RU2820047C1 - Method for optimal control of large-scale pipeline systems of power engineering using artificial intelligence technology - Google Patents

Abstract

FIELD: power engineering.

SUBSTANCE: invention relates to the field of power engineering systems and can be used to form control actions on equipment of a large-scale pipeline system (LPS) of power engineering in general in real time, taking into account the declared supply of the product and minimization of energy costs. Method for optimal control of power engineering LPS using artificial intelligence technology includes creation of information control system (ICS) and equipping it with measuring instruments for remote control of parameters of operating modes of the main process equipment of pipeline transport. ICS receives data from measuring instruments, parameters of the processes are transmitted through the ADMS to the simulation and optimization system based on neural networks, designed to perform multivariate calculations and optimize operating modes in accordance with the dispatcher's task, where the most efficient scenario is selected, which is realized under control of a dispatcher by generating in ADMS and subsequent transmission of corresponding control commands to equipment. Optimization criterion can be expenditures of fuel and energy resources for transport needs in natural or value terms, transportation of goods, use of gas stock accumulated in pipelines of the system (for gas transportation systems).

EFFECT: intellectualization of power engineering LPS control by using a simulation system to determine optimal controls in real time, which will make it possible to significantly increase quality of made decisions on system control, including to control distribution of product flows so as to minimize energy costs for transportation, and, as a result, operating costs.

1 cl, 3 dwg

Description

The invention relates to the field of energy pipeline systems and is aimed at improving approaches to finding optimal control actions in large-scale energy pipeline systems (LPS). In particular, the invention can be applied to large-scale gas transmission systems.

There is a known method [US007587326B1], which represents a balance model of the entire gas transportation system. The method allows you to solve the problem of determining the amount of gas and gas distribution simultaneously and systematically throughout the gas transportation network. The invention relates to gas transportation systems and allows maintaining the balance of gas entering and leaving the gas transmission network. It is assumed that logical points (pools) appear in the gas transportation system where it is possible to add or remove gas from the system. Pools are not physical points, so no gas can remain there at the end of the simulation period.

There is a known optimization method [US006829566B2], which allows optimization according to specified criteria. The method is demonstrated using the example of a water supply network. The method makes it possible to determine the peak flow requirements of each pipe and takes into account the operating criteria of the network.

There is a known method for automatically optimizing the operation of a natural gas transportation system [RU2007116343A], with which you can determine the values of continuously changing variables (pressure and gas flow at any point in the system) and discrete values (state of starting compressors, state of opening compressor stations, state of opening control valves, condition of compressor station bypass elements, condition of control valve bypass elements, orientation of compressor stations and orientation of control valves). The method is characterized by the fact that intervals of values of continuously changing variables and sets of values of discrete variables are selected as the initial state of optimization.

There is a known method for automatic optimization of a gas transmission network [EP1852820A1], including determining values for continuous variables, such as pressure and flow at all points of the gas transmission network, and determining values for discrete variables, for example, the opening state of control valves, etc. The method differs in that ranges of values for continuous variables and sets of values for discrete variables are selected as the initial state of optimization.

There is a known useful model of a unified control system for a pipeline system [RU140620U1], which is used for integrated continuous management and monitoring of energy pipeline systems.

The disadvantage of the known methods is the inability to take into account the hydraulic component when modeling the operating mode of the CTS. As a consequence, this drawback leads to the impossibility of creating sufficiently accurate control methods and, consequently, to a significant error in the control of technological processes. In particular, today there is no effective method for managing the CTS that allows taking into account the physical processes of product transport.

The technical result, which the proposed technical solution is aimed at achieving, is to increase the efficiency of management of the power supply system, including ensuring the optimal distribution of product flows through the pipeline system according to the selected optimization criterion.

The essence of the proposed method is as follows.

The specified technical result is achieved by the fact that the Method for optimal control of the CTS of the energy industry using artificial intelligence technology includes the creation and equipping of an information and control system (ICS) with equipment that provides remote control over the specified parameters of product transport (transportation volumes, pressure distribution throughout the system, control actions on KS), modeling in real time the process of product transport and performing multivariate calculations of operating modes of the pipeline system, managing calculations in real time (real-time), selecting the operating mode of the system based on the task and control criterion, as well as the formation and transfer of control commands into an automated dispatch control system for visual control of personnel and the decision maker (Fig. 1).

Currently, almost all PTS are equipped with automated dispatch control systems (ADCS), which operate under the control of dispatch personnel and provide control of the parameters of product transportation processes and remote control of the operating modes of the pipeline system and equipment. If the equipment of the automated control system does not provide control and management to the required extent for adequate modeling and optimization of product transportation processes, then the system is retrofitted.

In fig. 1 shows a diagram of the information management system (IMS), Fig. 2 shows a diagram of the integration of neural networks of subsystems, Fig. Figure 3 shows a schematic diagram of the integration of neural networks of subsystems into a common neural network according to the topology of the UGSS GTS.

The following symbols are shown in these figures:

1.1 - measuring equipment;

1.2 - automated dispatch modeling system (ASDS);

1.3 - modeling and optimization system using neural networks;

1, 2, 3 - subsystem numbers.

The method involves the creation of a computer system that includes a modeling and optimization system, which is based on the use of a neural network model.

The parameters of the product transportation processes are monitored using measuring equipment (1.1). The measurement results enter the ASDU (1.2), from where they are transferred to the modeling and optimization system using neural networks (1.3), which is a software complex. In the modeling and optimization system, the operation of the system is simulated, taking into account incoming data on the parameters of transport processes, which provide information about the current state of the system, possible scenarios for the operation of the pipeline system are calculated in real time, taking into account all restrictions on the operation of equipment and subsystems. According to a given algorithm, calculations are automatically managed and, based on the specified objective functions, the most effective scenario is selected, taking into account the product transportation plan and other limiting factors. The found optimal options for control actions are sent to the dispatcher, who monitors them, corrects them if necessary, and then transfers them to the process equipment using the automated control system.

The modeling of energy pipeline systems is based on the theory of hydraulic circuits [A.P. Merenkov, V.Ya. Khasilev. Theory of hydraulic circuits. Moscow "Science", 1985]. In each hydraulic system, the main components are distinguished: sources of pressure or flow (for example, compressor stations (CS), storage tanks, etc.), a hydraulic network (interconnected pipelines, air ducts, etc.) and consumers. Previously, classical methods of hydraulic calculation were applied to hydraulic systems, however, the invention under consideration involves the use of “hybrid” models - a set of hydraulic and neural network models for each subsystem of the pipeline system.

The method assumes that the pipeline system is a set of adjacent subsystems, which can be either active and passive elements (pipes, pressure reduction points (PRP), compressor stations or pumping stations), or fragments of a pipeline system consisting of several such elements. Each subsystem is equipped with supervisory control measuring devices. Data coming from sensors installed at system facilities are transferred to a database in the dispatch control system, where they are processed and then used to carry out multivariate calculations.

A neural network model is created separately for each subsystem (if subsystems are elements, then for each type of object (CS, LU, etc.). At the first stage, information coming from equipment and measuring instruments is collected. Operating mode parameters are received from sensors in real time. The dimension of the problem can be reduced by aggregating some elements (unification of consumers between two neighboring CS, model of the CS as a whole, rather than individual workshops, etc.).

The optimization model evaluates all control criteria, generates a task according to the criterion set by the dispatcher, solves the problem and offers an optimal scenario for controlling the CTS in the current state, which is used by the dispatcher to change the operating parameters of the system. The problem of efficient flow distribution of the product throughout the system as a whole is solved, characterized in that the ICS with a modeling system is formed on the basis of a neural network model of the pipeline system, which is a combination of neural network models of its subsystems, which are differentiable functions, supply an automated process control system and a dispatch control ICS, create a database for a modeling system with specified parameter values in the form of limit values of controlled operating and technological parameters, the measurement data enters an automated process control system, in which reliable information about the measurement data from the dispatch control information system is transferred to the modeling system, which in real time carries out optimization calculations using the obtained reliable information about measurement data through a mathematical model, which determines the operating mode taking into account the specified control criterion of the pipeline system, allows, with a high degree of detail and accuracy, to determine changes in the operating mode of large-scale PTS, including the distribution of flows, pressures and temperatures of the transported product, parameters of production facilities, due to changes in controls or external conditions, allows for large-scale CTS to determine optimal control actions that ensure minimization (maximization) of a given criterion while ensuring the necessary supplies to consumers and fulfilling technological restrictions, including pressures, volumes and temperatures of the transported fluid, as well as production capacity parameters. The presented technical solution is an element of the dispatch control system.

Statement of the problem of searching for optimal control of the CTS using an ICS modeling system.

The problem statement is shown using the example of minimizing energy costs for gas transportation.

The design diagram of the subsystem is formed by an expert taking into account the results of the analysis of characteristic operating modes in the season under consideration. The model for each subsystem includes functions of the cost of fuel and energy resources for its own technological needs (STN), the gas reserve accumulated in the pipes of the subsystem and the vector function of power consumption at the CS of the subsystem, which have the form

and also each subsystem is characterized by a vector of flow distribution at the joint nodes of the i-th subsystem

Where - function of fuel and energy resources costs for STN of the i-th GTS subsystem; pressure distribution vector in the nodes of the i-th subsystem, which are input to the subsystem; pressure value in nodes (which are the inputs of the i-th subsystem); pressure distribution vector in the nodes of the i-th subsystem, which are the output for the subsystem; pressure value in nodes (which are the outputs of the i-th subsystem); butt nodes of the i-th subsystem, number of joint nodes of the i-th subsystem; - vector of compression ratios of the CS of the i-th subsystem; - vector of characteristics of the i-th subsystem; - gas reserve accumulated in the pipes of the i-th subsystem; - vector function of power consumption at the CS of the i-th GTS subsystem; vector of flow distribution in the junction nodes of the i-th subsystem, flow value in nodes of the i-th subsystem.

A selection of operating modes for training each subsystem is formed by calculating functions for various combinations , which makes it possible, after training the neural network, to determine vector functions of the form

(1)

which are represented by deep neural networks that connect the input parameters of the GTS subsystem model and the corresponding values of the volume of transported gas and consumption of fuel and energy resources.

Controls for the system as a whole are the operating parameters of the active elements of the system - the compression ratio at the compressor station. As a result of solving the optimization problem, the gas pressure is determined for each joint node of the subsystems in accordance with the optimal control actions, and the flows and other simulated characteristics of the subsystem are determined for each subsystem.

Technological calculations for the system are carried out in compliance with restrictions in the form of equalities - nonlinear equations of gas balance in butt joints:

(3)

Where - the set of all butt joints; - a set of subsystems for which node k is a butt node; - gas flow through the i-th subsystem through the butt joint k; ( - the difference between the local gas supply to the k-th butt joint and its distribution (residual), ; pressure in the k-th joint node, if the k-th node is the output node of the i-th subsystem; pressure in the k-th joint node, if the k-th node is the input node of the j-th subsystem.

In the problem under consideration for the entire system, the required variables are gas pressure values in butt joints and the values of the degrees of compression of the CS, on which inequality restrictions of the form are imposed:

(4) (5) (6)

Where - set of CS of the considered GTS.

In addition, in a similar way to (4), conditions must be specified on nodes that are not interface nodes of subsystems in an amount sufficient to obtain a unique solution.

Condition (3) is ensured by constructing a unified neural network model of the system. It is formed like this:

- the input layer consists of neurons to which values are supplied ;

- the input layer is connected to the input layers of the neural networks for the subsystems in accordance with (3) for pressures in joint nodes, and identically for other parameters;

- the output layer aggregates data from the output layers of subsystem models.

An example of constructing a combined neural network model for a simple system of 3 subsystems with one interface node is presented in Fig. 2 and fig. 3. For each neural network model there is a set of input and output parameters . When combining (Fig. 2), the following equalities are satisfied:

,

,

,.

the flow balance condition (2) in this case will take the form

In fig. Figure 3 shows a schematic diagram of the integration of neural networks, which shows the pressure in the joint node as a value ( transmitted to the input of the neural network.

The objective function for the system model has the form:

(7)

Where weight coefficient.

The set (1), (2), (4), (5), (6), (7) is an optimization problem formed using smooth differentiable functions. This formulation of the problem allows the use of optimization packages based on gradient methods, which make it possible to use the automatic differentiability of such functions. The quality of the resulting optimal management decisions depends on the optimization package used.

The invention can be applied to various energy pipeline systems using appropriate terminology.

The proposed invention can be used in the development of an integrated control system using a modeling and optimization system based on neural networks, which will make it possible to intellectualize the control of gas flows in the CTS with automation of the decision-making process on the selection of system operating parameters in real time, ensuring delivery of the product to consumers and the optimal solution with acceptable accuracy.

The proposed invention improves the energy efficiency of the pipeline system.

The method can be used when planning the modes of pipeline systems using nonlinear criteria (criterion for the minimum energy or cost of fuel and energy resources for transport needs, etc.)

The method is characterized by a fast and efficient solution to the optimization problem by using the differentiability of neural networks.

The invention allows you to balance the system and select the optimal operating mode, according to the selected optimization criterion.

Claims (1)

A method for optimal control of large-scale pipeline systems (LPS) in the energy sector using artificial intelligence technology, including the creation and equipping of an information management system (IMS) with equipment that provides remote control over the specified parameters of transportation processes, and remote control of the operating modes of the pipeline system, simulating in real mode time process of product transport, collection of measured data from the ICS equipment, performing multivariate calculations of operating modes of power engineering equipment, managing calculations in real time, selecting the operating mode, generating and transmitting control commands to the dispatch control ICS to change the operating mode, modeling system with software - a hardware complex included in the IMS, measuring specified parameters using measuring devices located with the ability to measure and transmit them, transmitting measurement data in real time with the possibility of their visual control by dispatch personnel and with the possibility of multivariate forecast calculations in the modeling system, implementation in the modeling system in real time of multivariate calculations with the ability to select the operating mode of the energy pipeline system, taking into account the technological limitations of the operation of system objects, characterized in that the IMS with the modeling system is formed on the basis of a neural network model of the pipeline system, which is a combination of neural network models of its subsystems representing are differentiated functions, supply an automated process control system and a supervisory control system, create a database for a modeling system with specified parameter values in the form of limit values of controlled operating and technological parameters, measurement data enters an automated process control system, in which reliable information about measurement data from the automated dispatch control system is transferred to the modeling system, which in real time carries out optimization calculations using the received reliable information about the measurement data through a mathematical model that determines the operating mode taking into account the specified control criterion of the pipeline system, allows for a high degree of detail and accurately determine changes in the operating mode of large-dimensional CTS, including the distribution of flows, pressures and temperatures of the transported product, production capacity parameters, due to changes in controls or external conditions, allows for large-dimensional CTS to determine optimal control actions that ensure minimization (maximization) of a given criterion at ensuring the necessary supplies to consumers and meeting technological restrictions, including pressure, volume and temperature of the transported product, as well as production capacity parameters.