Thermal power plant scheduling method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a thermal power plant scheduling method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a sample information set of a thermal power plant; performing information division on multiple pieces of sample information in the sample information set based on preset parameter association information to obtain multiple working condition clusters; each working condition cluster comprises at least one part of sample information; for each working condition cluster, determining a cluster center of the working condition cluster; wherein each working condition cluster comprises a plurality of intra-cluster samples; determining the in-cluster sample with the minimum weighted distance to the cluster center from the in-cluster samples as a target sample; and performing constraint evaluation on each target sample, and using the target sample with the evaluation result meeting the evaluation condition to control the work of the thermal power plant. The method can improve the scheduling flexibility of the thermal power plant.

Description

Technical Field

[0001] This application relates to the field of intelligent dispatching technology, and in particular to a dispatching method, apparatus, computer equipment, computer-readable storage medium, and computer program product for thermal power plants. Background Technology

[0002] With the continuous development of power industry technology, thermal power plants, as the core link in energy supply, are crucial to the overall operation of the energy system due to their dispatch efficiency and stability. Currently, thermal power plant dispatch technology has gradually introduced data-driven methods, optimizing dispatch strategies by analyzing historical operating data. Traditional dispatch methods mainly rely on manual experience or models based on fixed rules. For example, by setting fixed load allocation thresholds or simple operating condition classification rules, thermal power plant operating data is divided into several fixed operating condition intervals, and then a dispatch plan is formulated for each interval.

[0003] However, the actual operating conditions of thermal power plants are complex and changeable, and are dynamically affected by multiple factors such as fuel quality, equipment status, and environmental conditions. Traditional methods are difficult to adapt to real-time changes due to their coarse division of operating conditions and rigid scheduling strategies. Furthermore, traditional methods lack in-depth mining of the inherent correlations in sample data and cannot flexibly adjust scheduling parameters to adapt to the optimization needs of different operating conditions. Therefore, traditional thermal power plant scheduling methods have the problem of poor flexibility. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for dispatching thermal power plants that can improve the dispatching flexibility of thermal power plants, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a method for dispatching a thermal power plant, including:

[0006] Obtain a sample information set of thermal power plants;

[0007] Based on pre-set parameter association information, multiple sample information in the sample information set are divided to obtain multiple working condition clusters; each working condition cluster includes at least a portion of sample information.

[0008] For each of the aforementioned operating condition clusters, a cluster center for the cluster is determined; wherein each of the aforementioned operating condition clusters includes multiple intra-cluster samples;

[0009] From the samples within each cluster, the sample within the cluster with the smallest weighted distance to the cluster center is determined as the target sample.

[0010] Constraint evaluations are performed on each of the target samples, and the target samples whose evaluation results meet the evaluation conditions are used to control the operation of the thermal power plant.

[0011] In one embodiment, obtaining the sample information set of the thermal power plant includes:

[0012] Obtain equipment information from a thermal power plant; the equipment information includes multiple real-time operating parameter information and constraint information corresponding to various operating scenarios;

[0013] In response to a received natural language scheduling instruction, a thermal power plant work scenario label matching the natural language scheduling instruction is determined;

[0014] From the constraint information described above, determine the target constraint information that matches the working scenario label of the thermal power plant;

[0015] Based on the target constraint information, constraint filtering is performed on each of the real-time working parameter information to determine the target working parameter information that matches the target constraint information, thereby obtaining a sample information set containing the target working parameter information and the risk weights corresponding to the target working parameter information.

[0016] In one embodiment, the method further includes:

[0017] Based on the degree of constraint satisfaction between the target working parameter information and the target constraint information, the risk weight of the target working parameter information is determined;

[0018] Based on the working scene label of the thermal power plant, determine the previous historical working condition that matches the working scene label of the thermal power plant.

[0019] The process of obtaining a sample information set containing the target working parameter information and the risk weights corresponding to the target working parameter information includes:

[0020] Based on the working condition information in the previous historical working condition, information is supplemented for other working parameter information besides the target working parameter information to obtain a sample information set containing the target working parameter information, the other working parameter information, and the risk weights corresponding to the target working parameter information and the other working parameter information respectively.

[0021] In one embodiment, obtaining a sample information set including the target operating parameter information, the other operating parameter information, and risk weights corresponding to the target operating parameter information and the other operating parameter information respectively includes:

[0022] Obtain the parameter weights corresponding to the target working parameter information and the other working parameter information, respectively;

[0023] The working parameter information that does not meet the weight conditions is deleted, resulting in a sample information set containing the target working parameter information, the other working parameter information, and the risk weights corresponding to the target working parameter information and the other working parameter information, which meet the weight conditions.

[0024] In one embodiment, the step of dividing multiple sample information in the sample information set based on pre-set parameter association information to obtain multiple working condition clusters includes:

[0025] The pre-set parameter association information and the sample information set are input into the sparse subspace clustering model;

[0026] Multiple operating condition clusters are determined based on the output of the sparse subspace clustering model.

[0027] In one embodiment, the constraint evaluation of each of the target samples, and the use of target samples whose evaluation results meet the evaluation conditions for controlling the operation of the thermal power plant, includes:

[0028] Based on a pre-defined evaluation index system that includes equipment performance constraints, safe operation constraints, and environmental emission constraints, the evaluation result for each target sample is determined.

[0029] The target samples whose evaluation results meet the evaluation criteria are used to control the operation of the thermal power plant.

[0030] Secondly, this application also provides a power plant dispatching device, comprising:

[0031] The information set acquisition module is used to acquire a sample information set of thermal power plants;

[0032] The information segmentation module is used to segment multiple sample information in the sample information set based on pre-set parameter association information to obtain multiple working condition clusters; each working condition cluster includes at least a portion of sample information.

[0033] The cluster center determination module is used to determine the cluster center of each operating condition cluster; wherein each operating condition cluster includes multiple intra-cluster samples.

[0034] The target sample determination module is used to determine the intra-cluster sample with the smallest weighted distance to the cluster center from the intra-cluster samples as the target sample;

[0035] The constraint evaluation module is used to perform constraint evaluation on each of the target samples, and to use the target samples whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant.

[0036] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0037] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the method described above.

[0038] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps of the method described above.

[0039] The aforementioned thermal power plant dispatching method, apparatus, computer equipment, computer-readable storage medium, and computer program product achieve refined classification of complex operating conditions by acquiring a set of thermal power plant sample information and dividing it into multiple operating condition clusters based on parameter correlation information. This avoids the problem of incomplete operating condition coverage caused by traditional fixed-rule division. For each operating condition cluster, a cluster center is determined, and the target sample with the smallest weighted distance is selected from the samples within the cluster. This accurately captures the optimal operating parameters under similar operating conditions, overcoming the limitations of reliance on human experience. Furthermore, constraint evaluation is used to select the target sample with the best evaluation result to control the operation of the thermal power plant, ensuring that the dispatching strategy achieves global optimization under constraints such as safety and environmental protection. Compared with existing technologies, this method significantly improves the flexibility of thermal power plant dispatching through dynamic operating condition division and data-driven optimization strategies. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is an application environment diagram of a thermal power plant dispatching method in one embodiment;

[0042] Figure 2 This is a flowchart illustrating a thermal power plant dispatching method in one embodiment;

[0043] Figure 3 This is a flowchart illustrating the power plant dispatching method in another embodiment;

[0044] Figure 4 This is a flowchart illustrating the steps for determining the operating condition cluster in one embodiment;

[0045] Figure 5 This is a flowchart illustrating the process of determining the operating condition cluster in another embodiment;

[0046] Figure 6 This is a flowchart illustrating the power plant dispatching method in yet another embodiment;

[0047] Figure 7 This is a structural block diagram of a thermal power plant dispatching device in one embodiment;

[0048] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] The thermal power plant dispatching method provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located on the cloud or other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. Specifically, during the scheduling of the thermal power plant, server 104 obtains a set of sample information of the thermal power plant from terminal 102; based on pre-set parameter association information, it divides the multiple sample information in the sample information set into multiple operating condition clusters; each operating condition cluster includes at least a portion of sample information; for each operating condition cluster, it determines the cluster center; wherein each operating condition cluster includes multiple intra-cluster samples; from the intra-cluster samples, it determines the intra-cluster sample with the smallest weighted distance to the cluster center as the target sample; it performs constraint evaluation on each target sample, and uses the target sample whose evaluation result meets the evaluation conditions to control the operation of the thermal power plant.

[0051] In one exemplary embodiment, such as Figure 2 As shown, a method for dispatching thermal power plants is provided, which can be applied to... Figure 1 Taking server 104 as an example, the explanation includes:

[0052] Step S202: Obtain a set of sample information from thermal power plants.

[0053] Thermal power plants are power plants that generate heat energy by burning fuels such as coal and natural gas, and then convert that heat energy into electrical energy through equipment such as steam turbines or gas turbines. The sample information set is a collection of various data related to the operation of thermal power plants. This information can cover multiple parameters of the thermal power plant under different times and operating conditions, such as unit load, fuel consumption, steam temperature, and pressure.

[0054] Specifically, the server obtains a set of sample information from the thermal power plant from the terminal to facilitate subsequent information processing and scheduling. For example, this acquisition can be proactive or reactive. Further, the server can first construct a scheduling knowledge graph, writing the thermal power plant's equipment nodes, parameter nodes, and corresponding hard constraints such as safety red lines, emission limits, and start-stop intervals into a graph database using unified encoding. It then sets up a retrieval interface for task tags → constraint links, receives natural language scheduling instructions input by technicians, parses them to obtain a set of scenario tags, calls the knowledge graph to extract the corresponding constraint subgraph, maps the hard constraints to a list of measurement point limits, performs rigid filtering and dynamic threshold calculation on the real-time data stream, and generates a set of sample information with risk weights.

[0055] Step S204: Based on the pre-set parameter association information, the information of multiple samples in the sample information set is divided to obtain multiple working condition clusters.

[0056] Each operating condition cluster includes at least a subset of sample information. Parameter correlation information is pre-defined information that reflects the interrelationships and influence patterns among different operating parameters of a thermal power plant. For example, there may be positive or negative correlations between certain parameters, or certain combinations of parameters may reflect specific operating condition characteristics. Operating condition clusters are different categories obtained by dividing the sample information set according to parameter correlation information. The sample information within each operating condition cluster has similar characteristics and operating patterns, representing a specific operating condition or range of operating conditions for the thermal power plant.

[0057] Specifically, the server, based on pre-set parameter association information, partitions multiple samples in the sample information set, grouping samples that satisfy the parameter association relationship into the same working condition cluster, thus obtaining multiple working condition clusters. Optionally, the server can use conventional K-Means to partition the data by randomly selecting initial cluster centers and iteratively updating the cluster centers and sample affiliations. To incorporate parameter association information, parameters can be preprocessed, such as normalized, to make parameters of different dimensions comparable. Then, an appropriate number of clusters K is determined based on parameter association analysis, for example, using the elbow rule (observing the change in the loss function under different K values ​​and selecting the K value corresponding to the point where the decrease in the loss function slows down). During the iteration process, for each sample, not only is its Euclidean distance to each cluster center calculated, but parameter association weights are also introduced. For example, if two parameters are strongly correlated, the weights of the corresponding dimensions of these two parameters are increased when calculating the distance. In this way, samples with similar parameter features are more likely to be assigned to the same cluster, thus obtaining working condition clusters that conform to the parameter association relationship.

[0058] Optionally, the server can also input pre-set parameter association information and sample information set into the sparse subspace clustering model, and determine multiple operating condition clusters based on the output of the sparse subspace clustering model.

[0059] Step S206: For each operating condition cluster, determine the cluster center of the operating condition cluster.

[0060] Each operating condition cluster comprises multiple intra-cluster samples. The cluster center is a central point that represents the overall characteristics of an operating condition cluster. Numerically, it is typically the average value of all samples within the cluster across various parameter dimensions, or a representative value calculated using a specific algorithm. Intra-cluster samples are specific sample information belonging to the same operating condition cluster, and each sample information contains the specific operating parameters of the thermal power plant under that operating condition.

[0061] Specifically, the server can collect data on all samples within the operating condition cluster across various parameter dimensions. These parameters can include numerical features corresponding to information extracted from the knowledge graph, such as load range, emission indicators, and start-up / shutdown costs. For each parameter dimension, the arithmetic mean of all sample values ​​in that dimension is calculated.

[0062] Step S208: From the samples within each cluster, determine the sample within the cluster with the smallest weighted distance to the cluster center as the target sample.

[0063] Among them, the target sample is the sample within each working condition cluster that has the smallest weighted distance from the cluster center, and is the sample that best represents the typical characteristics of that working condition cluster.

[0064] Specifically, the server can iterate through all intra-cluster samples within the working condition cluster, calculate their weighted distances to the cluster center, compare the magnitudes of these weighted distances, and select the intra-cluster sample with the smallest weighted distance as the target sample.

[0065] For example, the server can read the cluster identifier of each operating condition cluster one by one. Indices of representative samples within the cluster The system sequentially calls three forms from the knowledge graph: the operating load interval table retrieves the mean and range of the main steam load column within a cluster, generating typical load labels; the emission specification table retrieves the values ​​of the same cluster within the main steam load interval; and the emission specification table retrieves the values ​​of the same cluster within the main steam load interval. , The 95th percentile value of the three categories of particulate matter emissions is compared with the regulatory threshold and written into the compliant emission label; the start-up and shutdown cost table uses the number of times the equipment is started and stopped and the fuel unit price as the key to query the corresponding start-up and shutdown costs and generate cost labels.

[0066] The three tags are concatenated into a three-dimensional semantic tag string using the load-emission-cost sequence and written into the cluster-level metadata row, completing the structured encapsulation of key business indicators and avoiding recalculation on the interface. For each tagged cluster, the sample set is traversed again, and three types of key control fields are identified based on the mapping table: main control valve opening column, fuel ratio column, and air volume ratio column; field weights are stored using a knowledge graph. Calculate the weighted Euclidean distance from each sample within a cluster to the cluster center, and sort the results in ascending order. Select the sample record with the smallest distance within the cluster as the representative control scheme, and write its valve opening, fuel ratio, and air volume ratio values, along with the source sample number and cluster identifier, into the database to form a list of directly executable commands. This ensures that the extracted scheme is both mathematically closest to the cluster center and physically reusable from real operating parameters. Assemble the three-dimensional semantic labels with the representative control scheme to generate a complete operating condition template. Perform a global sort of all templates from low to high typical load, and then apply the configured thresholds. Only the first part Each template package is written to the executable condition pool. A unique serial number is generated for each template package. Record the applicable effective period and associated security mask version number, and push the results to the technical personnel's terminal; simultaneously, write the results to the system log. Triple index.

[0067] Step S210: Perform constraint evaluation on each target sample, and use the target samples whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant.

[0068] Constraint assessment involves examining and evaluating a target sample under a series of limiting conditions. These constraints may include safety requirements (such as equipment operating parameters not exceeding safety thresholds), environmental requirements (such as pollutant emissions not exceeding prescribed standards), and economic requirements (such as operating costs being controlled within a certain range). Assessment conditions are the conditions used to determine the assessment results of the target sample; for example, assessment conditions can be best, optimal, or numerically optimal.

[0069] Specifically, the server can first define various constraints on the operation of the thermal power plant. These constraints may include safety constraints (such as equipment operating parameters not exceeding safety thresholds to prevent equipment damage and accidents), environmental constraints (such as pollutant emission concentrations not exceeding national and local standards to reduce environmental pollution), and economic constraints (such as operating costs being controlled within a certain range to improve the economic efficiency of the thermal power plant). The parameter values ​​of each target sample are then substituted into these constraints for inspection and evaluation. For example, it checks whether the pollutant emission parameters in the target sample meet environmental constraints and whether the equipment operating parameters meet safety constraints. Based on the evaluation results, the target samples are ranked, and the target sample that best satisfies all constraints and has the best overall performance is selected. The operating parameters corresponding to this target sample with the best evaluation result are used as control commands to control the actual operation of the thermal power plant, enabling it to achieve optimal operation while meeting all constraints.

[0070] Optionally, the server can determine the evaluation result of each target sample based on a preset evaluation index system that includes equipment performance constraints, safe operation constraints, and environmental emission constraints, and use the target samples whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant.

[0071] Optionally, the server can also assign corresponding weights to different objectives such as safety constraints, environmental constraints, and economic constraints based on the actual needs and priorities of the thermal power plant. For example, during periods of stringent environmental requirements, the weight of environmental constraints can be appropriately increased; when economic pressure is high, the weight of economic constraints can be increased. The evaluation results of each objective sample under each constraint condition are weighted and summed to obtain a comprehensive score. Then, the objective sample with the highest comprehensive score is selected as the optimal solution for controlling the operation of the thermal power plant.

[0072] The aforementioned thermal power plant scheduling method achieves refined classification of complex operating conditions by acquiring a set of sample information from thermal power plants and dividing them into multiple operating condition clusters based on parameter correlation information. This avoids the problem of incomplete operating condition coverage caused by traditional fixed-rule division. For each operating condition cluster, a cluster center is determined, and the target sample with the smallest weighted distance is selected from the samples within the cluster. This accurately captures the optimal operating parameters under similar operating conditions, overcoming the limitations of reliance on manual experience. Furthermore, constraint evaluation is used to select the target sample with the best evaluation results to control the operation of the thermal power plant, ensuring that the scheduling strategy achieves global optimization under constraints such as safety and environmental protection. Compared with existing technologies, this method significantly improves the flexibility of thermal power plant scheduling through dynamic operating condition division and data-driven optimization strategies.

[0073] In an exemplary embodiment, obtaining a sample information set of a thermal power plant includes: obtaining equipment information of the thermal power plant; the equipment information includes multiple real-time operating parameter information and constraint information corresponding to various operating scenarios; in response to a received natural language scheduling instruction, determining a thermal power plant operating scenario label that matches the natural language scheduling instruction; from each constraint information, determining target constraint information that matches the thermal power plant operating scenario label; based on the target constraint information, performing constraint filtering on each real-time operating parameter information to determine target operating parameter information that matches the target constraint information, thereby obtaining a sample information set containing the target operating parameter information and the risk weight corresponding to the target operating parameter information.

[0074] The equipment information can exist in the form of a knowledge graph. Real-time operating parameter information refers to various parameters corresponding to the real-time operation of thermal power plant equipment, such as equipment nodes and parameter nodes. Specifically, it can include, for example, the DCS (Distributed Control System Point Number) number within the plant, fuel ledgers, and equipment abbreviations, parameter names, and time periods from environmental monitoring records and maintenance work orders. This reflects the current operating status of the equipment. Constraint information consists of restrictions set for different operating scenarios of the thermal power plant, such as hard constraints like safety red lines, emission limits, and start-stop intervals. Natural language dispatch instructions are dispatch requirements for the operation of the thermal power plant given in natural language, such as "increase power generation" and "reduce pollutant emissions." Thermal power plant operating scenario tags are used to identify different operating scenarios of the thermal power plant, such as "high load power generation scenario" and "low load maintenance scenario," facilitating the management and processing of information under different scenarios. Target constraint information is the constraint information corresponding to the thermal power plant operating scenario matched with the natural language dispatch instructions, used to filter operating parameters that meet the requirements. The target operating parameter information is the real-time operating parameter information that matches the target constraint information, that is, the equipment operating parameters that meet the constraints of the current working scenario.

[0075] Specifically, the server can pre-write the equipment nodes, parameter nodes, and corresponding hard constraints such as safety red lines, emission limits, and start-stop intervals of the thermal power plant into the graph database with unified encoding as equipment information, and set up a retrieval interface for task tags → constraint links. It can receive natural language scheduling instructions and parse them to obtain a set of thermal power plant working scenario tags, call the knowledge graph to extract the corresponding constraint subgraph, map the constraint information into a list of measurement point limits, perform rigid filtering and dynamic threshold calculation on real-time working parameter information, and generate a set of sample information with risk weights.

[0076] Furthermore, the server can aggregate the equipment abbreviation, parameter name, and time period from the DCS point number, fuel ledger, environmental monitoring records, and maintenance work orders within the plant, and assign them numbers using the unified coding system for the thermal power industry; using the functional level as the index, each equipment field is assigned a unique identifier. Assign a unique identifier to each parameter field. .

[0077] Hard constraints involving safety red lines, emission limits, and mandatory start-stop intervals are simultaneously incorporated into the constraint domain and generated one by one. The applicable scope and priority are marked. Based on the finalized glossary, the original records of each business system are scanned in chronological order. First, a whitelist mapping is used to lock the device primary keys. Secondly, regular expressions are used to map the names of measurement points within the same row of records to... Import constraint codes from the corresponding table. And add the effective period Assemble into ⟨ Quadruple.

[0078] Write the continuously generated quadruples in batches into the graph database: and Construct entity nodes, to , Attribute nodes are constructed and connected by directed relationships to form a scheduling semantic network covering safety red lines, emission limits, and start / stop intervals. A load-oriented adjustment low-NOx operation scenario tag query entry is configured for this network: following a mapping method of task tag → constraint link → node set, a single interface is provided for the real-time filtering module to ensure that the same source is always referenced during scheduling, completely avoiding omissions or conflicts caused by scattered rules.

[0079] Set the input natural language string as ;right Perform word segmentation to obtain a word sequence. A pre-compiled keyword and tag mapping matrix for thermal power scenarios is adopted. Perform matching on the word sequence and output a label vector:

[0080]

[0081] in: This is a label vector, with elements corresponding to scenario labels such as load increase and nitrogen oxide constraint, in order of their positions. This is a keyword-tag mapping matrix, with elements as follows: ; This is a sparse vector representation of a word sequence, with elements of . .

[0082] vector Median The indexes are summarized into a tag set. and put The instruction timestamp and operator number are written into the cache to lock the task identifier for subgraph extraction and permission tracking.

[0083] tag set Input the knowledge graph one by one, with each Starting from the task label, perform a depth-first traversal along the direction of the constraint node, retaining only the paths where the node type belongs to the safety red line, emission limit, or start-stop interval, to obtain the node set. and relation set .

[0084] Subgraph The nodes marked as mandatory constraints are mapped to the list of measurement points, resulting in an index set. And retrieve the limit value for each index. Construct a Boolean mask matrix accordingly. Let the first... Row sample Perform the following on all samples:

[0085]

[0086] in: For the sample At the measuring point The reading at the location; For measuring points The corresponding limit value; This is the sample set after rigid filtering. Any reading that violates this rule... The row samples are immediately removed, and the violation test points and timestamps are recorded simultaneously to prevent the violation data from entering the subsequent wideband calculation and clustering process.

[0087] The sample set that passed the rigid filter Re-corresponding to lightweight subgraph ,right Each hard constraint node reads three scene attributes: static limit value. Threshold scaling factor and remaining tolerance time Based on the sample arrival time Record the remaining time until the end of this constraint takes effect. Combined with the current reading (in To constrain The associated measurement point index), is constrained according to the following formula. Calculate the dynamic bandwidth:

[0088]

[0089] in: To constrain At any moment Acceptable bandwidth width; This is the threshold scaling factor, preset by the production technology department, and its value is [value to be filled in]. ; For the sample At the measuring point Real-time readings; To constrain The hard limit value; To constrain At any moment Remaining tolerance time; To constrain The absolute failure moment.

[0090] when satisfy At that time, the sample Recorded in the yellow alert set And write it in its row-level metadata. and If the inequality does not hold for all constraints, then... Deemed fully compliant to remain Neither type of sample will be deleted.

[0091] Based on this, the dynamic threshold band provides room for subsequent correction of data rows approaching the risk edge without violating hard compliance. For each sample row... Summarize the number of yellow alerts and calculate the row-level risk weight:

[0092]

[0093] in: For the sample The overall risk weight; For the sample The set of constraint indexes that trigger a yellow alert. Write the warning weight for the newly added column to provide a quantitative basis for selective weight reduction in subsequent modules.

[0094] Then, Early warning samples and its corresponding Write them together to the output buffer to generate the feature-sample matrix. The row index is the timestamp, the column index is the measurement point number, and a warning weight column is appended at the end.

[0095] The total number of samples in the output buffer and the task threshold Comparison: If the sample size is insufficient, a red alert is immediately sent to the terminal where the technician is located, requesting them to reset the target load or relax the constraints; if the sample size meets the requirements, a unique batch identifier is generated for this batch of data, and this identifier, along with... The data is then passed to the map incremental completion module. Based on this dual-layer threshold band filtering, it ensures that all samples meet the hard requirements while retaining sufficient and correctable working condition samples for subsequent clustering.

[0096] In this embodiment, by acquiring equipment information and combining it with natural language scheduling instructions, it is possible to accurately filter out working parameter information that meets the requirements of the current working scenario, forming a sample information set. This provides a targeted data foundation for subsequent analysis and decision-making, which helps to improve the targeting and efficiency of thermal power plant operation.

[0097] In an exemplary embodiment, the thermal power plant scheduling method further includes: determining the risk weight of the target operating parameter information based on the degree of constraint satisfaction between the target operating parameter information and the target constraint information; determining the previous historical operating condition that matches the thermal power plant operating scenario label based on the thermal power plant operating scenario label; and obtaining a sample information set containing the target operating parameter information and the risk weight corresponding to the target operating parameter information, including: supplementing the other operating parameter information besides the target operating parameter information according to the operating condition information in the previous historical operating condition, to obtain a sample information set containing the target operating parameter information, other operating parameter information, and the risk weights corresponding to the target operating parameter information and other operating parameter information respectively.

[0098] Among them, the previous historical operating condition is the operating condition that has occurred before and is matched with the current thermal power plant operating scenario label, including various operating condition information under that condition.

[0099] Specifically, the server can retrieve the previous historical operating conditions along the device-parameter-constraint link based on the scene tag, and fill in the missing fields of the compliant samples with confidence weights based on freshness weights and coupling weights to obtain the sample information set.

[0100] Furthermore, the server can denote the set of power plant work scene tags returned in the above embodiments as follows: For each tag The set of working parameter information that satisfies the label is obtained by calling the inverse index of the graph. The set intersection operation is used to lock the set of target working parameter information that simultaneously satisfies all label constraints. This refers to the complete set of information that must be involved in scheduling under the current task. For the first Each scene tag; To satisfy the label A set of working parameter information.

[0101] right For each working parameter, depth-first diffusion is performed along the equipment, parameter, and constraint directions to piece together a unique ordered link that satisfies the requirements of low nitrogen, load increase, and current start-stop interval. The link number, node sequence, and corresponding constraint code are then written into a temporary table. This provides a unique starting point for tracing history. Indicates working parameters The historical operation record list attached to the graph includes timestamps. With legal status identifier Along the link Perform a parallel scan on all working parameters and calculate a joint valid identifier for each record:

[0102]

[0103] in For joint valid identifiers, a value of 1 indicates that all working parameters on the chain are recorded. All nodes are in a valid state; a value of 0 indicates that at least one node is invalid. For record Working parameters The legal identifier.

[0104] Traverse records from nearest to farthest timestamp, when a record appears... Stop immediately and lock the timestamp of the record. And copy the corresponding device status and parameter values ​​to the reference working parameter set. The operation ensures that all referenced values ​​are derived from a complete and compliant real-world operating condition, avoiding physical inconsistencies caused by manual patchwork.

[0105] Reference working parameter set With the current sample matrix Perform field alignment. (Set) This indicates a double key for device number and parameter name. Middle field If a null value is found, the following mapping function will be called.

[0106]

[0107] in, For fields The first candidate value; For , and timestamp Use as the search key, from A mapping function that retrieves the corresponding numerical value; To standardize equipment numbering; To standardize parameter numbering; The timestamp for the compliance record locked in the previous sub-step.

[0108] All Collect into a list of candidate values If the field already has a real value, then... The record is set as an empty candidate and its original value is retained. At this point, each candidate value has three sources: device number, parameter name, and real timestamp, and can be directly entered into subsequent confidence score calculation without any averaging or extrapolation operations, ensuring the physical traceability of the completion action.

[0109] For the candidate value list Each record in (index , (Each corresponds to a unified device and parameter number), and the data freshness weights already saved in the knowledge graph are called. Weight of physical coupling strength The attention coefficient preset by the dispatching department is adopted. Calculate the overall confidence score:

[0110]

[0111] in: Candidate values The overall confidence score; This represents the time decay weight of the value from the current moment; the fresher the value, the higher the value. This represents the physical coupling strength between this value and the target working condition path; the closer the coupling, the higher the value. The weighting factor is set by the scheduler to control the proportion of freshness and coupling in the confidence score.

[0112] Will With system threshold Comparison: If Then Write back the sample matrix The field status column is marked as hard-filled. It can be understood that in this embodiment, information is filled in by comprehensive confidence. That is, if the comprehensive confidence is greater than or equal to the system threshold, then the working parameter information can be hard-filled.

[0113] The matrix that will be hard-filled Write to the output buffer and update the field status label for each column; then generate a new version number for this batch of data. And write to the log system, including , Number and confidence threshold Key information.

[0114] In this embodiment, by determining risk weights and supplementing other operating parameter information, the sample information set is made more complete and accurate, which can more comprehensively reflect the operating status of the thermal power plant under the current working scenario, and provide more reliable data support for subsequent operating condition analysis and control decisions.

[0115] In an exemplary embodiment, obtaining a sample information set including target working parameter information, other working parameter information, and risk weights corresponding to the target working parameter information and other working parameter information respectively includes: obtaining parameter weights corresponding to the target working parameter information and other working parameter information respectively; deleting working parameter information whose parameter weights do not meet the weight conditions, and obtaining a sample information set including target working parameter information, other working parameter information, and risk weights corresponding to the target working parameter information and other working parameter information respectively, which meet the weight conditions.

[0116] Among them, parameter weight is an indicator reflecting the importance of target working parameter information and other working parameter information in the overall context. Weight conditions are pre-defined conditions used to determine whether parameter weights meet the requirements; only working parameter information that meets the conditions will be retained.

[0117] Specifically, the server can denote the set of thermal power plant work scene tags returned in the above embodiments as follows: For each tag The set of working parameter information that satisfies the label is obtained by calling the inverse index of the graph. The set intersection operation is used to lock the set of target working parameter information that simultaneously satisfies all label constraints. This refers to the complete set of information that must be involved in scheduling under the current task. For the first Each scene tag; To satisfy the label A set of working parameter information.

[0118] right For each working parameter, depth-first diffusion is performed along the equipment, parameter, and constraint directions to piece together a unique ordered link that satisfies the requirements of low nitrogen, load increase, and current start-stop interval. The link number, node sequence, and corresponding constraint code are then written into a temporary table. This provides a unique starting point for tracing history. Indicates working parameters The historical operation record list attached to the graph includes timestamps. With legal status identifier Along the link Perform a parallel scan on all working parameters and calculate a joint valid identifier for each record:

[0119]

[0120] in For joint valid identifiers, a value of 1 indicates that all working parameters on the chain are recorded. All nodes are in a valid state; a value of 0 indicates that at least one node is invalid. For record Working parameters The legal identifier.

[0121] Traverse records from nearest to farthest timestamp, when a record appears... Stop immediately and lock the timestamp of the record. And copy the corresponding device status and parameter values ​​to the reference working parameter set. The operation ensures that all referenced values ​​are derived from a complete and compliant real-world operating condition, avoiding physical inconsistencies caused by manual patchwork.

[0122] Reference working parameter set With the current sample matrix Perform field alignment. (Set) This indicates a double key for device number and parameter name. Middle field If a null value is found, the following mapping function will be called.

[0123]

[0124] in, For fields The first candidate value; For , and timestamp Use as the search key, from A mapping function that retrieves the corresponding numerical value; To standardize equipment numbering; To standardize parameter numbering; The timestamp for the compliance record locked in the previous sub-step.

[0125] All Collect into a list of candidate values If the field already has a real value, then... The record is set as an empty candidate and its original value is retained. At this point, each candidate value has three sources: device number, parameter name, and real timestamp, and can be directly entered into subsequent confidence score calculation without any averaging or extrapolation operations, ensuring the physical traceability of the completion action.

[0126] For the candidate value list Each record in (index , (Each corresponds to a unified device and parameter number), and the data freshness weights already saved in the knowledge graph are called. Weight of physical coupling strength The attention coefficient preset by the dispatching department is adopted. Calculate the overall confidence score:

[0127]

[0128] in: Candidate values The overall confidence score; This represents the time decay weight of the value from the current moment; the fresher the value, the higher the value. This represents the physical coupling strength between this value and the target working condition path; the closer the coupling, the higher the value. The weighting factor is set by the scheduler to control the proportion of freshness and coupling in the confidence score.

[0129] Will With system threshold Comparison: If Then Write back the sample matrix And mark the status column as hard-filled;

[0130] like If the parameter weights do not meet the weight conditions, the gap is preserved and marked as soft missing in the field status column, so that subsequent clustering can automatically reduce the weights. After processing, the status labels and corresponding values ​​of all fields are updated. and threshold Write the completed results into the list. The list is numbered and logged to support post-event auditing.

[0131] The working condition prototype chain Reload, for each parameter node in the chain Calling the task relevance weights of the graph field This weight is determined by the operations and maintenance department based on years of scheduling experience. The higher the consistency with the task objectives (load, emissions, safety), the higher the weight. The larger:

[0132]

[0133] in: For parameters Task relevance weights; For all from within the atlas A set of feasible paths pointing to the task target node; For path The task-oriented weights are automatically updated by the graph maintenance script.

[0134] Press all fields Perform a one-time screening: If (The threshold is a hyperparameter that can be set by the technical lead), then the corresponding column is retained; if Then in the sample matrix Column-level hiding is performed. Semantic dimension selection ensures that highly relevant coupling parameters are fully preserved while eliminating low-contribution redundant dimensions, preventing cluster shifts in subsequent sparse subspace clustering due to irrelevant data. Finally, the matrix metadata table is updated to record the values ​​of each column. The compressed matrix, along with its metadata, is written to the output buffer, retaining / hiding the state. This completes the hard-filling, soft-missing annotation, and dimension selection of the matrix. Write to the output buffer, update the field status label (original value, hard-filled, soft-missing) for each column; then generate a new version number for this batch of data. And write to the log system, including , Number and confidence threshold Key information.

[0135] In this embodiment, by deleting unimportant working parameter information, the amount of data is reduced and the data processing efficiency is improved, while key information is retained, making subsequent analysis and decision-making more focused and effective.

[0136] In one exemplary embodiment, such as Figure 3 As shown, the thermal power plant dispatching methods include:

[0137] Step S301: Obtain equipment information from the thermal power plant;

[0138] The equipment information includes multiple real-time operating parameters and constraint information corresponding to various operating scenarios.

[0139] Step S302: In response to the received natural language scheduling instruction, determine the thermal power plant work scenario label that matches the natural language scheduling instruction;

[0140] Step S303: Determine the target constraint information that matches the working scenario label of the thermal power plant from the constraint information.

[0141] Step S304: Based on the target constraint information, perform constraint filtering on each real-time working parameter information to determine the target working parameter information that matches the target constraint information;

[0142] Step S305: Determine the risk weight of the target working parameter information based on the degree of constraint satisfaction between the target working parameter information and the target constraint information;

[0143] Step S306: Based on the thermal power plant work scenario label, determine the previous historical working condition that matches the thermal power plant work scenario label.

[0144] Step S307: Based on the working condition information in the previous historical working condition, supplement the working parameter information other than the target working parameter information.

[0145] Step S308: Obtain the parameter weights corresponding to the target working parameter information and other working parameter information respectively;

[0146] Step S309: Delete the working parameter information that does not meet the weight conditions, and obtain a sample information set containing the target working parameter information, other working parameter information, and the risk weights corresponding to the target working parameter information and other working parameter information that meet the weight conditions.

[0147] Step S310: Based on the pre-set parameter association information, the information of multiple samples in the sample information set is divided to obtain multiple working condition clusters;

[0148] Step S311: Perform constraint evaluation on each target sample, and use the target samples whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant.

[0149] In one exemplary embodiment, such as Figure 4 As shown, based on pre-set parameter association information, multiple sample information in the sample information set are divided into multiple working condition clusters, including:

[0150] Step S402: Input the pre-set parameter association information and sample information set into the sparse subspace clustering model;

[0151] Step S404: Determine multiple operating condition clusters based on the output of the sparse subspace clustering model.

[0152] Among them, the sparse subspace clustering model is a clustering model used to divide data into different subspaces. It achieves clustering by mining the sparsity features of the data. In this scenario, it is used to divide the sample information set into multiple working condition clusters.

[0153] Specifically, the server can construct parameter association information using the physical coupling and mutual exclusion relationships of measurement points recorded in the knowledge graph. The pre-set parameter association information and sample information set are then input into the sparse subspace clustering model. When solving the sparse representation coefficient matrix, rigid constraints are applied to elements in the mask that must or must not be linked. Using the nearest real-time load interval as a seed, the model iteratively expands through load corridors to obtain a connected subspace that satisfies the target load range, thereby dividing the operating condition clusters. In this embodiment, the parameter association information exists in the form of a safety mask matrix.

[0154] Furthermore, such as Figure 5 As shown, the server can output the sample matrix from the above embodiments. Along with batch markings Import the clustering task pool and perform unified test point code matching on the primary key of the column header; where the test point number is... A mapping function is used to map each measurement point number to a unique column sequence number, generating a mapping table. Subsequently, on Perform a two-level check: first, compare for duplicate mappings; if found... and If an error occurs, report it immediately; then check if any essential measurement points are missing. If any are missing, the technical staff should supplement them according to the map task template. Lock the map upon completion. The row and column dimensions and time boundaries, and put Batch Identifier Write it together to the task metadata area.

[0155] The metadata area records the physical location information (furnace number, loop number, denitrification channel number) of each column, while also indicating its functional category (combustion, air supply, emission, etc.). This ensures that the subsequent mask matrix can directly identify the coupling or mutual exclusion relationships between measuring point pairs by querying the physical location information, avoiding additional inter-table join operations. The knowledge graph interface is called to read the list of relationships within the same furnace and loop. and a list of mutually exclusive / overheating risk relationships. Both lists are formatted by measurement point number. As a basic element. ( Initialize the mask matrix in the dimension of (number of columns). .

[0156] Based on business relationships Element-by-element assignment;

[0157]

[0158] Wherein: is the element of the -th row and -th column of the mask matrix; is the set of measurement point pairs that need to maintain linkage; is the set of measurement point pairs that are prohibited from being grouped into the same cluster simultaneously; is the measurement point pair not subject to hard coupling constraints.

[0159] After construction, perform a symmetry check on . If , then backtrack to the data source for correction to ensure that the left and right sides of the mask matrix are consistent; subsequently, assign a security version number to and store it in the cache to ensure that the same security benchmark is referenced throughout the solution process. Synchronously input the sample matrix and the security mask into the sparse subspace solution; when solving the sparse representation coefficient matrix , introduce the following optimization model with mask constraints: Wherein:

[0160]

[0161]

[0162] Wherein: is the sparse representation coefficient matrix; is the Frobenius norm; is the Hadamard (element-wise) product; is the main diagonal identity matrix; is the -th row and -th column element of the matrix .

[0163] During the iteration process, force the position of to be non-zero to lock the physical linkage; force the position of to be for to avoid misconnection of potentially mutually exclusive measurement points; the remaining elements are adaptively adjusted by the optimizer according to sparsity. After the solution, output the weighted coefficient matrix and record the injected security mask version identifier in the metadata area. Denote the current unit's real-time load value as , and read the lower limit - upper limit list of the existing operating load - economic interval nodes in the knowledge graph. Calculate the distance using the midpoint of the interval, and calculate according to the following formula:

[0164]

[0165] in, The range index that best approximates the real-time load; This is the current load reading; For the first The upper and lower limits of an economic zone.

[0166] Locked range Then, the weighted coefficient matrix The middle column belongs to The sample indexes are collected as the initial set. Then the matrix Calculate the connectivity score:

[0167]

[0168] in, For the sample The cumulative connectivity with the seed interval; These are the elements of the sparse coefficient matrix; For interval Internal sample index set.

[0169] Will Sort from highest to lowest, select the top The first set of cluster seeds is generated from each sample. And encapsulate it into the starting subspace The subspace is the sole starting point for subsequent load iterations, ensuring that the seed is physically closest to the real-time load. Perform load iterative control and read the target load corridor from the knowledge graph. In each iteration From Extracting from the current subspace Candidate samples with the highest interconnectivity that have not yet been clustered ;right The following decision logic is adopted:

[0170]

[0171] in, The result is a Boolean decision. For the corresponding elements of the security mask matrix, Indicates mutual exclusion. or Indicates that it can be connected; Candidate samples Real-time load readings; The upper and lower limits of the target load corridor for scheduling.

[0172] like Then Merge into subspace And update the coverage area; if the determination fails, discard the data. And continue iterating at the current layer. Repeat this process until... The load range is fully covered Or there may be no more acceptable samples to expand upon.

[0173] Iteration ensures that the subspace satisfies both secure coupling relationships and falls entirely within the load range set by the scheduler, providing a safe and business-appropriate sample set for connected component partitioning. The expanded subspace... Labels are generated, and connected components are partitioned into clusters for connected samples to obtain cluster sets. For each cluster, the load statistics interval is calculated and verified to fall entirely within the allowed corridor. Then, the cluster center sample is selected as a representative sample and combined into a cluster label and representative sample output package. Finally, the security mask version number, load locking trajectory, and cluster statistics information are written to the log to provide a complete chain of evidence for operating condition template extraction and audit tracking.

[0174] In this embodiment, the sparse subspace clustering model can automatically and accurately divide sample information into multiple operating condition clusters, which helps to discover different patterns and rules in the operation of thermal power plants and provides a basis for subsequent refined management and control.

[0175] In an exemplary embodiment, constraint evaluation is performed on each target sample, and the target samples whose evaluation results meet the evaluation conditions are used to control the operation of the thermal power plant. This includes: determining the evaluation result of each target sample based on a preset evaluation index system that includes equipment performance constraints, safe operation constraints, and environmental emission constraints; and using the target samples whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant.

[0176] The evaluation index system is a pre-set set of indicators used to evaluate the target sample. It includes indicators such as equipment performance constraints, safe operation constraints, and environmental emission constraints, and is used to determine the evaluation results of the target sample.

[0177] Specifically, the server can denote the entire template set output in the above embodiments as follows: And perform batch task matching; read four real-time constraints for the current scheduled task: target load Remaining number of start / stop attempts Emissions Limit With maintenance window The data is then merged to form a task feature vector. The knowledge graph interface is called to retrieve three weight tables: an economic weight table, storing the weighting factors for each fuel unit consumption under different load conditions; a wear weight table, recording the stress coefficients of key equipment at different start-stop frequencies; and a safety priority table, indicating the accident threshold margin corresponding to each template in historical operation.

[0178] For template Each record in the database extracts three sets of indicators: fuel consumption per unit area, cumulative equivalent stress, and safety margin. These are then used for feature mapping according to the aforementioned weight table to form a task adaptation vector with unified dimensions. The adaptation vectors are batch-processed for multi-criteria decision-making, employing a hierarchical weighting strategy of prioritizing safety, centering on economy, and then considering wear. First, templates are scored according to safety priority; templates with scores below the safety threshold are immediately eliminated to ensure the entire candidate set meets the safety red line. For the remaining templates, fuel cost increments are calculated based on economic weights and placed in the main weight area for comprehensive score calculation. Finally, wear weights are used to make secondary corrections to the equipment's cumulative stress, preventing short-term economically optimal solutions that cause long-term excessive wear from taking the lead. After hierarchical scoring, the comprehensive scores are sorted in descending order, and the top three are automatically selected to generate candidate solutions. Each solution is supplemented with three summary pieces of information: load change magnitude (relative to the current load), estimated emissions (converted according to the emission model), and additional fuel cost (converted according to the daily benchmark unit price). The top three candidate schemes, along with three summary information items, are written into the scheduling recommendation table: the scheme with the highest comprehensive score is set as the default recommendation, and the other two are marked as alternatives. At the interface layer, representative control commands are automatically expanded, and the main control valve opening, fuel ratio, and air volume ratio are packaged to generate a distribution list.

[0179] In this embodiment, by establishing an evaluation index system to comprehensively evaluate the target sample, the optimal control scheme can be selected to ensure that the thermal power plant improves its operating efficiency and economic benefits under the premise of safety and environmental protection.

[0180] In a specific embodiment, such as Figure 6 As shown, a method for dispatching a thermal power plant is also provided, including:

[0181] Step S601: Obtain equipment information from the thermal power plant;

[0182] The equipment information includes multiple real-time operating parameters and constraint information corresponding to various operating scenarios.

[0183] Step S602: In response to the received natural language scheduling instruction, determine the thermal power plant work scenario label that matches the natural language scheduling instruction;

[0184] Step S603: Determine the target constraint information that matches the working scenario label of the thermal power plant from the constraint information.

[0185] Step S604: Based on the target constraint information, perform constraint filtering on each real-time working parameter information to determine the target working parameter information that matches the target constraint information;

[0186] Step S605: Determine the risk weight of the target working parameter information based on the degree of constraint satisfaction between the target working parameter information and the target constraint information;

[0187] Step S606: Based on the thermal power plant work scenario label, determine the previous historical working condition that matches the thermal power plant work scenario label.

[0188] Step S607: Based on the working condition information in the previous historical working condition, supplement the information of other working parameters except for the target working parameter information.

[0189] Step S608: Obtain the parameter weights corresponding to the target working parameter information and other working parameter information respectively;

[0190] Step S609: Delete the working parameter information that does not meet the weight conditions, and obtain a sample information set containing the target working parameter information, other working parameter information, and the risk weights corresponding to the target working parameter information and other working parameter information that meet the weight conditions.

[0191] Step S610: Input the pre-set parameter association information and sample information set into the sparse subspace clustering model;

[0192] Step S611: Determine multiple working condition clusters based on the output of the sparse subspace clustering model;

[0193] Each operating condition cluster includes at least a portion of sample information;

[0194] Step S612: For each working condition cluster, determine the cluster center of the working condition cluster; wherein each working condition cluster includes multiple intra-cluster samples.

[0195] Step S613: From the samples within each cluster, determine the sample within the cluster with the smallest weighted distance to the cluster center as the target sample;

[0196] Step S614: Based on a preset evaluation index system that includes equipment performance constraints, safe operation constraints, and environmental emission constraints, determine the evaluation result for each target sample.

[0197] Step S615: Use the target sample whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant.

[0198] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0199] Based on the same inventive concept, this application also provides a thermal power plant dispatching device for implementing the above-described thermal power plant dispatching method. The solution provided by this device is similar to the solution described in the above-described method; therefore, the specific limitations in one or more embodiments of the thermal power plant dispatching device provided below can be found in the limitations of the thermal power plant dispatching method described above, and will not be repeated here.

[0200] In one exemplary embodiment, such as Figure 7 As shown, a thermal power plant dispatching device 700 is provided, including: an information set acquisition module 702, an information partitioning module 704, a cluster center determination module 706, a target sample determination module 708, and a constraint evaluation module 710, wherein:

[0201] The information set acquisition module 702 is used to acquire a sample information set of a thermal power plant;

[0202] The information segmentation module 704 is used to segment multiple sample information in the sample information set based on pre-set parameter association information to obtain multiple working condition clusters; each working condition cluster includes at least a portion of sample information.

[0203] The cluster center determination module 706 is used to determine the cluster center of each operating condition cluster; wherein each operating condition cluster includes multiple intra-cluster samples.

[0204] The target sample determination module 708 is used to determine the intra-cluster sample with the smallest weighted distance to the cluster center from the intra-cluster samples as the target sample;

[0205] The constraint evaluation module 710 is used to evaluate the constraints of each target sample and to use the target samples whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant.

[0206] In one exemplary embodiment, the information set acquisition module 702 includes:

[0207] The equipment information acquisition unit is used to acquire equipment information of the thermal power plant; the equipment information includes multiple real-time operating parameter information and constraint information corresponding to various operating scenarios.

[0208] The dispatch instruction response unit is used to respond to the received natural language dispatch instructions and determine the thermal power plant work scenario label that matches the natural language dispatch instructions.

[0209] The target constraint information determination unit is used to determine the target constraint information that matches the working scenario label of the thermal power plant from the various constraint information.

[0210] The constraint filtering unit is used to perform constraint filtering on each real-time working parameter information based on the target constraint information, determine the target working parameter information that matches the target constraint information, and obtain a sample information set containing the target working parameter information and the risk weights corresponding to the target working parameter information.

[0211] In an exemplary embodiment, the thermal power plant dispatching device 700 further includes a previous historical operating condition determination module, specifically used for:

[0212] Based on the degree of constraint satisfaction between target working parameter information and target constraint information, the risk weight of target working parameter information is determined;

[0213] Based on the working scene tags of thermal power plants, determine the previous historical working condition that matches the working scene tags of thermal power plants.

[0214] In this embodiment, the constraint screening unit includes an information completion component, which is used to: complete the information of other working parameters besides the target working parameter information according to the working condition information in the previous historical working condition, so as to obtain a sample information set containing the target working parameter information, other working parameter information, and risk weights corresponding to the target working parameter information and other working parameter information respectively.

[0215] In one exemplary embodiment, the information completion component is specifically used for:

[0216] Obtain the parameter weights corresponding to the target working parameter information and other working parameter information respectively;

[0217] The working parameter information that does not meet the weight conditions is deleted, resulting in a sample information set that includes the target working parameter information, other working parameter information, and the risk weights corresponding to the target working parameter information and other working parameter information, all of which meet the weight conditions.

[0218] In one exemplary embodiment, the information segmentation module 704 is specifically used for:

[0219] Input the pre-set parameter association information and sample information set into the sparse subspace clustering model;

[0220] Multiple operating condition clusters are determined based on the output of the sparse subspace clustering model.

[0221] In an exemplary embodiment, the constraint evaluation module 710 is specifically used for:

[0222] Based on a pre-defined evaluation index system that includes equipment performance constraints, safe operation constraints, and environmental emission constraints, the evaluation result for each target sample is determined.

[0223] The target samples whose evaluation results meet the evaluation criteria will be used to control the operation of thermal power plants.

[0224] Each module in the aforementioned thermal power plant dispatching device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0225] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a power plant scheduling method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0226] Those skilled in the art will understand that Figure 8The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0227] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0228] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0229] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described above.

[0230] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0231] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0232] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0233] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for dispatching a thermal power plant, characterized in that, The method includes: Obtain a sample information set of thermal power plants; Based on pre-set parameter association information, multiple sample information in the sample information set are divided to obtain multiple working condition clusters; each working condition cluster includes at least a portion of sample information. For each of the aforementioned operating condition clusters, a cluster center for the cluster is determined; wherein each of the aforementioned operating condition clusters includes multiple intra-cluster samples; From the samples within each cluster, the sample within the cluster with the smallest weighted distance to the cluster center is determined as the target sample. Constraint evaluations are performed on each of the target samples, and the target samples whose evaluation results meet the evaluation conditions are used to control the operation of the thermal power plant.

2. The method according to claim 1, characterized in that, The acquisition of the sample information set of thermal power plants includes: Obtain equipment information from a thermal power plant; the equipment information includes multiple real-time operating parameter information and constraint information corresponding to various operating scenarios; In response to a received natural language scheduling instruction, a thermal power plant work scenario label matching the natural language scheduling instruction is determined; From the constraint information described above, determine the target constraint information that matches the working scenario label of the thermal power plant; Based on the target constraint information, constraint filtering is performed on each of the real-time working parameter information to determine the target working parameter information that matches the target constraint information, thereby obtaining a sample information set containing the target working parameter information and the risk weights corresponding to the target working parameter information.

3. The method according to claim 2, characterized in that, The method further includes: Based on the degree of constraint satisfaction between the target working parameter information and the target constraint information, the risk weight of the target working parameter information is determined; Based on the working scene label of the thermal power plant, determine the previous historical working condition that matches the working scene label of the thermal power plant. The process of obtaining a sample information set containing the target working parameter information and the risk weights corresponding to the target working parameter information includes: Based on the working condition information in the previous historical working condition, information is supplemented for other working parameter information besides the target working parameter information to obtain a sample information set containing the target working parameter information, the other working parameter information, and the risk weights corresponding to the target working parameter information and the other working parameter information respectively.

4. The method according to claim 3, characterized in that, The process of obtaining a sample information set containing the target working parameter information, the other working parameter information, and the risk weights corresponding to the target working parameter information and the other working parameter information respectively includes: Obtain the parameter weights corresponding to the target working parameter information and the other working parameter information, respectively; The working parameter information that does not meet the weight conditions is deleted, resulting in a sample information set containing the target working parameter information, the other working parameter information, and the risk weights corresponding to the target working parameter information and the other working parameter information, which meet the weight conditions.

5. The method according to claim 1, characterized in that, The information is divided into multiple sample information clusters based on pre-set parameter association information, resulting in multiple working condition clusters, including: The pre-set parameter association information and the sample information set are input into the sparse subspace clustering model; Multiple operating condition clusters are determined based on the output of the sparse subspace clustering model.

6. The method according to claim 1, characterized in that, The constraint evaluation of each target sample, and the use of target samples whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant, includes: Based on a pre-defined evaluation index system that includes equipment performance constraints, safe operation constraints, and environmental emission constraints, the evaluation result for each target sample is determined. The target samples whose evaluation results meet the evaluation criteria are used to control the operation of the thermal power plant.

7. A dispatching device for a thermal power plant, characterized in that, The device includes: The information set acquisition module is used to acquire a sample information set of thermal power plants; The information segmentation module is used to segment multiple sample information in the sample information set based on pre-set parameter association information to obtain multiple working condition clusters; each working condition cluster includes at least a portion of sample information. The cluster center determination module is used to determine the cluster center of each operating condition cluster; wherein each operating condition cluster includes multiple intra-cluster samples. The target sample determination module is used to determine the intra-cluster sample with the smallest weighted distance to the cluster center from the intra-cluster samples as the target sample; The constraint evaluation module is used to perform constraint evaluation on each of the target samples, and to use the target samples whose evaluation results meet the evaluation conditions to control the operation of the thermal power plant.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images