Energy-saving optimization method and system for power plant power system

Abstract

The application provides a power plant power system energy-saving optimization method and system, and belongs to the technical field of power plant energy-saving optimization. The energy-saving optimization method comprises the following steps: collecting multi-source operation data to generate a standardized multi-source operation data set; generating a standardized system state vector sequence, establishing a state dynamic evolution model; constructing an electricity consumption-state bidirectional sample set, establishing a multi-solution electricity consumption driven state inversion model; generating a plurality of candidate state solution sets according to the multi-solution electricity consumption driven state inversion model, and constructing a candidate set of the lowest energy consumption reachable state under multiple working conditions; constructing a perturbation structured mapping model to form a perturbation constraint feasible operation domain; constructing a time reachable graph topology structure to screen a globally optimal electricity consumption operation trajectory; building a cross-layer control parameter mapping system, and generating a multi-device collaborative control instruction sequence. The application realizes parallel output of multiple energy-saving operation schemes, smooth and non-impact switching of different energy-saving states, and bidirectional consideration of the global energy consumption optimization of the system and the stable operation of a single device.

Description

Technical Field

[0001] This invention relates to the field of energy-saving optimization technology for thermal power plants, and specifically to an energy-saving optimization method and system for the power consumption system of a thermal power plant. Background Technology

[0002] The main source of power consumption in thermal power plants is auxiliary equipment such as fans and pumps. Optimizing auxiliary equipment operation strategies and reducing the power consumption rate are key aspects of energy conservation and emission reduction in thermal power units. Current conventional energy-saving optimization methods in thermal power plants rely on forward calculation of energy consumption based on equipment operating status, followed by targeted adjustments to auxiliary equipment operating parameters—a reactive, post-implementation optimization model. In actual operation, the same overall plant power consumption level can correspond to multiple sets of different auxiliary equipment load distribution conditions. However, existing technologies default to a single optimal operating condition, ignoring the equivalent energy consumption characteristics of multiple operating conditions. Therefore, it remains difficult to balance the accuracy of overall energy-saving optimization with operational safety.

[0003] Existing energy-saving optimization methods for thermal power plant power systems still have the following shortcomings: First, they cannot adapt to the multi-condition operation characteristics of power plants with the same energy consumption. Traditional optimization models can only output a single optimal operating state, ignoring multiple feasible auxiliary equipment load allocation schemes under the same power consumption. The flexibility of operating condition selection is extremely poor, and it cannot adapt to different peak-shaving needs of the units. Second, they lack state transition constraints and topology relationship modeling, focusing only on the steady-state optimal operating state without considering the switching paths, transition losses, and adjustment limitations between different energy-saving operating conditions. The switching process of operating conditions involves large parameter jumps, threatening the safe operation of the units. Third, they have not built cross-layer collaborative decoupling control logic for auxiliary equipment groups. The optimal operating trajectory at the system level cannot be smoothly decomposed into controllable control commands for each auxiliary equipment. Coupling interference between equipment cannot be eliminated, and the overall energy consumption optimization effect of the system is greatly reduced. Summary of the Invention

[0004] The purpose of this invention is to provide an energy-saving optimization method and system for the power system of thermal power plants, so as to solve the technical problems that existing energy-saving optimization methods for the power system of thermal power plants cannot adapt to the operation scheme of multiple auxiliary machines with the same power consumption, the state switching is prone to causing fluctuations in unit parameters, and the global optimal trajectory is difficult to be effectively implemented.

[0005] To achieve the above objectives, this invention provides an energy-saving optimization method for the power system of a thermal power plant. The method includes: collecting multi-source operational data through data acquisition equipment deployed on the target thermal power plant's electrical equipment, and performing data correction processing based on a unified time reference to generate a standardized multi-source operational data set; based on the standardized multi-source operational data set, extracting features from the operating parameters of the target thermal power plant's electrical equipment to generate a standardized system state vector sequence, and establishing a dynamic state evolution model; based on the standardized system state vector sequence and the historical power consumption data of the target thermal power plant's electrical equipment, constructing a bidirectional sample set of power consumption and state, and establishing a multi-solution power consumption-driven state inversion model; and based on the multi-solution power consumption-driven state inversion model... Multiple candidate state solutions are generated and filtered within a multi-dimensional joint constraint space to construct a candidate set of minimum energy consumption reachable states under multiple operating conditions. Based on the multi-source operational disturbance factors of the target thermal power plant's electrical equipment, a disturbance structured mapping model is constructed to perform state extension calculations on the candidate set of minimum energy consumption reachable states, forming a disturbance-constrained feasible operating domain. A time-reachable graph topology is constructed based on the disturbance-constrained feasible operating domain to generate a continuous time-series candidate path set, and a path-level comprehensive energy consumption evaluation system is established to filter the globally optimal power consumption operating trajectory. Furthermore, based on the globally optimal power consumption operating trajectory, a cross-layer control parameter mapping system is established to decompose and obtain the control parameters of the target thermal power plant's electrical equipment, and a multi-equipment collaborative control command sequence is generated.

[0006] Optionally, the process of collecting multi-source operational data and performing data correction processing based on a unified time benchmark includes: dividing the operational data of the target thermal power plant's electrical equipment into three data sources: electrical data, thermal data, and environmental data; sampling electrical data at the second level, thermal data at the minute level, and environmental data at the time window level; reconstructing the timestamps of operational data with different sampling frequencies based on a unified time benchmark, and repairing missing data using an interpolation completion method; identifying and classifying abnormal data through a preset statistical feature analysis method to obtain a standardized multi-source operational data set.

[0007] Optionally, the step of extracting features from the operating parameters of the target thermal power plant's electrical equipment to generate a standardized system state vector sequence includes: extracting operating parameter features of the target thermal power plant's electrical equipment from a standardized multi-source operating data set, including equipment power level, equipment load rate, operating stability index, and operating efficiency index; fusing the operating parameter features to construct a system-level state vector; introducing coupling relationships between equipment into the system-level state vector and performing normalization processing to generate a standardized system state vector sequence.

[0008] Optionally, establishing the state dynamic evolution model includes: performing segmented pattern recognition on the standardized system state vector sequence and distinguishing and marking each type of pattern; quantitatively analyzing the transition logic and constraint path between different operating states based on the multi-mode vector sequence after the distinction and marking, and establishing a state evolution transition rule base; embedding multi-level physical constraints in the state evolution model so that all state vector sequences correspond to the equipment operating rules, thereby obtaining the state dynamic evolution model of the target thermal power plant's electrical equipment.

[0009] Optionally, the construction of the power consumption-state bidirectional sample set and the establishment of a multi-solution power consumption-driven state inversion model include: reconstructing the time series of historical power consumption data of the target thermal power plant's electrical equipment, using time series nodes of a preset time period as independent sample units, matching each sample unit with the corresponding power consumption value and system-level state vector to obtain bidirectional samples; performing hierarchical labeling of operating condition semantics on all bidirectional samples, and performing semantically enhanced structured encapsulation on each bidirectional sample to obtain the power consumption-state bidirectional sample set; using the power consumption-state bidirectional sample set as the input layer, the system-level state vector as the output layer, and embedding a multi-level physical constraint screening mechanism to construct a power consumption-driven state inversion mapping function system; introducing multiple target constraints into the power consumption-driven state inversion mapping function system to generate a multi-solution power consumption-driven state inversion model.

[0010] Optionally, the step of performing screening within a multi-dimensional joint constraint space to construct a candidate set of minimum energy consumption achievable states under multiple operating conditions includes: quantitatively modeling the equipment safety operation constraints, thermal system balance constraints, electrical system safety constraints, and external scheduling operation constraints of the target thermal power plant's electrical equipment, and integrating them into a multi-dimensional joint constraint space; within the multi-dimensional joint constraint space, establishing a standardized energy consumption assessment calculation function to calculate the target energy consumption index corresponding to each set of multiple candidate state solutions; sorting each set of target energy consumption indices in ascending order and performing screening according to a preset energy consumption threshold screening rule; introducing a state dynamic switchability evaluation index during the screening process, adding a switching feasibility verification to the screened low-energy consumption states, and generating a candidate set of minimum energy consumption achievable states under multiple operating conditions.

[0011] Optionally, the construction of the disturbance structured mapping model, which performs state extension calculations on the candidate set of minimum energy consumption reachable states and forms a disturbance-constrained feasible operating domain, includes: classifying the multi-source operational disturbance factors of the target thermal power plant's electrical equipment into three standardized categories and defining quantitative indicators, and building a disturbance structured mapping model; using each set of basic steady-state states within the candidate set of minimum energy consumption reachable states as a reference node, traversing all disturbance intervals, and carrying out hierarchical state extension for the three types of disturbances; embedding a global constraint propagation mechanism during the state extension process, combined with a global reachability filtering mechanism, to generate a disturbance extension state set; and after conducting dynamic stability judgment screening, state switching path smoothness verification, and equipment safety margin quantitative evaluation mechanisms on the disturbance extension state set, uniformly integrating and dividing the boundaries to obtain the disturbance-constrained feasible operating domain.

[0012] Optionally, the construction of the time reachability graph topology, generation of a continuous time-series candidate path set, and establishment of a path-level comprehensive energy consumption evaluation system to screen the globally optimal power consumption operating trajectory includes: extracting stable low-energy-consumption states within the perturbation-constrained feasible operating domain as independent nodes of the time reachability graph, and obtaining the time reachability graph topology according to preset timestamp binding rules; embedding hard constraints on time continuity to restrict time-series state jump behavior; performing multi-path structured combination operations based on the time reachability graph topology and hard constraints on time continuity to generate a set of multiple continuous time-series candidate paths; constructing a path full-cycle cumulative energy consumption evaluation model based on the continuous time-series candidate path set, and introducing a path perturbation anti-offset stability scoring mechanism, and using hierarchical screening logic to screen out the globally optimal power consumption operating trajectory.

[0013] Optionally, the energy-saving optimization method further includes: based on the deviation feedback between the operating status of the target thermal power plant's electrical equipment and the global optimal power consumption operating trajectory, rolling updates are made to the multi-solution power consumption driving state inversion model and the global optimal power consumption operating trajectory generation process to achieve closed-loop adaptive optimization control.

[0014] On the other hand, the present invention provides an energy-saving optimization system for a thermal power plant power system. The energy-saving optimization system includes a control module, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the energy-saving optimization method for a thermal power plant power system as described above.

[0015] Through the above technical solutions, this invention constructs a multi-solution power consumption-driven state inversion model, compatible with multiple feasible operating conditions corresponding to the same power consumption, and achieves the goal of parallel output of multiple energy-saving operation schemes, adapting to different peak-shaving operation needs of the unit and improving the flexibility of operating condition selection; through the minimum energy consumption reachable state topology network, it standardizes the operating condition switching migration path, adjustment rate and transition energy consumption constraints, achieving the goal of smooth and shock-free switching between different energy-saving states, avoiding the oscillation of unit thermal parameters caused by operating condition adjustment, and ensuring the safety of unit operation; by building a cross-layer control parameter mapping system and a multi-device collaborative control command sequence, it completes the precise decoupling and splitting of system trajectory to equipment commands, eliminates coupling interference between auxiliary machines, and achieves the dual goal of optimizing global system energy consumption and stable operation of individual equipment.

[0016] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the energy-saving optimization method for the power system of a thermal power plant according to the present invention. Figure 1 ; Figure 2 This is a schematic diagram of the process for generating a standardized multi-source operational data set in this invention; Figure 3 This is a schematic diagram of the process for generating a standardized system state vector sequence in this invention; Figure 4 This is a schematic diagram of the process for establishing a dynamic state evolution model in this invention; Figure 5 This is a schematic diagram of the process for establishing a multi-solution power consumption-driven state inversion model in this invention; Figure 6 This is a schematic diagram of the process for constructing a candidate set of minimum energy consumption achievable states under multiple operating conditions in this invention; Figure 7 This is a schematic diagram of the process for forming a feasible operating domain with perturbation constraints in this invention; Figure 8 This is a flowchart illustrating the process of selecting the globally optimal power consumption operating trajectory in this invention; Figure 9 This is a flowchart illustrating the energy-saving optimization method for the power system of a thermal power plant according to the present invention. Figure 2 . Detailed Implementation

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0019] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0020] Please refer to Figure 1 and Figure 9 This invention provides an energy-saving optimization method for a thermal power plant's power system, which may include: Step S100: Collect multi-source operation data by deploying data acquisition equipment on the electrical equipment of the target thermal power plant, and perform data correction processing based on a unified time reference to generate a standardized multi-source operation data set.

[0021] Please refer to Figure 2 In this embodiment of the invention, collecting multi-source operational data and performing data correction processing based on a unified time reference may include: Step S110: Divide the operating data of the target thermal power plant's electrical equipment into three data sources: electrical data, thermal data, and environmental data.

[0022] In a preferred embodiment of the present invention, at the data acquisition level, major auxiliary equipment such as induced draft fans, forced draft fans, primary air fans, feedwater pumps, and circulating water pumps can be used as core data acquisition targets. Simultaneously, electrical parameters such as voltage, current, active power, reactive power, power factor, and inverter operating frequency are collected. Furthermore, key operating parameters from both the boiler and turbine sides are included in the data acquisition scope, such as unit load, main steam pressure, reheat steam temperature, and fuel supply-related parameters. Additionally, continuous 24 / 7 acquisition of environmental boundary data is required, including seasonal operating parameters such as plant ambient dry-bulb temperature, circulating water inlet temperature, circulating water outlet temperature, ambient atmospheric pressure, seasonal humidity, and cooling water pool water level.

[0023] Step S120: Electrical data is sampled at the second level, thermal data is sampled at the minute level, and environmental data is sampled at the time window level.

[0024] Step S130: Based on a unified time reference, reconstruct the timestamps of the running data at different sampling frequencies, and use an interpolation completion method to repair the missing data.

[0025] In a preferred embodiment of the present invention, firstly, using the plant-level monitoring system time as the reference time axis, timestamp remapping is performed on all equipment data so that data from different sources can be mapped to a unified time series structure; secondly, data missing due to communication packet loss or equipment anomalies is filled in, and the filling method not only relies on simple interpolation methods but also needs to be constrained by the equipment operation logic; in addition, different data types are structurally reconstructed, and electrical quantities, thermal quantities, and environmental quantities are uniformly converted into standardized multidimensional data vectors so that the data at each time point can be represented as a system-level state input vector.

[0026] Step S140: Identify and classify abnormal data using a preset statistical feature analysis method to obtain a standardized multi-source operational data set.

[0027] In a preferred embodiment of the present invention, firstly, abnormal data is identified using a preset statistical feature analysis method, such as power surges, negative currents, and abnormal values ​​that do not conform to the physical constraints of equipment operation. Simultaneously, a sliding window analysis method is used to identify discontinuous fluctuations occurring within a short period. Secondly, the identified abnormal data is classified and processed. For instantaneous abnormal data, a smoothing correction method is used for replacement. For continuous abnormal data, logical judgments are made based on the equipment operating status, such as distinguishing between equipment shutdown and data abnormalities. Then, equipment operating constraint rules are introduced during the data correction process, such as auxiliary machine start / stop logic constraints and power change rate constraints. Finally, a standardized multi-source operating data set after anomaly correction is output.

[0028] Step S200: Based on the standardized multi-source operation data set, feature extraction is performed on the operating parameters of the target thermal power plant's electrical equipment to generate a standardized system state vector sequence and establish a dynamic state evolution model.

[0029] Please refer to Figure 3 In this embodiment of the invention, feature extraction of the operating parameters of the target thermal power plant's electrical equipment to generate a standardized system state vector sequence may include: Step S210: Extract the operating parameter characteristics of the target thermal power plant's electrical equipment from the standardized multi-source operating data set, including equipment power level, equipment load rate, operating stability index and operating efficiency index.

[0030] Step S220: The running parameter features are fused and processed to construct a system-level state vector.

[0031] In a preferred embodiment of the present invention, the system-level state vector not only includes the independent state information of each device, but also the correlation between devices, such as load distribution relationship, energy coupling relationship and operation coordination relationship; then, the overall operating parameters of the unit are incorporated into the state expression system, such as the unit load level, thermal system efficiency change, etc., so that the system state not only reflects the electrical side operation, but also reflects the thermal side operation characteristics.

[0032] Step S230: Introduce the coupling relationship between devices into the system-level state vector and perform normalization processing to generate a standardized system state vector sequence.

[0033] In a preferred embodiment of the present invention, all equipment state parameters can be arranged according to a unified structure to form a high-dimensional state vector. Each dimension of the vector corresponds to a specific operating characteristic, such as the power state of an auxiliary machine or the operating load state of a system. Structured relationship information is introduced into the state vector to express the coupling relationship between different devices. For example, when the load of a device changes, the states of other related devices will also change in a coordinated manner. This relationship is implicitly expressed through the internal structure of the state vector. Furthermore, the state vector is normalized so that data of different dimensions can be compared and calculated on a unified scale, ultimately forming a standardized system state vector sequence. This sequence is used to describe the changes in the operating state of the target thermal power plant's electrical equipment at different points in time.

[0034] Please refer to Figure 4 In this embodiment of the invention, establishing a dynamic state evolution model may include: Step S201: Perform segmented pattern recognition on the standardized system state vector sequence and distinguish and label each type of pattern.

[0035] In a preferred embodiment of the present invention, segmented pattern recognition can classify three basic operating modes based on the variation amplitude of unit load and total auxiliary power within a vector: stable steady-state operating mode, linear load increase / decrease adjustment mode, and sudden disturbance fluctuation mode. The stable steady-state mode is defined as a unit load fluctuation amplitude ≤ 5% of rated load for 30 consecutive minutes, with no significant adjustment in auxiliary load rate; the linear load increase / decrease mode is a continuous unidirectional increase / decrease in unit load, with auxiliary equipment synchronously adjusting its output; the sudden disturbance fluctuation mode includes operating conditions such as sudden changes in coal calorific value, sudden changes in ambient temperature, and start / stop switching of a single auxiliary machine. These three modes are automatically distinguished and marked using vector time-series characteristic variance and change slope, comprehensively identifying the evolution trend and inherent change law of different operating states of the system.

[0036] Step S202: Based on the multi-mode vector sequence with distinguishable markings, quantitatively analyze the transition logic and constraint paths between different operating states, and establish a state evolution transition rule base.

[0037] For example, the load increase path of the unit from 400MW to 600MW is as follows: the load of the primary air fan and the forced draft fan increases synchronously first, the induced draft fan adjusts lagging behind the change in furnace negative pressure, the feedwater pump flow rate increases synchronously, and the circulating water pump adjusts its output slightly according to the circulating water temperature. The state changes of the entire set of auxiliary equipment have a fixed sequence and amplitude matching relationship. The load reduction path of the unit presents the reverse adjustment logic, and the switching of different operating conditions corresponds to a unique state evolution path that conforms to the thermal process.

[0038] Step S203: Embed multi-level physical constraints in the state evolution model to make all state vector sequences correspond to the operating laws of the equipment, and obtain the state dynamic evolution model of the target thermal power plant's electrical equipment.

[0039] In a preferred embodiment of the present invention, for multi-level physical constraints, the first layer is the equipment response delay constraint: the frequency converter adjusts the auxiliary machine load with a response lag of 30~180 seconds, and instantaneous full-amplitude jumps are not allowed between adjacent state vector time sequences; the second layer is the load change rate constraint: the maximum load increase or decrease of the unit per minute does not exceed 3% of the rated capacity, and the corresponding auxiliary machine load rate is synchronously limited in its change range; the third layer is the thermal balance time sequence constraint: after the fuel quantity is adjusted, the main steam temperature and furnace negative pressure have a fixed delay response range, and the state evolution results of time sequence reversal are prohibited. Finally, a complete dynamic state evolution model of the target thermal power plant's electrical equipment is output. The model simultaneously includes three core contents: static cross-sectional state expression, time sequence dynamic transfer path, and multi-level physical time sequence constraints, which can accurately reproduce the continuous state change process of the auxiliary machine group under all operating conditions of the unit.

[0040] Step S300: Based on the standardized system state vector sequence and the historical power consumption data of the target thermal power plant's electrical equipment, construct a bidirectional sample set of power consumption-state, and establish a multi-solution power consumption-driven state inversion model.

[0041] Please refer to Figure 5 In this embodiment of the invention, constructing a bidirectional sample set of power consumption and state, and establishing a multi-solution power consumption-driven state inversion model, may include: Step S310: Reconstruct the time series of historical power consumption data of the target thermal power plant's electrical equipment, using the time series nodes of the preset time period as independent sample units, so that each sample unit is matched with the corresponding power consumption value and the system-level state vector to obtain bidirectional samples.

[0042] In a preferred embodiment of the present invention, high-precision time consistency secondary calibration can be performed on all historical time-series data. A preset time period (e.g., 1 minute) is used as an independent sample unit. Each sample is bound to two sets of core data within the same time segment: the first set is the complete system state vector (e.g., including all auxiliary machine characteristics, unit thermal conditions, and environmental disturbance parameters); the second set is the total plant power consumption and the cumulative power consumption of each subsystem's auxiliary equipment, ensuring that each set of power consumption samples uniquely matches the complete system state set within the corresponding time window, eliminating time-series misalignments and mismatched operating conditions. During modeling, power consumption is no longer simply defined as a state-derived result variable, but rather as a comprehensive characterization variable measuring the overall operational performance of the entire plant's auxiliary equipment group, participating with equal weight in the construction of bidirectional sample features.

[0043] Step S320: Perform semantic hierarchical labeling on all bidirectional samples and perform semantically enhanced structured encapsulation on each bidirectional sample to obtain a set of power consumption-state bidirectional samples.

[0044] In a preferred embodiment of the invention, the unit is divided into three basic operating condition subsets according to its core operating characteristics. The first layer is the stable load range sample: the unit load fluctuation range is ≤5% of the rated capacity, and there is no large-scale auxiliary machine adjustment operation. The second layer is the load dynamic adjustment range sample: the unit continuously increases or decreases the load, and the auxiliary machine adjusts its output synchronously. The third layer is the external disturbance recovery range sample: including transitional operating conditions such as sudden changes in coal quality, significant changes in ambient temperature, and the system returning to steady state after the start-up and shutdown of a single auxiliary machine. All samples are automatically divided into corresponding operating condition subsets according to their time series labels. Different subsets are stored independently and participate in modeling training independently. This avoids the chaotic combination of multiple operating conditions caused by the mixing of multiple operating conditions, which results in a large number of disordered state combinations with no physical logic corresponding to the same power consumption, and solves the inherent ambiguity problem of multiple solutions in reverse mapping.

[0045] In a preferred embodiment of the present invention, when semantically enhancing and structurally encapsulating each bidirectional sample, each sample, in addition to the two core fields of basic power consumption value and system state vector, is additionally bound with four categories of auxiliary background information: operating condition semantic label, load change trend slope, auxiliary machine start / stop mark, and thermal system constraint boundary, forming a "semantically enhanced power consumption-state bidirectional sample set".

[0046] Step S330: Use the power consumption-state bidirectional sample set as the input layer, the system-level state vector as the output layer, and embed a multi-level physical constraint screening mechanism to construct a power consumption-driven state inverse mapping function system.

[0047] In a preferred embodiment of the present invention, the input layer of the model only inputs the time-series cumulative plant power consumption and the power consumption of auxiliary equipment in the subsystems, and the output layer of the model is a complete high-dimensional system-level state vector. Unlike the traditional single-output supervised learning architecture, this mapping function does not force the solution of a unique optimal state, but retains all feasible state combinations that satisfy the physical boundaries, forming a multi-solution set mapping structure, which fits the objective operating characteristics of power plants in actual operation where, under the same total power consumption, the load distribution schemes of multiple auxiliary equipment are different but the energy consumption is the same.

[0048] In a preferred embodiment of the present invention, after embedding a multi-level physical constraint screening mechanism, invalid solutions that violate equipment, thermal, and electrical operating rules are automatically eliminated. The multi-level physical constraint screening mechanism includes equipment safety constraints (e.g., auxiliary machine load rate must not exceed 100% of rated load and must not be lower than the minimum stable combustion load), thermal balance constraints (e.g., total fuel quantity matches unit load, and flue gas volume ratio meets furnace negative pressure requirements), and electrical constraints (e.g., three-phase power balance, power factor range limitation). After constraint screening, only state combinations with practical engineering feasibility are retained, significantly reducing the computational workload of subsequent model solution set screening.

[0049] It should be noted that, considering the significant differences in the coupling patterns of auxiliary power consumption under different load ranges and disturbance conditions of the unit, a regional local mapping function fusion architecture is adopted: for the three major operating condition subsets of stable load, peak load adjustment and disturbance recovery, local inverse mapping sub-functions are trained independently, and different sub-functions have mapping weights adapted to their own operating conditions; during operation, the corresponding local mapping function is automatically matched according to the real-time operating condition semantic label, and then the complete state solution set is uniformly output through the global fusion layer, finally forming a power consumption-driven state inverse mapping function system, enabling the system to deduce the state space distribution from power consumption.

[0050] Step S340: Introduce multiple objective constraints into the power consumption-driven state inverse mapping function system to generate a multi-solution power consumption-driven state inverse model.

[0051] In a preferred embodiment of the present invention, the multiple objective constraints may include three dimensions: operational stability constraints, state change continuity constraints, and interpretability constraint mechanisms. First, operational stability constraints are introduced to perform an initial filtering of all candidate state solutions output by the model. Unstable state combinations that exhibit equipment overload, load below the minimum stable operating load, thermal parameters deviating from the safe range, or severe imbalance of three-phase electrical parameters are directly eliminated. For example, if the inverted solution set contains states that violate the equipment's safe operating range, such as a single feedwater pump operating at 110% load or an induced draft fan operating at a load below 20%, these states are directly marked as invalid and deleted to ensure that all output states meet the basic conditions for long-term continuous steady-state operation of the unit.

[0052] Secondly, a continuous constraint on state changes is introduced, and a smoothing limit is applied to the state output sequence corresponding to the continuous time-series power consumption input to eliminate non-physical state switching of large instantaneous jumps in a single time series. The constraint rule stipulates that the change amplitude of each dimension of the system state vector corresponding to adjacent minute time series shall not exceed the maximum adjustment limit of the equipment in 1 minute. If the model output continuous state shows a sudden result of the auxiliary machine load rate jumping from 10% to 90% in a short time, the solution set is automatically smoothed to ensure that the evolution process of the inverted time series state conforms to the adjustment response speed of the equipment.

[0053] Furthermore, an interpretability constraint mechanism is introduced to bind a contribution traceability label to each set of inversion output states, realizing the quantitative decomposition of the sources of power consumption changes and the power consumption contribution of each auxiliary machine. The mechanism has a built-in feature contribution calculation module. When the total plant power consumption increases or decreases, it automatically quantifies and calculates the contribution ratio of the output adjustment of each type of auxiliary machine, such as induced draft fan, feed water pump, and circulating water pump, to the overall energy consumption change. It also displays the energy consumption offset caused by the disturbance of ambient temperature and coal calorific value. The black box mapping is decomposed into quantifiable and traceable sub-contribution indicators, and finally a stable, time-series smooth, physically and logically self-consistent, and quantitatively interpretable multi-solution power consumption driven state inversion model is formed.

[0054] Step S400: Generate multiple sets of candidate state solutions based on the multi-solution power consumption driven state inversion model, perform screening processing in the multi-dimensional joint constraint space, and construct a candidate set of minimum energy consumption achievable states under multiple operating conditions.

[0055] Please refer to Figure 6 In this embodiment of the invention, a screening process is performed within a multi-dimensional joint constraint space to construct a candidate set of minimum energy consumption achievable states under multiple operating conditions, which may include: Step S410: Quantitatively model the equipment safety operation constraints, thermal system balance constraints, electrical system safety constraints, and external dispatch operation constraints of the target thermal power plant's electrical equipment, and integrate them into a multi-dimensional joint constraint space.

[0056] In a preferred embodiment of the present invention, when modeling the constraints for safe operation of equipment, the mechanical and electrical operating limits of all auxiliary machines are covered, including detailed rules such as the maximum rated load rate of a single auxiliary machine, the minimum stable operating load threshold, the upper limit of motor winding temperature, the bearing vibration safety range, the upper and lower limits of inverter frequency adjustment, the upper limit of the number of daily start-ups and shutdowns of the equipment, and the continuous operating time limit. All constraints are quantified into numerical ranges and written into the constraint space. If any auxiliary machine parameter in any state vector exceeds the range, it is determined to be an unreachable state.

[0057] When modeling the thermal system balance constraints, it is ensured that the selected state meets the boiler-turbine energy matching production logic. The constraint rules include the matching range of the total fuel quantity of the unit and the power generation load, the allowable fluctuation range of the main steam / reheat steam pressure and temperature, the safe range of furnace negative pressure, the air preheater differential pressure limit, the coupling relationship between feedwater flow and unit load, and the cooling water temperature constraint corresponding to the circulating water flow.

[0058] When modeling electrical system safety constraints, the operating boundaries of the plant's auxiliary power bus are controlled. The constraints include electrical safety rules such as the upper and lower limits of the auxiliary power bus voltage, the threshold of three-phase current imbalance, the reactive power compensation range of the whole plant, the power factor control range, and the upper limit of the load of the auxiliary power transformer, so as to avoid the power quality of the auxiliary power system exceeding the standard under low energy consumption conditions.

[0059] When modeling external dispatching operation constraints, the modeling process matches the hard requirements of the power grid's power generation load instructions. The core constraint is that the real-time power generation load of the generating units must closely match the load range given by the power grid dispatching, allowing for small normal fluctuations, but prohibiting the selection of energy-saving states where the load deviates significantly from the dispatching instructions, thus ensuring the fulfillment of the power plant's power generation tasks. Finally, a unified constraint fusion calculation framework is established to integrate single-device local constraints, thermal system coupling constraints, electrical safety constraints, and global dispatching constraints into a multi-dimensional joint constraint space.

[0060] Step S420: Within the multidimensional joint constraint space, construct a standardized energy consumption assessment calculation function to calculate the target energy consumption index corresponding to each set of multiple candidate state solutions.

[0061] In a preferred embodiment of the present invention, when building a standardized energy consumption assessment calculation function, for each set of system-level state vectors that have passed the constraint verification, the corresponding steady-state total plant power consumption and power supply auxiliary power rate are accurately calculated based on the load rate of all auxiliary machines and the output adjustment parameters within the vector. The power consumption of the flue gas system, water supply system and desulfurization system are calculated simultaneously to form a complete energy consumption assessment index set corresponding to each state.

[0062] Step S430: Sort each group of target energy consumption indicators in ascending order and filter them according to the preset energy consumption threshold filtering rules.

[0063] In a preferred embodiment of the present invention, after calculating the energy consumption values ​​of all candidate states, the states are sorted in ascending order from low to high according to the total power consumption of the entire plant. An energy consumption threshold screening rule is set: based on the theoretical minimum energy consumption value under the current operating conditions, all system states with total energy consumption 0-3% higher than the benchmark value are retained, and high energy consumption states with high energy consumption and weak energy-saving benefits are eliminated. This screening rule can ensure that all states in the set are near the theoretical energy-saving boundary of the current operating conditions of the unit and have significant energy-saving optimization value.

[0064] Step S440: During the screening process, a dynamic switchability evaluation index is introduced to perform a switchability verification on the screened low-energy-consumption states and generate a candidate set of the lowest energy-consumption achievable states under multiple operating conditions.

[0065] In a preferred embodiment of the present invention, the state dynamic switchability evaluation index quantifies the auxiliary machine adjustment range, adjustment time, and additional transition energy consumption required to complete the adjustment and switching between two sets of states. Only low-energy-consumption states with low switching costs and smooth adjustment processes without exceeding the limit risks are retained, while extremely optimal states that, although with extremely low steady-state energy consumption, require significant adjustment of auxiliary machine output and are prone to thermal fluctuations when switching from the current operating condition to that state are eliminated. Finally, a candidate set of multiple minimum energy-consumption achievable states is formed.

[0066] In a preferred embodiment of the present invention, all low-energy-consumption states within the set can be topologically encapsulated to construct a network topology with state transition relationships. First, each steady-state operating state within the set of lowest-energy-consumption achievable states is defined as an independent node in the topology network. Each node is bound to four basic attributes: a complete system state vector, steady-state energy consumption indicators, equipment constraint boundaries, and switching costs. Node attributes are stored fixed throughout the process, facilitating rapid retrieval of energy consumption and constraint-related parameters during subsequent path planning. Second, directed migration edges are constructed between nodes, representing executable switching paths between two low-energy-consumption states. Each migration edge is equipped with complete switching constraint rules and transition loss parameters: the constraint rules include the upper limit of the auxiliary machine output adjustment rate during the switching process, the allowable range of thermal parameter fluctuations, and the shortest switching time; the transition loss parameters quantify the additional short-term power consumption generated by the dynamic adjustment of auxiliary machines during state switching, distinguishing between steady-state energy consumption and total energy consumption during dynamic switching, thus conforming to the actual energy consumption change patterns during the unit's adjustment process.

[0067] In addition, a migration edge determination logic is established. A valid migration edge is generated only when the entire process of switching between two states meets the triple constraints of equipment response delay, thermal stability, and no electrical over-limit. If the switching between two states requires simultaneous large-scale adjustment of the output of multiple auxiliary machines, which may easily cause furnace negative pressure and drastic fluctuations in main steam temperature, a direct migration edge is not generated. Instead, indirect switching through intermediate transition states is allowed to avoid the risk of dynamic adjustment. Finally, a low-energy-consumption state topology network is formed.

[0068] Step S500: Based on the multi-source operational disturbance factors of the target thermal power plant's electrical equipment, construct a disturbance structured mapping model, perform state extension calculations on the candidate set of minimum energy consumption achievable states, and form a disturbance-constrained feasible operating domain.

[0069] Please refer to Figure 7 In this embodiment of the invention, a perturbation-structured mapping model is constructed to perform state expansion calculations on the candidate set of minimum energy consumption reachable states, forming a perturbation-constrained feasible operating domain, which may include: Step S510: Divide the multi-source operational disturbance factors of the target thermal power plant's electrical equipment into three standardized categories and define quantitative indicators, and build a disturbance structured mapping model.

[0070] In a preferred embodiment of the present invention, the three standardized categories may include load-related disturbances, equipment-related disturbances, and environmental disturbances. Load-related disturbances originate from grid dispatch command adjustments and regional power load fluctuations, and are further subdivided into two types: step-type load disturbances (e.g., grid commands for short-term large increases or decreases in load) and continuous linear load disturbances (e.g., continuous increases or decreases in load during stable peak shaving by generating units). The quantified disturbance indicators are the magnitude of unit load changes, the rate of load change, and the duration of load shift. This type of disturbance will simultaneously drive the synchronous adjustment of the output of all flue gas and feedwater auxiliary equipment, and is the most critical external disturbance affecting the system state.

[0071] Equipment-related disturbances originate from long-term wear and tear of auxiliary equipment, decreased motor efficiency, and ash accumulation and blockage in heat exchangers, resulting in equipment performance degradation. Specific indicators include the percentage decrease in operating efficiency of a single auxiliary machine, the increase in resistance of fan / pump pipelines, the increment of frequency converter adjustment response delay, and the vibration offset of equipment bearings. This type of disturbance will cause a continuous increase in auxiliary machine power consumption under the same output conditions, and the fixed optimal load distribution state will gradually shift as the equipment deteriorates, belonging to long-term slow internal disturbances.

[0072] The sources of environmental disturbances are seasonal changes, diurnal temperature differences, circulating water temperature changes, and atmospheric humidity fluctuations. The quantitative disturbance indicators are the ambient temperature deviation, the range of circulating water inlet temperature changes, and atmospheric pressure fluctuations. These disturbances directly affect the operating energy consumption of circulating water pumps and desulfurization cooling systems. Under the same unit load, the optimal auxiliary equipment allocation state continuously shifts with changes in environmental parameters.

[0073] It should be noted that after completing the classification and quantification index definition of the three types of disturbances, a "disturbance-state influence mapping structure" is constructed to describe the influence of disturbances on the direction and magnitude of state shifts. For example, if the ambient temperature rises by 5°C, the optimal load rate of the circulating water pump increases by 4% to 7% simultaneously, the exhaust resistance of the induced draft fan increases slightly, and the load rate shifts by about 2% simultaneously. The mapping structure fully quantifies the influence of various disturbances on the shift of the low-energy-consumption operating state.

[0074] Step S520: Using each set of basic steady-state states in the candidate set of minimum energy consumption reachable states as the reference node, traverse all disturbance intervals and carry out hierarchical state expansion in three types of disturbances.

[0075] In a preferred embodiment of the present invention, the extension logic takes each set of basic steady-state states within the candidate set of minimum energy consumption achievable states as the core benchmark node, traverses all disturbance intervals of load type, equipment type, and environment type, and completes the state vector numerical offset according to the offset direction and offset magnitude defined by the disturbance mapping model. It carries out layered extension in three types of disturbances: single disturbance extension (e.g., only load / equipment / environment parameters change, and other operating conditions remain constant), dual disturbance coupling extension (e.g., load fluctuation superimposed with ambient temperature change), and full disturbance cascade extension (e.g., unit peak shaving, equipment efficiency decay, and seasonal temperature change occur simultaneously), which fully covers all disturbance coupling scenarios in the daily operation of power plants and generates a massive number of disturbance offset derived states.

[0076] Step S530: During the state expansion process, a global constraint propagation mechanism is embedded, combined with a global reachability filtering mechanism, to generate a perturbation expansion state set.

[0077] In a preferred embodiment of the present invention, when a parameter in a certain dimension shifts due to a disturbance, the constraint boundaries of the associated parameters are automatically and synchronously updated to prevent the deviation of a single parameter from causing the associated thermal and electrical parameters to exceed the operating limits. For example, if the unit load increases by 10% (e.g., load disturbance expansion), the allowable load range of the feedwater pump and fan is automatically adjusted upwards, and the constraint thresholds of the main steam flow and fuel quantity are simultaneously corrected. This achieves synchronous and linked updates of the constraints across the entire system when the disturbance expands, preventing logical contradictions such as local parameter compliance and overall thermal imbalance in the expanded state.

[0078] After all extended and derived states are generated, a global reachability filtering mechanism is activated to distinguish between two judgment dimensions: static constraint compliance and dynamic operating condition reachability. Static verification only checks whether the state vector values ​​meet the equipment limits. Dynamic reachability verification simulates whether the current baseline state can smoothly transition to the derived state through the equipment adjustment path after disturbance offset. It filters out virtual states that are theoretically compliant but cannot be actually reached due to equipment adjustment rate and response delay limitations. Finally, a complete set of disturbance extended states is generated, which covers the entire range of disturbance fluctuations under all operating conditions.

[0079] Step S540: After conducting dynamic stability judgment screening, state transition path smoothness verification and equipment safety margin quantitative evaluation mechanism on the disturbance extension state set, the boundary is uniformly integrated and divided to obtain the disturbance constraint feasible operating domain.

[0080] In a preferred embodiment of the present invention, the first layer performs dynamic stability judgment and screening, and establishes a time-series dynamic evolution verification mechanism. Each set of disturbance extension and derivative states is input into the state dynamic evolution model to simulate the time-series state change trend for 60 consecutive minutes under the continuous action of the disturbance. If unstable phenomena such as continuous drift of auxiliary machine load, oscillation and divergence of thermal parameters, and continuous increase of energy consumption occur during the simulation, it is directly determined that the state has insufficient anti-disturbance capability and is eliminated; only the highly stable state that can maintain a stable steady state without continuous offset oscillation under the continuous action of the disturbance is retained.

[0081] The second layer performs state switching path smoothness verification. Based on the low-energy-consumption state topology network, it matches all associated basic low-energy-consumption states for each set of disturbance-derived states, traverses all bidirectional switching migration paths, and verifies whether there are dynamic risks such as auxiliary machine instantaneous overload, large fluctuations in thermal parameters, and plant power voltage exceeding limits throughout the entire path. It eliminates derivative states that are prone to severe operating condition shocks during the switching process and retains feasible states with smooth switching paths and low transition losses.

[0082] The third layer introduces a quantitative evaluation mechanism for equipment safety margin. It calculates the safe distance between the operating parameters of all auxiliary machines in each disturbance-induced state and the rated upper limit and the minimum stable power generation lower limit, and converts them into a unified safety margin score of 0 to 1. It sets a minimum safety margin threshold, eliminates high-risk states with scores below the threshold or close to the equipment's operating limit boundary, and reserves sufficient adjustment buffer space to cope with sudden disturbance increments, so as to avoid the equipment touching the safety red line during unit operation.

[0083] Step S600: Construct a time reachability graph topology based on the feasible operating domain under perturbation constraints, generate a set of continuous time-series candidate paths, and build a path-level comprehensive energy consumption evaluation system to select the globally optimal power consumption operating trajectory.

[0084] Please refer to Figure 8 In this embodiment of the invention, constructing a time-reachable graph topology, generating a set of continuous time-series candidate paths, and establishing a path-level comprehensive energy consumption evaluation system to screen the globally optimal power consumption trajectory may include: Step S610: Extract stable low-energy states within the feasible operating domain under perturbation constraints as independent nodes of the time reachability graph, and obtain the topology of the time reachability graph according to the preset timestamp binding rules.

[0085] In a preferred embodiment of the present invention, all stable low-energy states within the perturbation-constrained feasible operating domain can be extracted as independent nodes of the time-reachable graph. The basic attributes of nodes, directed migration edges between nodes, and switching transition loss parameters in the topology network are reused. Additional time-series timestamp binding rules are added, and each migration edge is accompanied by a minimum switching duration and a time-series smoothing constraint threshold, thereby realizing the expansion and upgrade of the static state network to a time-series reachable graph.

[0086] Step S620: Embed hard constraints on time continuity to restrict the timing state transition behavior.

[0087] In a preferred embodiment of the present invention, the system state evolution can be forced to follow the principle of gradual and smooth evolution, prohibiting large state jumps across intervals without transition. The constraint details stipulate that, for example, the load rate change of each auxiliary machine in the system state vector corresponding to adjacent 1-minute time series shall not exceed 5% of the rated load per minute; state switching across multiple time series must be inserted with intermediate transition states to gradually adjust the output of auxiliary machines; if a candidate path shows a sudden and large state jump, the path is directly judged as illegal and marked as invalid, ensuring from the bottom layer that the generated time series path conforms to the actual adjustment response characteristics of the auxiliary machines.

[0088] Step S630: Based on the topology of the time reachability graph and the hard constraint of time continuity, perform multi-path structured combination operation to generate a set of multiple continuous temporal candidate paths.

[0089] In a preferred embodiment of the present invention, multi-path structured combination operations are performed based on time reachability graphs and temporal continuity constraints. Multiple complete temporally continuous candidate path sequences are generated in batches according to the typical operating sequence windows of the unit (e.g., a 24-hour daily operating cycle and a 4-hour peak-shaving cycle). Each candidate path consists of dozens to hundreds of temporally connected state nodes in an orderly manner, fully covering the entire process of unit load changes and environmental disturbance fluctuations within the corresponding time window. After the path is generated, a segment-by-segment physical consistency verification is initiated. For each state transition process within the path, the triple constraints of thermal balance, electrical safety, and equipment regulation rate are verified in sequence. Violation paths with parameter over-limits or thermal imbalances during the switching process are eliminated, and only legal candidate paths with no conflicts throughout, smooth temporal sequence, and reasonable regulation logic are retained, generating a set of multiple continuous temporally continuous candidate paths.

[0090] Step S640: Based on the continuous time-series candidate path set, construct a path full-cycle cumulative energy consumption evaluation model, introduce a path disturbance anti-offset stability scoring mechanism, and use hierarchical screening logic to select the globally optimal power consumption operation trajectory.

[0091] In a preferred embodiment of the present invention, when building a path full-cycle cumulative energy consumption evaluation model, the steady-state power consumption of all time-series nodes within the complete time window of each candidate path is summed to obtain the basic steady-state total energy consumption of the path; then, all state switching transition segments within the path are traversed, and the auxiliary machine adjustment additional transition energy consumption generated during each switching process is accumulated. The steady-state energy consumption and switching loss are added together to obtain the complete cumulative total energy consumption of the path, which is used as the core evaluation index of the path's energy-saving effect, accurately distinguishing between suboptimal paths with low steady-state energy consumption but frequent switching and high transition losses, and high-quality paths with low energy consumption throughout and few large-scale adjustments.

[0092] Secondly, a path disturbance resistance stability scoring mechanism can be introduced to quantitatively evaluate the operational deviation of each candidate path under small fluctuations in load, environment, and equipment disturbances. The score is calculated based on the safety margin of each time-series state within the path, the state transition buffer zone, and the disturbance tolerance range, with a maximum score of 100 points. The higher the score, the stronger the path's ability to cope with sudden disturbances and the less likely it is to deviate from the low-energy consumption range during operation. Under the premise that the cumulative total energy consumption of multiple paths is similar, the time-series path with the higher stability score is selected first to avoid vulnerable paths that have the best short-term energy consumption but are extremely prone to instability under disturbances.

[0093] Finally, considering the two core evaluation dimensions of cumulative total energy consumption and disturbance stability score, a hierarchical sorting and screening logic is adopted: The first layer sorts the paths in ascending order of cumulative total energy consumption, and selects the top 10% of low-energy-consumption candidate paths; the second layer compares the stability scores within the subset, and selects the time-series path with the lowest overall energy consumption and the best disturbance resistance as the globally optimal power consumption operating trajectory; if there are multiple paths with energy consumption differences of less than 0.5% and similar stability scores, multiple near-globally optimal power consumption operating trajectories are simultaneously retained for maintenance personnel to flexibly choose according to the on-site equipment conditions.

[0094] It should be noted that during the actual operation of thermal power plants, the grid load command, ambient temperature, and equipment efficiency degradation are in a state of real-time dynamic change. The initial global optimal power consumption operation trajectory generated by offline planning is only suitable for historical predicted disturbance conditions. When the actual on-site operating conditions deviate from the predicted values, the energy-saving effect of the trajectory will continue to decay. Therefore, a closed-loop mechanism of real-time deviation monitoring, local trajectory correction, and rolling optimization and updating can be established.

[0095] Step S700: Based on the globally optimal power consumption operation trajectory, establish a cross-layer control parameter mapping system, decompose the control parameters of the target thermal power plant's electrical equipment, and generate a multi-equipment collaborative control command sequence.

[0096] In a preferred embodiment of the present invention, during the conversion of the optimal power consumption operating trajectory into equipment control commands, a cross-layer control parameter mapping system is first constructed. This system decomposes system-level state variables into multiple independent but interconnected control subspaces. These subspaces are reconstructed according to the "energy flow path consistency principle," ensuring that each type of control subspace corresponds not only to a single device but also to the energy coupling relationship between devices.

[0097] Based on this structure, a control parameter decoupling mechanism is introduced to decompose system state changes into power regulation components, load distribution components, and dynamic response components, and map them to different device control variables respectively, thereby avoiding the control conflict problem that occurs in the traditional direct mapping method. In the cross-layer mapping process, a device response delay modeling mechanism is introduced to establish time response constraints for each type of control variable, and finally, a control parameter mapping system is formed.

[0098] In a preferred embodiment of the present invention, after the control parameter mapping is completed, the auxiliary equipment group collaborative control command generation stage begins. First, the coupling relationship of the control commands for all auxiliary equipment is modeled, identifying the control dependencies between different equipment, such as load sharing and energy compensation relationships, and transforming them into a constraint graph structure. A control conflict detection mechanism is introduced; when the control adjustment of one equipment causes an energy imbalance in another equipment, the system automatically triggers a conflict flag and enters a collaborative redistribution process. Subsequently, a consistency reconstruction algorithm is introduced to redistribute conflicting control commands, rebalancing all equipment control commands under the global energy consumption target constraint, thereby forming a consistent control solution, and finally generating a multi-equipment collaborative control command sequence.

[0099] After the coordinated control commands are generated, all control commands undergo hierarchical executability verification. This verification system consists of three levels: equipment-level constraints, system-level constraints, and time-level constraints. Equipment-level constraints verify whether individual devices meet operating range limitations, including power limits, minimum stable operating load, and start / stop frequency constraints. System-level constraints verify the overall energy balance between multiple devices, such as whether the overall load demand of the unit matches the energy consumption of auxiliary equipment. Time-level constraints verify the executability of control commands in the time sequence, including command execution order, response delay matching, and state continuity requirements. When constraint conflicts are detected, a hierarchical correction mechanism is introduced, making local adjustments only to the conflicting layer, ultimately forming a set of executable control commands that satisfy the multi-level constraint system.

[0100] Step S800: Based on the feedback of the deviation between the operating status of the target thermal power plant's electrical equipment and the global optimal power consumption operating trajectory, the multi-solution power consumption driving state inversion model and the global optimal power consumption operating trajectory generation process are continuously updated to achieve closed-loop adaptive optimization control.

[0101] In a preferred embodiment of the present invention, during system operation, real-time deviation detection is performed between the actual operating state and the optimal operating trajectory. This deviation detection constructs a multi-dimensional deviation structure, including state structure deviation, time deviation, and energy consumption deviation. The state structure deviation is used to measure the degree of deviation from the equipment combination method; the time deviation is used to measure the difference in the state evolution rhythm; and the energy consumption deviation is used to measure the difference in overall operating efficiency.

[0102] When the system exhibits persistent deviations, the power consumption-driven state inversion model is adaptively updated. The update process employs a recursive adjustment mechanism, allowing the model parameters to be gradually corrected without disrupting the original structure. During the update process, a stability constraint mechanism is introduced to prevent the model from experiencing overall structural oscillations due to local data fluctuations; simultaneously, a memory retention mechanism is introduced to allow the model to retain historical operating patterns during the update process, forming an adaptive power consumption-driven state inversion model.

[0103] After the model is updated, the optimal operating trajectory is reconstructed in a rolling manner. This reconstruction mechanism is based on time window division, breaking down the overall optimization problem into multiple local time periods, and recalculating the local optimum within each time period. At the same time, continuity constraints between windows are introduced to ensure a smooth transition between trajectories of different time windows and avoid abrupt changes. When a significant change in the system state is detected, only the affected time window is locally updated, ultimately forming a continuously adaptively updated optimal power consumption operating trajectory.

[0104] On the other hand, embodiments of the present invention also provide an energy-saving optimization system for a thermal power plant power system. The energy-saving optimization system includes a control module, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the energy-saving optimization method for a thermal power plant power system described above.

[0105] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0106] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0108] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0109] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0110] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0111] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0112] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0113] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for optimizing energy conservation in the power supply system of a thermal power plant, characterized in that, The energy-saving optimization method includes: By deploying data acquisition equipment on the electrical equipment of the target thermal power plant, multi-source operation data is collected, and data correction processing is performed based on a unified time benchmark to generate a standardized multi-source operation data set. Based on the standardized multi-source operation data set, feature extraction is performed on the operating parameters of the target thermal power plant's electrical equipment to generate a standardized system state vector sequence and establish a dynamic state evolution model. Based on the standardized system state vector sequence and the historical power consumption data of the target thermal power plant's electrical equipment, a bidirectional sample set of power consumption-state is constructed, and a multi-solution power consumption-driven state inversion model is established. Based on the multi-solution power consumption driven state inversion model, multiple sets of candidate state solutions are generated, and the screening process is carried out in the multi-dimensional joint constraint space to construct the candidate set of the lowest energy consumption reachable state under multiple operating conditions. Based on the multi-source operational disturbance factors of the electrical equipment in the target thermal power plant, a disturbance structured mapping model is constructed, and the state extension calculation is performed on the candidate set of minimum energy consumption reachable states to form a disturbance-constrained feasible operating domain. Based on the perturbation-constrained feasible operating domain, a time-reachable graph topology is constructed, a continuous time-series candidate path set is generated, and a path-level comprehensive energy consumption evaluation system is established to screen the globally optimal power consumption operating trajectory; and Based on the globally optimal power consumption operation trajectory, a cross-layer control parameter mapping system is established to decompose the control parameters of the target thermal power plant's electrical equipment and generate a multi-equipment collaborative control command sequence.

2. The energy-saving optimization method for power systems in thermal power plants according to claim 1, characterized in that, The process of collecting multi-source operational data and performing data correction processing based on a unified time benchmark includes: The operating data of the target thermal power plant's electrical equipment are divided into three data sources: electrical data, thermal data, and environmental data. Electrical data is sampled at the second level, thermal data at the minute level, and environmental data at the time window level. Based on a unified time base, timestamps are reconstructed for running data with different sampling frequencies, and interpolation completion methods are used to repair missing data. By identifying and classifying abnormal data through a pre-defined statistical feature analysis method, a standardized multi-source operational data set is obtained.

3. The energy-saving optimization method for power systems in thermal power plants according to claim 1, characterized in that, The step of extracting features from the operating parameters of the target thermal power plant's electrical equipment to generate a standardized system state vector sequence includes: Extract the operating parameter characteristics of the target thermal power plant's electrical equipment from a standardized multi-source operating data set, including equipment power level, equipment load rate, operating stability index, and operating efficiency index. The operational parameter features are fused and processed to construct a system-level state vector; The coupling relationship between devices is introduced into the system-level state vector and normalized to generate a standardized system state vector sequence.

4. The energy-saving optimization method for power systems in thermal power plants according to claim 3, characterized in that, The establishment of the state dynamic evolution model includes: The standardized system state vector sequence is subjected to segmented pattern recognition, and each pattern is distinguished and labeled. Based on the multi-mode vector sequences with distinguishable markings, the transition logic and constraint paths between different operating states are quantitatively analyzed, and a state evolution transition rule base is established. By embedding multi-level physical constraints into the state evolution model, all state vector sequences correspond to the operating laws of the equipment, thus obtaining the dynamic state evolution model of the target thermal power plant's electrical equipment.

5. The energy-saving optimization method for power systems in thermal power plants according to claim 1, characterized in that, The construction of the power consumption-state bidirectional sample set and the establishment of a multi-solution power consumption-driven state inversion model include: The historical power consumption data of the target thermal power plant's electrical equipment is reconstructed into a time series. The time series nodes of the preset time period are used as independent sample units, so that each sample unit is matched with the corresponding power consumption value and the system-level state vector to obtain bidirectional samples. All bidirectional samples are labeled with hierarchical semantic tags based on operating conditions, and each bidirectional sample is encapsulated with semantic enhancement structure to obtain a set of power consumption-state bidirectional samples. The system constructs a power consumption-state bidirectional sample set as the input layer, the system-level state vector as the output layer, and embeds a multi-level physical constraint screening mechanism to build a power consumption-driven state inverse mapping function system. Multiple objective constraints are introduced into the power consumption-driven state inverse mapping function system to generate a multi-solution power consumption-driven state inverse model.

6. The energy-saving optimization method for power systems in thermal power plants according to claim 1, characterized in that, The process of filtering within a multi-dimensional joint constraint space to construct a candidate set of minimum energy consumption achievable states under multiple operating conditions includes: The equipment safety operation constraints, thermal system balance constraints, electrical system safety constraints, and external dispatch operation constraints of the target thermal power plant's electrical equipment are quantitatively modeled and integrated into a multi-dimensional joint constraint space. Within a multidimensional joint constraint space, a standardized energy consumption assessment calculation function is constructed to calculate the target energy consumption index corresponding to each set of multiple candidate state solutions. The target energy consumption indicators for each group are sorted in ascending order and then filtered according to the preset energy consumption threshold filtering rules. During the screening process, a dynamic switchability evaluation index is introduced to verify the feasibility of switching the selected low-energy-consumption states, thereby generating a candidate set of states with the lowest achievable energy consumption under multiple operating conditions.

7. The energy-saving optimization method for power systems in thermal power plants according to claim 1, characterized in that, The construction of the perturbation-structured mapping model involves performing state expansion calculations on the candidate set of minimum energy consumption achievable states and forming a perturbation-constrained feasible operating domain, including: The multi-source operational disturbance factors of the target thermal power plant's electrical equipment are classified into three standardized categories and quantitative indicators are defined to build a disturbance structured mapping model. Using each set of basic steady-state states in the candidate set of minimum energy consumption reachable states as the reference node, we traverse all disturbance intervals and carry out hierarchical state expansion in three types of disturbances. A global constraint propagation mechanism is embedded in the state expansion process, combined with a global reachability filtering mechanism, to generate a perturbation expansion state set; After conducting dynamic stability judgment screening, state transition path smoothness verification and equipment safety margin quantitative evaluation mechanism on the disturbance extension state set, the boundary is uniformly integrated and divided to obtain the disturbance constraint feasible operating domain.

8. The energy-saving optimization method for power systems in thermal power plants according to claim 1, characterized in that, The construction of the reachable graph topology generates a set of continuous time-series candidate paths, and establishes a path-level comprehensive energy consumption evaluation system to screen the globally optimal power consumption operating trajectory, including: Stable, low-energy-consumption states within the feasible operating domain under perturbation constraints are extracted as independent nodes in the time reachability graph, and the topology of the time reachability graph is obtained according to the preset timestamp binding rules. Embed hard constraints on temporal continuity to limit the timing state transition behavior; Based on the topology of the time reachability graph and the hard constraint of time continuity, a multi-path structured combination operation is performed to generate a set of multiple continuous temporal candidate paths. Based on the continuous time-series candidate path set, a path full-cycle cumulative energy consumption assessment model is constructed, and a path disturbance anti-offset stability scoring mechanism is introduced. The globally optimal power consumption operation trajectory is selected using hierarchical screening logic.

9. The energy-saving optimization method for power systems in thermal power plants according to claim 1, characterized in that, The energy-saving optimization method also includes: Based on the feedback of the deviation between the operating status of the target thermal power plant's electrical equipment and the global optimal power consumption trajectory, the multi-solution power consumption driving state inversion model and the global optimal power consumption trajectory generation process are continuously updated to achieve closed-loop adaptive optimization control.

10. An energy-saving optimization system for the power supply system of a thermal power plant, characterized in that, The energy-saving optimization system includes a control module, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the energy-saving optimization method for the power system of a thermal power plant according to any one of claims 1-9.

Images