PNRPU Bulletin. Electrotechnics, Informational Technologies, Control Systems

The development of intrusion detection systems (IDS) is depended on the development both adaptive and non-adaptive algorithms today. On the one hand, the usage of non-adaptive algorithms provides the required speed of operation with the low rates false positive and false negative errors for the IDS. On the other hand, the ability of IDS to detect unknown threats is provided by the availability of the adaptive components. The intensive development of adaptive algorithms is provided by creating new methods associated with artificial immune systems, artificial neural networks, genetic algorithms, etc. The quality of new developed methods and algorithms are tested through to specialized datasets. The datasets are sets of queries which represented in a special form that are transmitted between systems in different levels of the OSI. There are number of such datasets such as: KDD99, NSL-KDD DataSet, ADFA Intrusion Detection Datasets, MACCDC, ISTS, ITOC, DEFCON CTF, etc. Each dataset contains training and test (work) subsets. The training set is need for the learning the adaptive algorithm. The work subset is need for the testing the quality of developed methods and algorithms. The datasets which described by the queries have full structure in terms of their entities. That’s why the specialists in security science are faced with the problem of optimization the attribute set in datasets. This optimization allows to increase the efficiency of algorithm learning and to decrease the number of its linear combination. The article considers the possibility of using the correspondence analysis method with the embedded mechanism of singular decomposition of the matrix representation of datasets to reduce the number of linear combinations its attributes.

The method of parametric identification of indeterminate dynamic objects is described, based on the piecewise-exponential approximation of self-oscillations in systems of two-position control. The structure of the model of indeterminate dynamic objects functioning in systems of on-off regulation is represented by a differential equation with variable coefficients, which consider dynamic parameters and perturbation of objects. The general scheme of parametric identification provides for splitting the experimental self-oscillating curves into separate sections, the description of self-oscillations in the regions using the method of piecewise-exponential approximation, the subsequent determination based on the obtained descriptions of the unknown coefficients of the model of uncertain dynamic objects. As an example of an indefinite dynamic object, the heating zone of an industrial apparatus for vacuum separation of sponge titanium is considered. The selection of informative sections from the experimental curves of self-oscillating regimes is based on the assumption that during the period of self-oscillations the dynamic parameters and perturbation of an undetermined object remain constant. In this case, the period of self-oscillations is proposed to be used as an identification interval, on which the real undetermined object is considered quasi-stationary, in the process of functioning at different identification intervals, as a family of quasi-stationary control objects. This approach allows us to describe the dynamics of an undetermined object on the intervals of identification by differential equations with constant coefficients. A system of finite equations is obtained that approximates self-oscillations on the intervals of identification by exponential functions. As a result of the solution of the system unknown coefficients of the differential equation are determined, and a model is obtained that is adequate to the real undefined object. The model is used to develop adaptive and robust control systems for the production processes of sponge titanium.

The paper investigates the performance of control systems taking into consideration the characteristics of these systems. In order for the efficient performance to be achieved, statistical data are used to build empirical regression models which are further expanded to control systems; as a result, control systems exhibit higher performance parameters than those with regulatory control. In the study, we use statistical data and machine learning methods which were applied for building regression models. Control models are built on the retrospective data being used for both simulating changes in room temperature depending on the temperature outside and the amount of heat obtained from heat supply, and for simulating changes in the temperature of heat carrier depending on the changes in room temperature and the temperature outside. Method lasso produced most accurate values. The models which were obtained with help of this method were used for building a control system and estimating system reaction. These models refer to the class of dynamic predictive models. The paper suggests how these models can be applied to control heat supply in high rise office buildings by taking into account the inertness of heat supply. The calculations evidenced higher efficiency of suggested solutions compared to the application of PID control systems on retrospective data and showed that the solution depends on the selection of data. The suggested approach allows to build control models which can operate in automatic mode and can be used as decision support systems by giving reasons for the integration of control systems and the assessment of potential savings.

The article substantiates the need to take into account comparable conditions in determining the energy savings that arose in the implementation of measures to improve energy efficiency. Peculiarities of the implementation process of the energy service contract and accounting for comparable conditions are described, with reference to projects implemented within the framework of this type of contracts. An algorithm for determining the basic level of consumption and verification of measurements within the framework of the implementation of energy service contracts is described. The selection as a method of accounting for comparable conditions of independent consumption adjustments arising under the influence of various factors is substantiated. An example of the implementation of the energy service contract, including the modernization of the thermal network of the thermal power plant of the Siberian Federal District, is considered. For this contract, the process of determining the savings as a difference between the actual loss of thermal energy through isolation and the standard losses, the achievement of which was planned within the framework of the project, is described. The main realized stages of the algorithm are briefly described. With the purpose of approbation of the put forward assumptions, the process of determining the relationship between the losses of thermal energy and the air temperature was carried out. As a model for determining the correction, the pairwise linear regression equation was used. Based on the received value of the coefficient, the adjusted losses for each year of the project implementation are calculated and the savings in natural and monetary terms obtained under comparable conditions are determined. The effect of a change in the economy on the economic efficiency of the project is shown. The conclusion is made on the reasons for refusal to correct heat losses in the original version of the project. Recommendations are given on accounting for various comparable conditions, depending on the target value of various project indicators. The publication was prepared within the framework of the RFBR-supported scientific project No. 16-12-54003.

The increasing of man-induced ecological stress implies the reducing of potable water stock which, in turn, causes the necessity of wastewater treatment processes efficiency increasing. Since the natural experiments performance is impossible then significance of mathematical modeling as an instrument of wastewater treatment processes researching is difficult to overestimate. The observations show that a day can be divided in periods of approximately constant of the wastewater inflow rate. Also, the system of balancing tanks can maintain the constant value of sewage water flow. In view of these factors, the further development of compartment mathematical model for wastewater biological treatment process is proposed in this paper. The model is the system of ordinary differential equations with variable structure. The hour inflow rate of wastewater is supposed to be constant. The presented model describes the concentration dynamics of two types of substrates, oxygen and overall activated sludge microorganisms concentration. The introduced threshold functions permit to describe the oxidation process switching from organic substance to ammonium and the microorganisms oxygen competition. Also these functions permit to avoid inadmissible values of model concentrations which may emerge as a result of computational errors. The Monod function is used to take into account the oxygen and substrate saturation of microorganisms. The software package in Python is developed. The parameter identification is presented. The model output concentrations corresponding different control parameter values are set in the article. The results of computer modeling show the adequacy of the developed model to the wastewater treatment process realizing at Petrozavodsk (Karelia) plant. The proposed model permits to solve the problem of forecasting and control of biological wastewater treatment process and to develop algorithms for its stabilizing and optimizing.

This article is about basic problems from water resources depletion. An actuality of impact of some factors on water resource depletion research is shown. A dynamics of the depletion is considered in a link with follow: volumes of fresh water use, volumes of recycled water, volumes of polluted water discharged, volumes of internal renewable water and percentage of marine protected areas. Factor and non-factor models of the depletion are built: linear multi-factor model, model in the state space, autoregressive models. Annual value series of factors are researched to pair correlation. Coefficients of models are determined by a least square method. A choice of the model is based on quality of approximation of source data and quality of post-forecast. The linear multi-factor model is the best in post-forecast and has satisfactory quality of the approximation. Controlled and independent from decision makers factors are substantiated. A forecast of water resources depletion with a persistence of factor trends is made. An increasing of the depletion is shown in conditions of the persistence. An analysis of possible situations in depend of uncontrolled factors is made for the nearest future. A worst combination of the independent factors is determined. It leads to the maximum depletion. Possibilities of decision makers are investigated for improving the situation by changing of controlled factors. It is shown that, although there are negative tendencies in the dynamics of water resources, and the effects of the external environment can worsen them, there are ways to effectively counteract them.

The article deals with the relevance of the problem of computer attacks as the most dangerous form of information security threat materializing. It describes information systems of general use that are the most susceptible to computer attacks. The objects of critical information infrastructure are considered as the targets of computer attacks. Special focus was given to Russian legislative requirements directed to the protection of critical information infrastructure. The research considers the characteristics of the targeted computer attack and its features. The generalized structural scheme of the computer attack and the sequence of its form are examined. The article includes the search for effective algorithms for detection and analysis of the attacks. It presents the generalized characteristic of the main methods for detecting and analyzing computer attacks that are methods for analyzing signatures and methods for detecting abnormal deviations. The information needed to the identification of computer attacks is analyzed. The directions for searching and identifying the effective ways to detect computer attacks are suggested. There is the task to consider the development of algorithmic models providing the recognition of the image of an attack based on a set of its distinctive features. The concept of model adequacy is formulated, it is considered in two aspects: the adequacy of the prototype and the adequacy of the application of information protection measures. There are statements about the advisability of presenting the computer attack detection model in the form of a composition of two main components. The article includes the formation of models of attack on objects of a critical information infrastructure. There is the generalized functional scheme for the formation of the corresponding model. There are necessary analytical expressions for the development of algorithms. The mathematical model of computer attack detection based on the functional approach is proposed. The use of the functional approach presupposes the full set construction of secure conditions of an informational system and its further detection of computer attacks signs.