necessary sensors {P S , P i , T S , G T , φ, t G }, where φ is air humidity, t G is the exhaust gas temperature. The controlling elements {SMW, SMF} are the servomotors, which would adjust water supply SMW to cool the supercharged air and fuel supply SMF.
A technique chosen to organize the work of programs similar to РRОРТ is an interruption procedure with a sampling period Т СО predetermined by the dynamic properties of the super-air cooling system and diesel engine. To facilitate the procedure of logical programming, the РRОРТ program is represented as a set of РROРТ={SBHTP, SBALF, SBSM, SBCNT} subroutines. Namely, subroutines SB for the control of high temperature SBHTP, calculations of deviations of the coefficient of excess air SBALF, control of servomotors SBSM, and calculations of the duration of enabling a servomotor СNT [23]. In this case, the rules of the system operation are re/presented in the form of instructions: that is, in case of deviations in the total coefficient of excess air from the optimal one, servo motors are enabled in the appropriate direction if it is possible to adjust. If not, when the valve is already fully open, control is carried out influencing the fuel supply if a coordinator approves of this (DCP=1). When the temperature of exhaust gases is exceeded, the process is sent to the SBНТР subroutine. The implementation of SВСNT subroutines is carried out by comparing the result of subtracting the number from the remaining ones that reflects the size of the time interval ΔB(t+1), unit 1(СNT) of the account, which is usually equivalent to the duration of the main stroke Т ТC . In the process of operating a vessel's SAEPS under a static mode, or during the period of change in its structure under the conditions of load optimization, emergencies and pre-emergencies may occur, caused by the run-out of controlled parameters beyond the maximum level. For these cases, the study must provide for the algorithms that would form commands to enable a backup GU and disable a failed GU, taking into consideration the assigned sequence.

The aim and objectives of the study
The aim of this study is to establish operating rules for the algorithm of optimal control over SAEPS when changing the load, GU technical condition, and environmental conditions. This would make it possible to avoid errors in the sequence of enabling GU when the specified combination is incomplete, the sequence is not set, when one GU is included several times in one sequence, and when the binary sets of different combinations of sequence coincide.
To accomplish the aim, the following tasks have been set: -to analyze the load characteristics of the units, decompose control tasks, and build databases that would determine the number of working GUs, their technical condition, load, fuel consumption, and environmental parameters; -to determine the total load on SAEPS, analyze equivalent characteristics of fuel consumption, select the optimal one according to the criterion of the minimum fuel consumption, the GU configuration taking into consideration restrictions on the upper and lower loading thresholds; -to determine the power conversion coefficient k RCL and adjust the upper and lower load thresholds, the optimal load distribution in terms of technical condition and meteorological conditions; -to optimize the operations of starting, synchronizing, transferring a load, and stopping GU, related to the formation of GU configuration under the conditions of optimality F ∑ min(P ∑ ). To improve the РRОРТ programs for optimizing primary GU engines, which implements the criterion(α С -α opt )→min.

Materials and methods for synthesizing basic algorithms for the higher levels of SAEPS control
Determining the current value of the upper loading threshold for SAEPS implies considering the influence of meteorological conditions of a vessel's voyage. We determined the dependence of change in indicative power according to a procedure recommended by the International Congress of Motor Builders that employed K. Zinner's formulae [25]: 293 293 , 99 where N iT , N iN is the indicative power under the current and normal conditions, respectively; D is the barometric pressure, kPa; P P is the partial pressure of water vapor, kPa; Т, Т w is the air temperature at suction and cooling water, respectively, K.
Managing the sequential process of enabling/disabling GU taking into consideration the emergency states of GU and control influences from the operator. We propose a sequence of synthesizing the algorithms for the program that controls the supervisor of the control system coordinator with a distributed two-level hierarchical structure.
Procedures for the transition from one level of generated power to another, taking into consideration the efficiency criteria, are executed taking into consideration the pre-emergency and emergency states of SAEPS and control influences from the operator. The structure and information connections of the control system, taking into consideration the requirements for the examined SAEPS, are strictly related to the principle of concentrating system control functions associated with the properties of adaptation and optimization.
The partial pressure is determined from the i-d diagram of moist air ( Fig. 1) [26] as a function of temperature and humidity of the environment; the indicators of the extent of m, n, q are determined depending on the normal values of the total coefficient α C of excess air [27]: This relationship can be represented in the form of a nomogram ( Fig. 2) [27] where the consumption of humid and dry air is related via the following ratio: A value of d i is found by using wet air tables or an i-d diagram. The nomograms in Fig. 2

demonstrate that when the load of DG is P(i) and the fuel consumption is ,
A T G other operating parameters would be characterized by the set {P S (i), T S (i), d i }, then the total coefficient of excess air would equal α C (i)<α opt less than the optimal value. The consumption of dry air is ΔG BC less than the norm, which would worsen the combustion process and lead to an increase in the temperature of exhaust gases.
Maintaining the optimal operation mode of each DG by adjusting α C helps α C solve the task of minimizing fuel consumption, but, for the current value of the system load, it is difficult to ensure the normal thermal mode of the diesel engine and the maximum value of its efficiency ( Fig. 3) [27]. However, in general, for SAEPS, the objective function of total fuel consumption may not have a global minimum: where F ∑ , F(i) is the fuel consumption per hour of SAEPS operation and the i-th GU; g e∑ , g e (i) is the specific fuel consumption by SAEPS and the i-th GU; P ∑ and P D (i) are the loads of SAEPS and the i-th GU; λ(i)=P D (i)/P ∑ is the load share of At the next stage, in accordance with the GU loading diagram in Fig. 5, it is necessary to consider issues related to the logical module SBCNT, which gives rise to a predicate: To find an extremum of the function g e∑ , a load distribution method is used so that the values θ(i) for each GU at the points of the predefined mode accept equal values, that is, In this case, a sufficient condition for optimal load distribution is the g e∑ →min minimum requirement, which is met at d 2 g e∑ >0. If GUs have the same load characteristics, then, in order to satisfy the required condition, it is necessary to distribute the load evenly at the points of SAEPS operation modes. To meet the condition of sufficiency, the following conditions must be satisfied: this is achievable if all derivatives from ( ) This position is extremely important since it makes it possible to calculate, under optimality conditions, the lower load thresholds of SAEPS (Fig. 6) [27]. In this case, according to Fig. 6, the region of generators' operation, limited by valid boundaries, may increase, then formula (2) is simplified to take the following form: However, during operation, the load characteristics of engines change [28], which can lead to their significant scattering and setting the task of optimizing the operation modes of SAEPS taking into consideration the actual load characteristics [29].
To solve the problem in such a statement, a method of dynamic programming [30] is used, which makes it possible to choose the most likely (quasi-optimal), for a certain region of operation mode, GU configuration in compliance with all restrictions. To perform the dynamic programming algorithm [31], equivalent characteristics must be constructed, which are the dependences of the minimum fuel consumption by assemblies on their total power, that is: Accounting for the mechanical efficiency η M , the power conversion coefficient k RCL is determined from the formula k RCL =(k-0.7(1−k)/η M ), and the permissible power P Di of the i-th DGfrom expression P Di =k RCL •P N [32]. It is obvious that when the air temperature changes from 27 to 62 °C, the k RCL value varies from 0.96 to 0.88, and when the three parameters change (T, D, φ) -from 0.93 to 0.85. In the analysis, it is accepted that the temperature of the cooling water varies within 25÷35 °C [33]. Thus, the law is established for determining k RCL whose application in algorithms (1) renders flexibility in the generation of predicates PH(l), РD(l). This law makes it possible to operate GU in regions as close as possible to the barrier characteristics while ensuring reliability and safety, and thereby improve a SAEPS load factor [34].
There is another way to improve control over SAEPS by maintaining the optimal mode of operation of the main engine if we accept as a criterion for assessing the mode the stability of the value for the total coefficient of excess air [35]. That is, min, where α C is the current value, α opt is the optimal value of the coefficient for a given mode of the diesel engine operation, at which the maximum possible power is achieved at the minimum fuel consumption and the permissible temperature of exhaust gases [36]. In the nearest approximation, the total coefficient of excess air can be determined [37] from the readings of measuring devices, using the following expression: where A 1 1 is the coefficient that takes into consideration the structural features of the diesel engine at the constant product of the filling and blowing coefficients; P S and T S are the air pressure and air temperature in the turbocharger; G T is the fuel consumption per unit of time; d(i) is the air humidity content under the i-th static mode. At this stage, there is a need to generate primary predicates P N (N l ) and P D (N l ), which identify that the load has reached the upper and lower loading thresholds for GU.
In this case, emissions should not be taken into consideration if their duration is shorter than the specified time interval t PN (min)±∆τ PN , where ∆τ PN is the time interval that corrects the instability of emission duration (Fig. 8).
In the case when t PN (min) is over and the ejection value P N (l)=1, then the value for the second delay time interval is selected depending on the established amount of power ∆P N , in accordance with the load characteristics of GU, for example, as follows or, as shown in Fig. 7, to count the delays t PN (min) and t PN ( j), one should predict the subroutines of the timer PR1TM and PRTM, giving rise to their corresponding predicates T1PN and TPN according to the following rule: Under the conditions of uncertainty in a change in the load, it becomes necessary to prohibit the change towards a decrease (from a certain valid number) in GUs running in parallel even if the predicate PD(l) is generated when assessing the load. This is achieved by assigning the minimum permissible (l min ) quantity of GUs. We shall link by the mutually unambiguous relationship the sets of these tasks with the set of predicates MIN(l), and such that, if MIN(l)=I, then the minimum permissible number of GUs would be l min . Thus, we have defined the rules to form a group of basic predicates WRK(l), PH(l), PD(l), TPH, TPD, MIN(l), NOTACC, NOTCNS. This allows us to write the resultant functions RFU NRY that give rise to these predicates: The next task arising in the study of the rules for operating the converter WRK(l)→NRY(l±1)→WRK(l±1) is to determine a procedure for setting the order of enabling/ disabling GU.
We believe that the system has a setter using can assign any sequence (order) SQ(f)∈SQ(N SQ ), taken from an ordered, numbered set SQ(N SQ ) of all possible sequences. Then, if there are m |GU|=m generators installed in SAEPS, the set of sequences would be |SQ(N SQ )=m!|, that is, with the growth of m, the order increases in factorial. Modern SAEPS do not seek to obtain the entire dimensionality m! as this is an excessive complication of synthesis. In this regard, when selecting elements of a subset SQ(f)∈SQ(N SQ ), it is advisable to be guided by ensuring that each GU is able to be set in any turn.
The synthesis of the sequence assignment algorithm (20) is much simpler than the synthesis of a full-size algorithm since it is possible to choose the current SQ(i) sequence among m sequences.
We defined a set of the critical and non-critical controlled accidents for each GU and in SAEPS in general in the following way: where X CR (i) is the set of controlled and numbered N CR accidents for each GU represented in the form of signals at the logical level {0, 1} from measuring transducers acquired from sensors i; , .
The tasks of determining the optimal information model and information encoding techniques, the development of individual frames belong to the class of ergatic tasks and tasks of ergonomics. Therefore, here we move on to the further study of the converter. For the recording of the resultant functions ∪ ST , ∪ SP , which were defined by the family of predicates ∪ ST , ∪ SP , the relationships between the elements of the set of input signals and internal states are as follows, The model of this converter largely depends on the structure and properties of the setter of the order for enabling/disabling GU. Consider some research results on determining the structure and properties of the order setter (as regards a 4-unit SAEPS). In this case, the converter takes the following form Converter (23), considering (24) for ST (1) and SP(1), is described as follows

3) WRK(3)&NRY(2); 4) WRK(4)&NRY(3).
As the strongest prerequisite to form SP(1), in addition to cases 2-4, we shall attribute the GU(1) state, the last in line, or which is an accident, if other GUs are in a non-accident state. If not, then SAEPS has other GUs in a state of emergency, SP(1) should not be formed, at least without the intervention of the operator. With these conditions in mind, one can describe as follows: where SPQA(1) is the command to stop a DG immediately, due to the critical accident СR (1). However, such models with homogeneous Boolean functions are more suitable for building devices with hard logic and are not quite effective in the development of programmed control systems. This is explained by the fact that with the growth of the established number of GUs in SAEPS, the dimensionality of programs increases significantly, and the tasks associated with the automation of ACS of the STS and C of a given class are significantly complicated. We shall solve the specified problem by searching for more compact Boolean functions, the main of which are the identifiers of data sets. Thus, the generalized functions STA(i) and SPA(i) at some values of indexes m, l, i, and others, demonstrate undefined identifiers that require the introduction of additional conditions and transitions when programming. This is the task of simplifying the operation of algorithms (26) and reducing their volume.

1. Semantic decomposition of control tasks
The formulated tasks in accordance with the analysis of the requirements by international classification societies [38] regarding the structure, functions, and components of SAEPS and the purpose of improving control with information support have contributed to the decomposition of control tasks.
According to the specified dependences (6), (8)-(10), and the characteristic diagrams ( Fig. 1-4), the method of decomposing the tasks of control over SAEPS, GU, and shaft generator unit (SGU) was used to determine the operational characteristics; the results are given in Tables 1-3. The resultant decomposition is represented in the form of a generalized converter structure to control the configuration of SAEPS, GU, and SGU, which corresponds to the formulated control tasks in Tables 1-3 (Fig. 6). Table 1 Decomposition of SAEPS control tasks

Control function
Tasks solved during control SAEPS control under normal, emergency, and pre-emergency modes Coordination of the levels of generated power and the power required at present Transition from one level of generated power to another, taking into consideration efficiency criteria Control over the configuration of GU with the accounting of pre-emergency and emergency states of SAEPS and controlling influences from the operator Organization of the sequence of enabling/disabling GU, taking into consideration the emergency states of GU and controlling influences from the operator Table 2 Decomposition of GU control tasks Control function Tasks solved during control

Hot reserve control
Processing signals from the switch that sets the type of control (automatic/manual) over lubrication, lubricant heating valve, lubricant pressure sensor. Formation of lubrication cycles with adjustable (variable by an operator) time intervals. Formation and control of signals for enabling/disabling a lubrication pump. Registration of emergency signals "No pumping", "No warming up"

Start process control
Processing of signals coming from the local coordinator, buttons "Start", "Emergency start", "Stop", "Emergency stop", from sensors of position of a rail, oil pressure, oil temperature, slow turn, speed sensor, block contact of generator machine. Formation (changeable by the operator) of time intervals of control: successful starting pumping of oil; slow turning; enabling an air valve; shutdown (pauses at repeated attempts of start) of the valve of starting air; enabling the servomotor of a controller; disabling blocking on oil pressure; emergency start; confirmation; successful excitation. Formation and control of execution of enabling/disabling signals: the pump of oil pumping; the servomotor of a controller; a working stop device; a slow turning device; a starting valve; unlocking of the generator machine. Formation of signals of process of start and ALARM: "Start", "Emergency start", "Fault", "No start", "No pumping", "No revs", "Ready to accept loading", "Warms up", "Synchronized", "Load enabled", "Accident" The generalized structure of the converter to control the configuration of GU, which corresponds to the formu-lated control tasks in Tables 1-3, takes the form shown in Fig. 8. Table 2 Stop process control

Continuation of
Processing of signals coming from the local coordinator, buttons "Stop", "Emergency stop", from speed sensors, the position of the rail of fuel pumps, a generator automatic switch. Formation of variable (by operator) time intervals of control: enabling a working stop device, enabling an emergency stop mode.
Formation and control of execution of signals on the shutdown of the generator automatic machine, enabling a working stop device, enabling/disabling a servomotor, closing the valves that cool seawater and fresh water. Formation of signals for a stop process and ALARM: "Stop", "Emergency stop", "Fault", "No stop", "Hot reserve enabled", "Diesel fuel", "Accident". Formation of the warning alarm system from gauges: Oil pressure; Water pressure; Oil temperature; Water temperature; Exhaust gas temperature; Overload; Rupture of fuel pipes: Presence of shavings in oil; Water in the collector; Starting air pressure

Protection and blocking control
Processing signals from the "Protection disabled" switch and the "Reset" button. Organization of time delays and formation of protection signals: -due to excessive speeds of the diesel engine without delay; -due to loss of oil pressure with a delay of 0÷15 s; -due to loss of circulation of cooling water with a delay of 0÷15 s; -due to an increase in the temperature of cooling water with a delay of 0÷30 s; -due to the reverse power with an adjustable power setting of 0÷15 % of the rated power Р N and the actuation time of 0÷15 s; -due to full-current overload with adjustable current setting 0÷1.5 from I N and a time delay corresponding to time-current characteristic; -due to phase breakage and improper alternation of phases; -due to voltage deviation with adjustable deviation setting (from 0.01 to 70 %) and a time delay from 0.01 to 15 s; -due to frequency deviation up to 10 % of the rated value and a delay of up to 15 s. Formation of a generalized signal "Accident" and a signal deciphering the type of accident.
Blocking the start-up when protection is triggered  Table 3 Decomposition of SGU control tasks

Control function
Tasks solved during control Control over the processes of enabling/disabling SGU Processing signals coming from the local coordinator, from the sensors of the clutch position, the speed of GD, the block-contact of the generator machine, buttons "Start" ("Clutch of SGU coupling"), "Stop" ("Disconnection of SGU coupling"). Control of the processes of clutching and disconnecting an SGU coupling. Formation and control of signals for enabling/disabling clutch valves and disconnecting an SGU coupling. Formation of signals of the process for enabling/disabling ALARM "Start", "Stop", "No stop", "Coupling in gear", "Coupling disconnected", "Ready to accept the load", "Synchronization", "Load enabled", "Malfunction", "Accident", "Pressure of working oil of SGU coupling", "Temperature of SGU windings"

2. Iterative adjustments of the capacity redistribution coefficient in accordance with the upper and lower load thresholds
In accordance with the loading diagram of GU (Fig. 5), consider the issues related to the logical module SBCNT, generating a predicate (11). As input information, we accept signals SWG(i)∈SWG(N GU ), coming to the controller from state signalers (on/off) of generator circuit breakers SWG(N GU ), |N GU |=m, forming at the input of the input device in each technological cycle Т TZ , m is the bit binary word where l j k is the conjunction, a constituent unit that includes l non-inverse and (m-l) inverse variables S SWG (i); ∨ is the disjunction that combines all possible conjunctions of full length.
The result is the correspondence in tabular form (Table 4) for a 4-unit (m=4) SAEPS [40]. Table 4 Combinations of GUs in a four-unit SAEPS

WRK(l) WRK(0) WRK(1) WRK(2) WRK(3) WRK(4)
In this case, the technique to form an algorithm of functioning of the SBCNT module implies the arithmetic addition of the values of input variables with the result selected as a valid predicate according to the following rule: The input word is used as the address of a valid predicate, and unambiguous groups of address words would match non-extreme predicates. For example, the predicate WRK(1) is generated by the following rule: The next stage of the study is associated with the PRN-RY module, which generates predicates NRY(l), NRY(l+1), NRY(l-1) for the individually operated, or running in parallel, GUs (depending on the specified load).

3. Analysis of equivalent characteristics and selection of optimal GU configuration according to the criterion of minimum fuel consumption
The generation of primary predicates P N (N l ) and P D (N l ), identifying that the load has achieved the upper and lower GU loading thresholds, should not take into consideration emissions if their duration is less than the specified time interval t PN (min)±∆τ PN . ∆τ PN is the time interval that corrects the instability of emission duration (Fig. 7).
In the case when, after t PN (min) is over, the emission value P N (l)=1, then the value for the second delay time interval is selected depending on the amount of the established power ∆P N , in accordance with the load characteristics of GU (17). Or, as shown in Fig. 7, to count the delays t PN (min) and t PN ( j), one should provide for the subroutines of the timer PR1TM and PRTM, generating their corresponding predicates T1PN and TPN according to rule (18).
We would like to emphasize that this approach to determining the time delay is fair for the established emission [41] when the probability of further increase in load is low over time t PH . At the same time, as practice shows [42], at some levels of SAEPS loading there are significant fluctuations in the standard deviation σ x relative to its mean values. Failure to consider these fluctuations can lead to engine overload during the delay, which is a significant drawback [43]. The disadvantage is eliminated by comparing the values of the produced, up to the time r CNT , energy W T with the permissible energy W D along the section t PN =B PN •∆t PN , obtained by the results of measuring the power in a zero cycle ∆P N (0) (Fig. 9). That is, we can conclude that the delay time would be over at a time point when: In this case, the delay time control procedure is reduced to the following rule: The CTM procedure based on ∆P PN (0) defines t PN , calculates B PN , and counts the timer program. After the time t PТ is over, if ∆P PТ >0, it is possible (in order to preserve the current GU configuration according to the criterion of technical and economic feasibility) to provide measures for their unloading. For example, by disabling secondary consumers (CNS) or enabling a power storage device. Naturally, the break in the power supply to secondary consumers, as well as the time of the return of electricity by storage devices, are limited. And, if, after this time is over, the deficit of generated power is unchanged, then the use of these techniques is impossible [44]. Therefore, there is a need to assess the load forecast of SAEPS after t PN .
We provide a special program to perform the procedure of calculating C S and evaluating the predicted energy ACC S S < by comparing it with the possible available energy of the storage device S ACC , as well as generating a SWACC predicate that algorithmizes enabling the storage devices: &  then  else if;   while  do procedure  :  ,  , , where NOTACC is the predicate, which indicates that the energy storage device is not able to solve the task of SAEPS unloading [45].
The structure of the PRCNS program, which forms the predicate OFCNS to disable secondary consumers, is similar to the PRACC program.

4. Improvement of operational operations under optimal conditions
Solving the task of synthesizing the sequence algorithm (20) is a continuation of the study of the properties of the converter (1), (2), its subroutine PRSEL (Fig. 8).
The PRSEL program structure is linked to the two resultant functions U ST and U SP , the weakest post-conditions of which would be the predicates ST(i) and SР(i). The prerequisites for these functions are the result of the work of the PRCNT and PRNRY programs, as well as a certain set of predicates from the database that characterizes the technical condition of GU. Set the following sets' predication: Thus, the abstract model of the converter, which forms, for each GА(i), a predicate on a generalized accident, takes the following form: where TEM is the predicate on ending the time delay of a generalized accident signal. The accepted designation of the elementary emergency signal x(i, j)∈X(N GU , N EM ) (21) is more expressive compared to x EM (f), since the symbol x(i, j) carries the information about a membership to a certain GU -GU(i)∈GU(N GU ), i∈N-GU , as well as a membership to a completely certain accident, since j∈(N CR ˅N NCR ). Thus, the predicates ЕМ, СR, NСR are expressed in the form of the following correspondences: where EM(N GU , N EM ), unlike ЕМ(i), is a generalized accident signal for the entire SAEPS. We shall demonstrate the transformation and translation from a semantic language into natural for the basic emergency signals of the i-th DG (22) Building such a model with homogeneous Boolean functions suitable to a greater extent for the construction of devices with hard logic is not quite effective in the development of programmable control systems. This is explained by the fact that with an increase in the established number of GUs in SAEPS, the dimensionality of programs grows significantly while the tasks associated with the automation of the ACS of STS and C of a given class are significantly complicated. Solving the specified task by searching for more compact To eliminate these shortcomings, an algorithmization technique was used, based on the use of a special extended data array ( Table 5).
The use of this data array makes it possible to record the functioning algorithm in the following form, The data of the array of the converter (25) (2). In this case, the QU converter must, through the shift operation MVE(N GA ), put GU Number 2 first in the queue, and the emergency converter -the third, that is, the last one among those operated (Table 6). Table 6 An example of setting a sequence of GUs for a five-unit SAEPS where MVE(N GU ) is the predicate that is valid at the end of N GU number shift operations. Based on the properties of the new QU converter, the predicates STA(i), SPA(i) are described in a general form as follows Predicates SТА(i), SРА(i) are the prerequisites for the system to execute the programs to start PRST, stop PRSP, and synchronize PRSY.

Discussion of results of synthesizing basic algorithms for the higher levels of SAEPS control
After analyzing the load characteristics of GU and completing the decomposition of control tasks, we have defined control functions and those tasks that are solved in the process of control (Tables 1-3). Based on these results, a generalized structure of the converter to control the configuration of GU was developed, which corresponds to the formulated control tasks (Fig. 8).
In accordance with a GU loading diagram (Fig. 5), a structure of the logical module SBCNT was determined, which generates a predicate (11). On this basis, depending on the input signals SWG(i)∈SWG(N GU ), entering the controller from the state detectors of circuit breakers, we obtained the possible combinations of values for the variables defined by expression 2 m= |N S |. In addition, according to the combinations of GUs given in Table 4, we suggested a procedure to form the algorithms of functioning of the SBCNT module based on a valid predicate and a rule (20). The technique makes it possible to use the input word as the address of a valid predicate corresponding to non-extreme predicates of unambiguous groups of address words (21).
To determine the power conversion coefficient k RCL , adjust the upper and lower load thresholds and optimal load distribution, according to the description of state predicates (20), the properties and relations among the predicates on the emergency states of GU have been established. Namely, based on a case where EM(N GU )≠Ø, it was found that at least one of the sets CR(N GU ) or NCR(N GU ) is non-empty. In addition, the following, mutually unambiguous, correspondences where TEM is a predicate about the end of the time delay in a signal of a generalized accident. However, it was found that the sets of critical and non-critical controlled accidents for each GU and in SAEPS in general are predetermined by the sets of measuring transducers where X CR (i) is the set of controlled accidents N CR numbered for each GU. Therefore, the display of signals at the logical level {0, 1} from the measuring transducers acquired from sensors i is possible only if the following sets are defined,  x EM (f) since the symbol x(i, j) carries information about a membership to a completely certain GU. Therefore, the rules of the set GU(i)∈GU(N GU ), i∈N GU of a membership to a completely certain accident would be fulfilled under the condition j ∈ (N CR ˅N NCR . Thus, the ЕМ, СR, NСR predicates can be represented in the form of the following correspondences We have optimized the operations of starting, synchronizing, transferring a load, and stopping GU under optimality conditions F ∑ min(P ∑ ) by improving the program РRОРТ that controls the primary motors of GU, which implements the criterion (α С -α opt )→min. It should be noted that in the data array (Table 5) most of the arguments at numerical indexes 1 and (m+1), 2 і (m+2) coincide. However, the task of minimizing the functions (25) and (26) is not exhausted. The most effective way to reduce functions is to design a rational converter ( ) ( ) The reduction is possible through the development of a self-adjustable converter, the output sequence SQ(i), by setting a part of GU to manual MNL and remote control and reconfiguring the order specified by the operator depending on the pre-emergency and emergency situations EM(N GU ).

1.
Using an example of the analysis of load characteristics of a five-unit automated SAEPS, the levels of generated power and the power required at a certain point in time have been established. It is proved that the procedure of transition from one level of generated power to another, taking into consideration the efficiency criteria, takes place considering the pre-emergency and emergency states of SAEPS and controlling influences by the operator. For example: a pause from 30 to 60 minutes and an operation of 1...10 minutes. The program enables automatic reset of the cycle pause when changing the configuration of operating GUs, as well as when disrupting the specified law of load distribution. The duration of the processes to start TM(ST) and synchronize TM(SY) is controlled and can be configured from 10 to 180 s. In case of the unsuccessful synchronization of DG EM(i,SY), the latter is stopped as a failed GU. A warning emergency signal is formed at a failed SGU synchronization.
2. We have proposed a sequence of the synthesis of algorithms in the control program for the supervisor of the control system coordinator with a distributed two-level hierarchical structure. The task of coordinated control over SAEPS with changes in load has been solved. A generalized structure of the converter to control the configuration of SAEPS GUs has been given; the principles for constructing control procedures based on the principle of "rigid and flexible" thresholds have been described. Taking into consideration the time delay adjustment diagram for enabling GU dependent on the demanded power, a technique to improve the reliability of SAEPS operation was proposed, by eliminating possible emergency modes when erroneous control combinations are assigned. The load transfer process is controlled by the predefined value SWGOP, for example, no more than (0.1 P H ), and time SWGOT, for example, from (1 to 180 s). It is possible to set the time delay for an emergency stop ESTD from 1 to 10 s according to the limit values of parameters such as the pressure and temperature of lubricant and cooling water, exhaust gas parameters, return power, low/high voltage, and frequency.
3. ACS of SAEPS has been proposed subject to minimum fuel consumption and taking into consideration meteorological navigating conditions. The coefficient of recalculation of the dependence of change in the indicative power k RCL accounting for mechanical efficiency has been determined. We have synthesized algorithms for optimizing the diesel engine operation modes by using the necessary sensors (air humidity, exhaust gas temperature, signals from the controlling elements of water supply control for cooling supercharged air, fuel supply, etc.). The sensors of analog signals determine the following: GU phase currents AMP (N m  c opt α − α → When synthesizing control over a five-unit SAEPS, a procedure of algorithmization has been proposed, based on the use of an extended data array, which makes it possible to simplify the algorithm of functioning in the operations of selecting the configuration for a five-unit SAEPS. As a result, time delays were optimized taking into consideration protective signals: from ultra-high diesel engine speed without a delay; from oil pressure loss with a delay of 0÷15 s; from loss of circulation of cooling water with a delay of 0÷15 s; from an increase in the temperature of cooling water with a delay of 0÷30 s; from the reverse power with an adjustable power setting of 0÷15 % of the rated power Р Н and the actuation time of 0÷15 s; from full-current overload with adjustable current setting 0÷1.5 from I H and a time duration corresponding to a time-current characteristic; from phase breakage and improper alternation of phases; from voltage deviation with adjustable deviation setting (from 0.01 to 70 %) and a time delay from 0.01 to 15 s; from frequency variance up to 10 % of the rated value and a delay of up to 15 s. Ministry of Education and Science of Ukraine, No. 1223, dated November 9, 2018 "On approval of expert evaluations of projects of scientific fundamental and applied research and scientific-technical (experimental) advancements".