Multi-layer linkage optimization method for thermal power generating unit control system

Abstract

The invention relates to the technical field of thermal power plant automation control, and discloses a thermal power generating unit control system multilayer linkage optimization method, which comprises the steps of discretizing a thermal power generating unit physical system into a control cellular network; for each cell, calculating an entropy potential fusing thermodynamics, security constraints, economy and control target information so as to construct a dynamic entropy potential field; each cell autonomously determines a control action according to a local gradient of an entropy potential field, and further, for the cells which belong to different units and are connected to the same physical sharing point, cooperative control without central scheduling is realized based on the entropy potential of each cell through a resonance coupling mechanism. According to the invention, by constructing a decentralized self-organizing control system based on entropy potential field driving, the dependence on an accurate global model is eliminated, the robustness and dynamic response speed of the system are improved, and rapid intelligent cooperation among multiple units is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic control of thermal power plants, in particular to a multi-layer linkage optimization method for a thermal power unit control system. BACKGROUND

[0002] As the main power source in the power system, the stability, economy and flexibility of the thermal power unit are crucial to the safe and reliable power supply of the power grid. In recent years, with the rapid growth of renewable energy generation, the power grid has made unprecedentedly high demands on the peak shaving capacity, response speed and operational flexibility of traditional thermal power units. Thermal power units not only need to bear the base load generation task, but also need to frequently deep peak shave in a wide load range to balance the intermittency and volatility of renewable energy.

[0003] In order to realize accurate control of the complex thermal system of the thermal power unit, the existing technology usually adopts advanced control strategies based on process mechanism models or data models. For example, coordinated control systems are widely used to uniformly regulate the response of the boiler and the steam turbine to meet the instructions of the unit load. However, these control methods gradually reveal their inherent limitations when dealing with the current complex and variable operating conditions.

[0004] The performance of the existing advanced control strategy is highly dependent on the accuracy of the mathematical model. The thermal power unit is a multivariable, strongly coupled, nonlinear time-varying system that includes combustion, heat transfer, phase change and flow. It is extremely difficult to establish a mechanism model that can accurately describe the dynamic characteristics of the unit in the entire operating condition range. At the same time, as the unit operates, factors such as equipment aging, coking and pollution of the heating surface, and changes in fuel properties will cause the actual operating characteristics of the unit to deviate from the initially established model, i.e., a model mismatch problem. This mismatch will directly lead to a decline in control performance, especially in non-design conditions such as rapid load change or low load operation, control overshoot and oscillation phenomena occur frequently, affecting the safety and economy of the unit operation.

[0005] In addition, the currently widely used control system architecture is mostly centralized and hierarchical. Under this architecture, the data collected by the bottom layer sensors needs to be uploaded to the central controller for calculation and decision-making, and the generated control instructions are then issued to the field actuators. This serial information processing and instruction transmission process inevitably introduces communication and calculation delay. In the steady-state operation of the unit, the impact of this delay is not significant; but in the dynamic process that requires fast response, such as primary frequency modulation or emergency load adjustment, this inherent time lag will become a key bottleneck that restricts the system response speed and control quality. When this centralized architecture is further applied to the coordinated control of multiple units, the information dimension and calculation complexity that the system needs to handle grow exponentially, making it extremely difficult to achieve real-time and dynamic optimal coordination among multiple units, and often relying on relatively static and simplified allocation strategies, making it difficult to fully tap the overall adjustment potential of the entire unit group. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application provides a multi-layer linkage optimization method for a thermal power unit control system, which solves the problem that the existing thermal power unit control method has poor control performance and difficulty in coordinated response in complex working conditions such as deep peak regulation due to its dependence on accurate mathematical models, inherent hierarchical delay of centralized control architecture, and coordination barriers between multiple units.

[0007] To achieve the above object, the present application is implemented by the following technical scheme: the present application provides a multi-layer linkage optimization method for a thermal power unit control system, which divides a complex physical system into a distributed control cell network and introduces a unified entropy potential field concept as the driving force for system regulation, replacing the traditional centralized optimization calculation.

[0008] By defining entropy potential to quantify heterogeneous information including thermodynamics, safety constraints, economy, and control objectives, the system can interact based on local information to form a global and adaptive regulation behavior.

[0009] Further, by establishing a resonance coupling mechanism across systems, multiple units can form a fast coordinated response without central instructions when facing a common disturbance.

[0010] Specifically, the method can include the following steps:

[0011] First, the physical system of one or more thermal power units is discretized into a network topology composed of multiple control cells according to its functional units or physical location.

[0012] The control cell is a software agent of the physical unit, and the network topology is constructed according to the actual transmission relationship of energy flow, material flow, and information flow.

[0013] Secondly, for each control cell i in the network topology, its entropy potential E i (t) at time t is calculated

[0014] E i (t) = w thermo E i,thermo (t) + w const E i,const (t) + w target E i,target (t) + w econ E i,econ (t) + w

[0015] E thermo (t) + w const E target (t) + w econ E i,thermo where w

[0016] The specific calculation of each component in the formula is as follows:

[0017] The thermodynamic entropy potential E i,thermo (t) is calculated according to the state vector of the control cell and its adjacent cells through a preset thermodynamic function, which represents the irreversible loss in the energy transfer or conversion process.

[0018] The constraint entropy potential E i,const (t) is calculated according to the deviation of the real-time operating state parameters of the control cell from its preset constraint boundary. One specific implementation is to calculate it through the following exponential function:

[0019]

[0020] where x i,k (t) is the jth constrained parameter of cell i, x k,opt is its ideal operating center value, x k,lim is its safe operating limit value, and a k is a preset coefficient.

[0021] The target deviation entropy potential E i,target (t) is calculated according to the deviation between the actual output O actual (t) of the unit and the target output O target (t) for the end control cell directly related to the final output of the unit, for example, by calculating

[0022] The economic entropy potential E i,econ (t) is calculated according to the weighted sum of quantifiable cost items related to the operation of the control cell.

[0023] The enthalpy potentials of all the control cells together form a dynamic enthalpy potential field.

[0024] Then, each control cell in the network topology determines its own control action according to the gradient of the enthalpy potential field formed locally by the control cell and according to a preset local interaction rule. i (t), which makes the predicted enthalpy potential of the control cell in the next time step minimum.

[0025] Finally, the collaborative control between different thermal power units is realized. This step is completed through a resonance coupling mechanism, which includes identifying the points physically shared between different thermal power unit systems and defining them as resonance nodes. For each resonance node k, a resonance enthalpy potential E j (t) is calculated according to the enthalpy potentials E k,res (t) of all control cells j connected to the node and belonging to different thermal power units. A specific implementation is to calculate it through the following weighted summation formula:

[0026]

[0027] wherein, is the set of all cells connected to the resonance node k, and γ j is a preset coupling weight coefficient. When the resonance enthalpy potential E k,res (t) changes due to external disturbances, the change amount acts on all thermal power units connected to the resonance node. Distributed adjustment is performed within each unit system to reduce its contribution to the change in the resonance enthalpy potential, ultimately forming a collaborative response of multiple units.

[0028] The second aspect of the present application provides a multi-layer linkage optimization system for a thermal power unit control system, which comprises:

[0029] a network construction module, configured to discretize the physical system of at least one thermal power unit into a network topology composed of multiple interconnected control cells;

[0030] an enthalpy potential calculation module, configured to calculate an enthalpy potential for each control cell in the network topology according to its own real-time operating state and a preset optimization target, and to construct a dynamic enthalpy potential field based on the enthalpy potentials of all the control cells;

[0031] a distributed decision-making module, configured in each control cell, for determining its own control action according to the gradient of the enthalpy potential field formed locally by the control cell and according to a preset local interaction rule;

[0032] The synergic control module is configured to realize synergic control among different thermal power generating units through a resonance coupling mechanism based on entropy potential of each control cell for multiple control cells belonging to different thermal power generating unit systems and connected to the same physical shared point.

[0033] In one embodiment, the entropy potential calculation module is further configured to obtain the entropy potential of the control cell by weighted sum of the thermodynamic entropy potential, the constraint entropy potential, the target deviation entropy potential and the economic entropy potential.

[0034] The present application provides a multi-layer linkage optimization method for a thermal power generating unit control system.

[0035] 1. The present application converts a complex multi-objective optimization problem into an adaptive adjustment process in a unified scalar field by discretizing a thermal power generating unit physical system into a control cell network and calculating an entropy potential for each cell, which can quantitatively unify heterogeneous information such as thermodynamics, safety constraints, economy and control targets, thereby eliminating the dependence on an accurate global mathematical model and improving the robustness and adaptability of the control system under the influence of uncertain factors such as equipment aging, fuel property changes and unmodeled dynamics.

[0036] 2. The present application distributes control decision-making rights to the execution unit closest to the disturbance source by making each control cell determine its own control action in parallel according to the gradient formed by the entropy potential field in its local area and in accordance with a pre-set local interaction rule, thereby realizing distributed and concurrent execution of control responses, significantly shortening the delay time from disturbance occurrence to control action execution compared to the traditional top-down centralized control architecture, and improving the dynamic response speed of the system.

[0037] 3. The present application establishes a resonance coupling mechanism at a physical shared point for control cells belonging to different thermal power generating units and calculates a resonance entropy potential based on the entropy potential of each cell, thereby providing a unified environmental pressure signal for multiple independent unit systems, enabling each unit to spontaneously perform competitive regulation based on bottom-up information interaction, and realizing fast and intelligent synergic control among multiple units without the need for upper-level centralized dispatching instructions, thereby improving the overall stability and coordination efficiency of the unit group when dealing with global tasks such as grid disturbance. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 The method flowchart of the present application;

[0039] Figure 2 The system architecture diagram of the present application. DETAILED DESCRIPTION

[0040] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0041] Embodiments:

[0042] Please refer to the accompanying Figure 1 , the embodiment of the present application provides a multi-layer linkage optimization method for a thermal power generating unit control system, comprising the following steps:

[0043] S1: discretizing the physical system of at least one thermal power generating unit into a network topology composed of multiple interconnected control cells;

[0044] In the embodiment, in order to realize a distributed self-organizing control which is independent of a global centralized model, the present application first provides a method for structurally abstracting and modeling the physical system of a thermal power generating unit. The method aims to transform a continuous, complex and strongly coupled industrial process into a discrete network system composed of standardized units, which is easy for distributed computing and information interaction.

[0045] Specifically, the first step of the present application is to discretize the physical system of at least one thermal power generating unit into a network topology composed of multiple interconnected control cells. This step is the basis for building the subsequent entropic potential field and realizing distributed control.

[0046] The control cell is not a physical entity, but a software abstraction or software agent of the smallest unit in the thermal power generating unit which has a specific function or is in a specific physical location. The granularity of discretization is adjustable, and in a preferred embodiment, a functionally relatively complete device or a physically representative area can be defined as a control cell according to the requirements of control accuracy and computing resources.

[0047] In order to enable each control cell to accurately represent the state of its corresponding physical unit and participate in subsequent calculation and interaction, the present embodiment defines the core attributes of each control cell i. Among them, the most important attribute is its state vector S i (t). The state vector is a set of real-time parameters describing the physical state of cell i at time t, which is composed of:

[0048] S i (t)={p i,1 (t),p i,2 (t),…,p i,m (t)};

[0049] In this definition, pi,m (t) represents the mth physical parameter collected by the sensor system in real time. These parameters comprehensively reflect the operating conditions of the physical unit represented by the cell, which can include but are not limited to temperature, pressure, flow, enthalpy, medium chemical composition, liquid level, valve opening, motor speed, etc. This state vector is a direct data source for subsequent calculation of the entropy potential of the cell, and constitutes the cornerstone of the cell's perception of the physical world.

[0050] Further, after all the control cells are defined, it is necessary to build the connection relationship between them to form a network topology that can reflect the internal relationship of the real physical system. In this embodiment, the network topology is constructed as a weighted directed graph G = (V, E).

[0051] In this network topology G, the set represents the set of all control cells generated by the discretization of the thermal power unit.

[0052] The set E represents the set of edges connecting the control cells. The construction of these edges is not arbitrary, but strictly follows the transmission path of the real energy flow, material flow and information flow in the physical system. Specifically: an edge from cell A pointing to cell B can represent the flow of matter from the physical location corresponding to A to the physical location corresponding to B, for example, water flowing from the feed water pump cell to the economizer cell. Or, the edge can represent the transmission relationship of energy, for example, heat is transmitted from the cell representing the flue gas side to the cell representing the water vapor side. Or, the edge can also represent the transmission path of certain information or control instructions, although the design of the present invention aims to minimize the dependence on external centralized instructions.

[0053] The network topology constructed in this way not only maps the physical entities of the thermal power unit in structure, but also reproduces its complex internal interaction relationship in function. This network topology provides a path for the subsequent propagation of entropy potential and the formation of gradient, so that local control actions can have a global impact along the real physical connection.

[0054] Thus, through the above-mentioned discretization and network topology construction, the present invention successfully converts a continuous physical system that is difficult to directly perform global optimization into a distributed network composed of a large number of standardized control cells with basic perception ability. This network provides the necessary technical premise for the subsequent introduction of the entropy potential field concept and the realization of self-organization and adaptive control based on local information interaction, so that the system can realize rapid, stable and economic response to operating conditions without the need to establish and solve complex global mathematical models.

[0055] S2: for each control cell in the network topology, a scalar value is calculated according to its real-time running state and preset optimization target, which is used to quantitatively unify the non-ideal degree of the control cell as its entropy potential; the entropy potentials of all control cells together form a dynamic entropy potential field;

[0056] In this embodiment, after the discretization of the physical system of the thermal power unit and the construction of the control cell network topology are completed, one of the core steps of the present application is to assign a quantitative index to each control cell in the network to guide its behavior. To this end, the present application provides a method for calculating and constructing an entropy potential field, which aims to unify multiple heterogeneous and even conflicting optimization targets into a single scalar field with clear physical meaning, thereby providing driving force for subsequent distributed self-organizing control.

[0057] Specifically, for each control cell in the network topology, a scalar value is calculated according to its real-time running state and preset optimization target, which is used to quantitatively unify the non-ideal degree of the control cell as its entropy potential. The entropy potentials of all control cells together form a dynamic entropy potential field.

[0058] The calculation of this entropy potential is not a single-dimensional evaluation, but a comprehensive quantitative process. In a preferred embodiment, for any control cell i in the network, the total entropy potential E i (t) of the control cell at time t is obtained by weighted summation of four key component entropy potentials. The specific calculation formula is as follows:

[0059] E i (t)=w thermo E i,thermo (t)+w const E i,const (t)+w target E i,target (t)+w econ E i,econ (t);

[0060] In this formula, w thermo ,w const w target ,w econ are all preset dimensionless weight coefficients. The setting of these coefficients reflects the emphasis degree of the specific unit on each performance index in different running stages, for example, the weight related to response speed can be appropriately increased during fast load variation, and the weight related to economy can be increased during steady-state operation.

[0061] The four component entropy potentials in the formula are described in detail as follows:

[0062] First, the thermodynamic entropy potential E i,thermo(t). The introduction of this term aims to quantify the irreversibility in energy conversion and transfer from the fundamental level of the second law of thermodynamics. The higher the thermodynamic entropy potential of a control cell, the greater the loss or entropy production, i.e. the more severe the degradation of energy quality, of the physical process it represents. Its calculation is a pre-defined function of the state vector of the cell and its neighboring cells:

[0063] E i,thermo (t) = f thermo (S i (t), {S j (t) | j e Neighbors(i)}); (2)

[0064] The specific form of the function f thermo depends on the physical process represented by the cell, for example, for a heat exchanger cell, it can be constructed based on the logarithmic mean temperature difference and the heat transfer coefficient; for a throttling process, it can be constructed based on the pressure and temperature before and after throttling.

[0065] Secondly, there is the constraint entropy potential E i,const (t). The purpose of this term is to ensure the safe and stable operation of the unit. It quantifies and punishes the approach of any real-time operating state parameter to the pre-set safety or process constraint boundary by forming a strong potential barrier. When the parameter operates within the normal range, the value of this term is close to zero; once the parameter approaches its limit, the value of this term will grow exponentially. A preferred calculation method is as follows:

[0066]

[0067] where x i,k (t) is the kth constrained parameter (such as the wall temperature of the steam drum) in the state vector of the cell; x k,opt is the ideal operating center value of the parameter; x k,lim is its non-crossable safety upper or lower limit; a k is a coefficient used to adjust the steepness of the barrier, the greater the value, the more important the constraint, and the stronger the repulsive force of the barrier.

[0068] Thirdly, there is the target deviation entropy potential E i,target (t). This term constitutes the ultimate guide of the entire system operation, i.e. the gravitational source of global optimization. It is mainly applied to the end cells directly related to the final output of the system, to quantify the deviation between the actual output and the desired target. Its calculation formula can be expressed as:

[0069] E i,target (t) = (O actual (t) - O target (t)) 2 ;​

[0070] where O actual (t) is the actual total output of the unit at time t, while O target (t) is the target total output from the grid dispatch or operation plan. The square form of the deviation ensures that any deviation in either direction will lead to an increase in the entropy potential, thus driving the system to converge towards the target value. For non-terminal cells in the network, the value of this term is usually set to zero.

[0071] Finally, there is the economic entropy potential E i,econ (t). The introduction of this term allows the control decisions to directly take into account the economic cost of operation. It quantifies parameters related to the operation of a particular control cell that can be converted into cost, such as the power consumption of auxiliary equipment, the consumption of chemical agents, the leakage or loss of working medium, etc. Its calculation method is the weighted sum of various cost parameters:

[0072]

[0073] where c i,j (t) is the instantaneous measured value of the jth economic-related parameter of cell i; β j is the price coefficient or weight coefficient that converts this physical quantity into a unified cost measure.

[0074] By calculating the above-mentioned comprehensive entropy potential value for each control cell in the network in real time and in parallel, a dynamic scalar field, i.e. an entropy potential field, is constructed throughout the entire thermal power generating unit. The visual form of this field is like a geographical topography with ups and downs, where the areas with ideal operating conditions appear as plains or basins, and the areas with problems or deviations from the target appear as mountains or highlands.

[0075] Thus, by constructing the entropy potential field, the present application successfully converts a complex control problem with multiple variables, multiple constraints, and multiple targets into a more physically intuitive problem of finding and maintaining the lowest potential energy path in a unified potential field. This dynamic entropy potential field provides clear and unified gradient information for subsequent distributed decision-making, so that the local regulation behavior of each cell can serve the overall goal of minimizing the global entropy potential, thus making it possible to achieve self-organizing optimization of the system.

[0076] S3: Each control cell in the network determines its own control action according to the gradient formed by the entropy potential field in its local area and in accordance with the pre-set local interaction rules;

[0077] In this embodiment, after the control cell network topology is constructed and the dynamic entropy potential field is established, the application provides a distributed and self-organizing decision and control execution mechanism. The core purpose of the mechanism is to enable each independent control cell in the network to autonomously make a control decision that is beneficial to the optimization of the global system state based on the local information available to it, thereby replacing the traditional control mode that relies on a central controller for centralized calculation and instruction issuance.

[0078] Specifically, each control cell in the network topology of the application determines its control action according to the gradient of the entropy potential field formed locally and in accordance with its pre-set local interaction rules.

[0079] The first step of this process is the perception of the local gradient. In one embodiment, each control cell i does not obtain global information of the entire entropy potential field, but only needs to perceive its own entropy potential E i (t) and interact with the neighborhood cells j that have a direct connection relationship with it through the network topology G to obtain the entropy potentials E j (t) of these neighborhood cells. The difference between the self-entropy potential and the neighborhood cell entropy potential constitutes the local gradient of the entropy potential field that the cell can perceive. This gradient intuitively reflects the "potential difference" of energy, matter or information and indicates the direction for the decision of control action.

[0080] Based on the perception of the local gradient, the next step is the decision process of determining the specific control action. Instead of using a fixed, top-down control logic, the method of the application decentralizes the decision-making power to each cell. The decision goal of each control cell is single and clear:

[0081] That is, to select a control action that can make its own entropy potential lowest in the next control period in the future. This is essentially a local and forward-looking optimization problem.

[0082] In a preferred embodiment, the process of determining the control action u i (t) of control cell i at time t can be realized by solving the following optimization problem:

[0083]

[0084] In this formula, u represents the set of all available control actions of cell i, for example, different opening setting values for a valve cell. represents the predicted value of the entropy potential of cell i itself at the next time t+1 when the control action u is executed at the current time t.

[0085] Predicted entropy potential The calculation can be based on a simplified mathematical model describing the local dynamic characteristics of the cell, or it can be based on a data-driven predictive model learned from historical operating data. This prediction-based decision-making approach endows each cell with a certain degree of foresight, enabling it to anticipate the potential consequences of an action, thereby avoiding short-sighted control behaviors that could lead to a surge in future entropy.

[0086] The above optimization process ultimately utilizes the pre-defined local interaction rules R for each cell. i These rules are implemented concretely, transforming the abstract principle of entropy gradient descent into logical carriers of executable instructions for specific devices. For example, if a cell senses that the entropy potential of its downstream neighboring cells is significantly higher than its own, indicating a downstream demand for energy or matter, its local interaction rules will trigger a control action aimed at increasing downstream output, such as increasing valve opening or pump speed. Conversely, if the entropy potential of its upstream neighboring cells is significantly lower than its own, indicating abundant upstream resources, its rules will trigger a control action aimed at increasing upstream input.

[0087] Specifically, when the constraint entropy potential E of a cell itself... i,const When (t) begins to grow rapidly, it indicates that it is approaching a certain safety boundary. Its local interaction rules will assign a high-priority control action aimed at keeping the state parameters away from that boundary to ensure the safety of the system.

[0088] Thus, through the aforementioned mechanism, each control cell continuously and in parallel executes a cycle of sensing local gradients, predicting future entropy potentials, and selecting the optimal action. The aggregation of these local decision-making behaviors of all cells macroscopically creates an effect where energy and matter flow spontaneously along the entropy potential field gradient. Just as fluids naturally flow from high-pressure areas to low-pressure areas, similarly, in the system constructed in this invention, the system's operating state spontaneously evolves and migrates from high-entropy potential regions to low-entropy potential regions.

[0089] The establishment of this distributed decision-making and execution mechanism enables the entire thermal power unit control system to exhibit high adaptability and robustness. When a local disturbance occurs, it is immediately sensed by neighboring cells, triggering a series of local, chain-like autonomous adjustments. This allows the disturbance to be quickly absorbed locally before it spreads, without waiting for time-consuming recalculation and command issuance from the upper-level control system. Ultimately, the system as a whole exhibits an emergent behavior that allows it to spontaneously find and maintain the globally optimal operating point under the current conditions without centralized coordination.

[0090] S4: For multiple control cells belonging to different thermal power unit systems but connected to the same physical shared point, a resonant coupling mechanism is used to achieve coordinated control between different thermal power units based on the entropy potential of each control cell.

[0091] In this embodiment, building upon the self-organizing control based on entropy potential fields implemented within a single thermal power unit, the present invention further provides a technical solution that seamlessly extends this control philosophy from the scope of a single unit to a system-level linkage composed of multiple cooperating units. This solution aims to address the communication delays, model barriers, and dependence on centralized coordinators present in traditional plant-level or grid-level coordinated control.

[0092] To achieve this objective, this invention provides a method for coordinated control between different thermal power units, based on the entropy potential of each control cell, through a resonant coupling mechanism, targeting multiple control cells belonging to different thermal power unit systems but connected to the same physical shared point. This mechanism enables coordinated actions between multiple units to emerge spontaneously from underlying physical information interaction as a macroscopic behavior, without the need for intervention from a top-level, independent central dispatching unit.

[0093] Specifically, the implementation of this resonant coupling mechanism first requires the identification and definition of physical shared points. These points are referred to as resonant nodes in this invention. A resonant node is a physically real and inseparable coupling point between multiple independent thermal power unit systems. Typical resonant nodes may include, but are not limited to: the upstream gateway or busbar to which the generator outlets of multiple units are connected, the common chimney to which the flue gas of multiple boilers is discharged, and the main inlet or outlet pipe of the circulating water system shared by multiple units.

[0094] After identifying the resonance nodes, this invention defines and calculates a novel potential function for each resonance node k, called the resonance entropy potential E. k,res (t). The calculation of this potential function does not depend in isolation on a single cell, but is jointly determined by the entropy potentials of all cells connected to this resonant node and belonging to different unit systems, thus forming a cross-system coupling effect. In a preferred embodiment, its calculation can be expressed in the following functional form:

[0095]

[0096] In this formula, This represents the set of all control cells connected to the resonant node k, where cells can come from two or more different thermal power unit systems. g is a predefined coupling function used to fuse entropy potential information from different systems.

[0097] For ease of engineering implementation, the coupling function g can be concretized as a weighted summation, and its calculation formula is as follows:

[0098]

[0099] Where, γ j It is a dimensionless coupling weight coefficient whose value is preset and used to characterize the strength or importance of the influence of the operating state of cell j on the overall state of the resonant node.

[0100] Based on the establishment of the aforementioned resonant entropy potential, coordinated control among different thermal power units is achieved in the following way: when an external disturbance or global command acts on the system, the disturbance will first manifest itself at the associated resonant node. For example, changes in the grid frequency will directly affect the target deviation entropy potential E of the terminal cell at the upstream gateway. i,target (t), which in turn leads to the resonance entropy potential E of the resonance node. k,res (t) changed significantly.

[0101] The change in the resonant entropy potential is like throwing a stone into a calm water surface, creating a potential field disturbance. This disturbance acts as a unified, undifferentiated environmental pressure signal, broadcast simultaneously through the network topology and affecting all thermal power units coupled to the resonant node.

[0102] Upon receiving the environmental pressure signal, the self-organizing control system within each thermal power unit will immediately and independently react. Each unit will autonomously perform internal adjustments, the underlying motivation of which is to reduce its contribution to the increase in entropy potential at the resonant node.

[0103] In this process, a key collaborative behavior emerges spontaneously: for a given regulation task, units with larger internal regulation margins, healthier equipment, and faster response times can absorb the external disturbance more quickly and economically through their internal cellular networks, i.e., reduce their own entropy potential's contribution to the resonant entropy potential more rapidly. This allows them to naturally assume more regulation tasks in this competitive adaptation process with other units.

[0104] Thus, through this competitive adjustment process based on the coupling of underlying physical information, the macroscopic manifestation is ultimately a highly efficient and intelligent collaborative action among multiple units without any central dispatch instructions. The system automatically and in a distributed manner finds the optimal resource allocation scheme to mitigate external disturbances and restore the resonant node entropy potential to stability as quickly as possible, thereby realizing the spontaneous collaborative optimization control of the unit group.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-level linkage optimization method for thermal power unit control systems, characterized in that, Includes the following steps: S1: Discretize the physical system of at least one thermal power unit into a network topology consisting of multiple interconnected control cells; S2: For each control cell in the network topology, a scalar value is calculated based on its real-time operating state and preset optimization target to uniformly quantify its degree of non-ideality, which serves as its entropy potential; the entropy potentials of all control cells together constitute a dynamic entropy potential field. S3: Each control cell in the network topology determines its own control action based on the gradient formed locally by the entropy potential field and according to the preset local interaction rules. S4: For multiple control cells belonging to different thermal power unit systems but connected to the same physical shared point, a resonant coupling mechanism is used to achieve coordinated control between the different thermal power units based on the entropy potential of each control cell.

2. The method according to claim 1, characterized in that, The step of calculating the entropy potential of each control cell specifically includes: Calculate the thermodynamic entropy potential of the control cell based on the irreversible loss during the energy conversion process; The constraint entropy potential is calculated based on the deviation of the real-time operating state parameters of the control cell from its safety or process constraint boundary. Based on the deviation between the actual output and the target output of the thermal power unit, the target deviation entropy potential is calculated for the end control cell associated with the final output. Calculate the economic entropy potential based on the operating costs associated with the control cell; The entropy potential of the control cell is obtained by weighted summing of the thermodynamic entropy potential, constraint entropy potential, target deviation entropy potential, and economic entropy potential.

3. The method according to claim 2, characterized in that, The steps for calculating the constraint entropy potential are as follows: When the real-time operating state parameters of the control cell approach its safety or process constraint boundary, the constraint entropy potential is calculated using an exponential function, such that the constraint entropy potential increases exponentially as the deviation decreases.

4. The method according to claim 1, characterized in that, The steps for determining the respective control actions are as follows: Each control cell selects a control action that minimizes its predicted entropy potential in the next control cycle, thereby enabling the energy flow or matter flow to spontaneously adjust along the local gradient descent direction of the entropy potential field.

5. The method according to claim 1, characterized in that, The resonant coupling mechanism specifically includes: The physical shared points among the different thermal power units are identified and defined as resonance nodes; For each resonance node, a resonance entropy potential is calculated based on the entropy potentials of all control cells connected to this node and belonging to different thermal power units.

6. The method according to claim 5, characterized in that, The steps for calculating the resonance entropy potential are as follows: The resonant entropy potential is obtained by weighted summation of the entropy potentials of all control cells connected to the resonant node.

7. The method according to claim 5, characterized in that, The specific implementation method of the aforementioned collaborative control is as follows: When an external disturbance or change in command causes a change in the resonance entropy potential of a certain resonance node, this change acts as a unified environmental pressure on all thermal power units coupled to that resonance node. Each thermal power unit system performs self-organized adjustment to reduce its own contribution to the change in the resonance entropy potential, thereby giving rise to a coordinated response behavior on a macroscopic level that does not require central command.

8. The method according to claim 1, characterized in that, The control cell is a software agent of the smallest functional unit or physical location unit in the physical system of the thermal power unit.

9. The method according to claim 1, characterized in that, The network topology is constructed based on the actual energy flow, material flow, and information flow transmission relationships within the thermal power unit.

10. A multi-layer linkage optimization system for thermal power unit control system, characterized in that, include: A network building module is used to discretize the physical system of at least one thermal power unit into a network topology consisting of multiple interconnected control cells. The entropy potential calculation module is used to calculate a scalar value for each control cell in the network topology, based on its real-time operating state and preset optimization target, to uniformly quantify its degree of non-ideality, as its entropy potential. And a dynamic entropy potential field is constructed based on the entropy potential of all control cells; A distributed decision-making module, configured in each of the control cells, is used to determine the respective control actions based on the gradient formed locally by the entropy potential field and according to preset local interaction rules. The collaborative control module is used to achieve collaborative control between different thermal power units based on the entropy potential of each control unit through a resonant coupling mechanism for multiple control cells belonging to different thermal power unit systems but connected to the same physical shared point.

Images