Power generation enterprise operation index self-adaptive early warning system and method based on dynamic boundary

Abstract

The application provides a power generation enterprise operation index adaptive early warning system and method based on a dynamic boundary. The system comprises a data acquisition module for real-time acquisition and preprocessing of operation parameter data of a generator set and extraction of a multi-dimensional feature vector; a working condition recognition module for classification of a current operation state to a target working condition category according to a matching degree between the multi-dimensional feature vector and each working condition category in a preset historical working condition library, and dynamic updating of a category center of the target working condition category; a boundary calculation module for calling corresponding historical operation data from a historical database according to the current working condition category, constructing an operation feasible region boundary, and outputting corresponding dynamic boundary parameters; an early warning analysis module for generating a graded early warning signal according to a position deviation degree of current operation state data relative to the dynamic boundary parameters and a time variation trend of the position deviation degree; and an early warning pushing module for pushing the graded early warning signal to a user terminal.

Description

Technical Field

[0001] This invention relates to the field of power generation enterprise operation monitoring technology, and in particular to an adaptive early warning system and method for power generation enterprise operation indicators based on dynamic boundaries. Background Technology

[0002] In the operation and management of power generation enterprises, key operating indicators (such as main steam temperature, flue gas temperature, coal consumption for power supply, etc.) are monitored. Monitoring and early warning of emissions concentrations, etc., are crucial for ensuring the safe and economical operation of generating units. Traditional early warning methods employ fixed thresholds, triggering an alarm when the indicator value exceeds preset upper and lower limits. Existing technologies already include indicator early warning systems for power generation enterprises. For example, CN118674300A discloses a new energy indicator data management system that configures upper and lower limit fields in indicator information and displays them in red and issues pop-up alarms when they are exceeded; CN112862314B discloses a report-based management system for environmental indicators of thermal power plant production and operation, which also uses an indicator over-limit alarm module to display over-limit indicators in different colors; CN103366500B discloses an alarm method for power grid systems that establishes a four-dimensional indicator system encompassing safety, economy, quality, and environmental protection, sending alarm signals by comparing various monitoring indicators with corresponding preset thresholds.

[0003] The common feature of the aforementioned existing technologies is that the alarm threshold is a pre-configured fixed value or a preset level threshold, and it does not automatically adjust with changes in operating conditions. However, with deep peak shaving of thermal power units becoming the norm, fixed threshold early warning systems reveal serious adaptability deficiencies: 1. High false alarm rate: Under low load conditions, parameters such as flue gas temperature and main steam temperature are lower than the design values. If a fixed threshold under high load conditions is used, some normal operating conditions will be misjudged as abnormal, causing operators to frequently deal with invalid alarms, creating a "boy who cried wolf" effect. When a real serious abnormality occurs, it is ignored.

[0004] 2. Increased risk of underreporting: Under high load conditions or specific coal quality conditions, some indicators may be within the fixed threshold range but are already close to the danger boundary of that condition. The fixed threshold cannot identify this normal but dangerous state in advance, thus missing the best time for intervention.

[0005] 3. Confusion between safety and economic boundaries: Fixed thresholds only distinguish between normal and abnormal conditions, but cannot distinguish between safe but uneconomical and unsafe conditions, which makes it impossible for operators to make accurate decisions based on early warning information.

[0006] Furthermore, CN120373946A proposed a dynamic multi-dimensional quality evaluation system for a smart monitoring model of thermal power plants, which retrieves historical data under a given operating condition from a historical database through operating condition matching to participate in indicator calculation. However, the core purpose of this scheme is to evaluate the quality of the model itself, rather than to provide dynamic early warnings for operating indicators; its method of retrieving historical data after operating condition matching is a static query, failing to form a dynamically evolving operating boundary.

[0007] Therefore, there is an urgent need for an adaptive early warning method for power generation enterprises that can dynamically change with operating conditions, distinguish between safety boundaries and economic boundaries, and provide graded early warnings. Summary of the Invention

[0008] The purpose of this invention is to provide an adaptive early warning system and method for power generation enterprise operating indicators based on dynamic boundaries, aiming to solve the above-mentioned problems in the prior art.

[0009] This invention provides an adaptive early warning system for power generation enterprise operating indicators based on dynamic boundaries, comprising: The data acquisition module is used to acquire the operating parameter data of the generator set in real time, and to preprocess the operating parameter data to extract a multi-dimensional feature vector that represents the current operating condition. The working condition identification module is connected to the data acquisition module and is used to receive the multi-dimensional feature vector, classify the current operating state into the target working condition category according to the matching degree between the multi-dimensional feature vector and each working condition category in the preset historical working condition database, dynamically update the category center of the target working condition category, and output the identification result of the current working condition category. The boundary calculation module, connected to the working condition identification module, is used to receive the identification result, retrieve the corresponding historical operation data from the historical database according to the current working condition category, construct the operational feasibility domain boundary under the current working condition category based on the historical operation data, and output the corresponding dynamic boundary parameters. The early warning analysis module, connected to the data acquisition module and the boundary calculation module, is used to receive the dynamic boundary parameters, acquire the current operating status data in real time, and generate a graded early warning signal based on the degree of deviation of the current operating status data from the dynamic boundary parameters and the time trend of the degree of deviation. The early warning push module, connected to the early warning analysis module, is used to receive the graded early warning signals and push the graded early warning signals to the user terminal.

[0010] This invention provides an adaptive early warning method for power generation enterprise operating indicators based on dynamic boundaries, comprising: The data acquisition module acquires the generator set's operating parameter data in real time, and preprocesses the operating parameter data to extract a multi-dimensional feature vector that represents the current operating condition. The multi-dimensional feature vector is received by the working condition recognition module. Based on the matching degree between the multi-dimensional feature vector and each working condition category in the preset historical working condition database, the current operating state is classified into the target working condition category, and the category center of the target working condition category is dynamically updated, and the recognition result of the current working condition category is output. The boundary calculation module receives the identification result and retrieves the corresponding historical operation data from the historical database according to the current working condition category. Based on the historical operation data, it constructs the operational feasibility domain boundary under the current working condition category and outputs the corresponding dynamic boundary parameters. The early warning analysis module receives the dynamic boundary parameters and acquires the current operating status data in real time. It generates a graded early warning signal based on the degree of deviation of the current operating status data from the dynamic boundary parameters and the time trend of the degree of deviation. The warning push module receives the graded warning signals and pushes them to the user terminal.

[0011] This invention also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the above-described adaptive early warning method for power generation enterprise operating indicators based on dynamic boundaries.

[0012] This invention also provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor, implements the steps of the above-described adaptive early warning method for power generation enterprise operating indicators based on dynamic boundaries.

[0013] The embodiments of this invention can include the following beneficial effects: This invention proposes an adaptive early warning system and method for power generation enterprise operating indicators based on dynamic identification of operating boundaries. The system includes a data acquisition and feature extraction module, an online operating condition identification module, a dynamic boundary calculation module, a multi-level early warning module, and an early warning push module. The data acquisition and feature extraction module collects DCS operating parameters in real time and extracts multi-dimensional feature vectors. The online operating condition identification module uses Mahalanobis distance to classify the current operating condition into a historical operating condition database and dynamically updates the operating condition category. The dynamic boundary calculation module uses support vector data description algorithm to calculate the dynamic safety boundary and dynamic economic boundary under the current operating condition. The safety boundary and economic boundary are differentiated by setting different penalty parameters. The multi-level early warning module generates four levels of early warning signals (safety, attention, warning, and alarm) based on the normalized distance and rate of change from the current indicator status to the center of the safety boundary hypersphere and calculates the estimated arrival time at the boundary. The early warning push module pushes the early warning information to the user terminal. The embodiments of the present invention enable the early warning boundary to be adaptively adjusted according to the operating conditions, eliminating false alarms and missed alarms under varying operating conditions with a fixed threshold, and for the first time in the power generation field, decoupling the safety boundary and the economic boundary, effectively improving the accuracy of the operation early warning and the decision-making guidance value. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of an adaptive early warning system for power generation enterprise operation indicators based on dynamic boundaries, according to an embodiment of the present invention. Figure 2 This is a flowchart of an adaptive early warning method for power generation enterprise operation indicators based on dynamic boundaries, according to an embodiment of the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0017] System Implementation Examples According to embodiments of the present invention, an adaptive early warning system for power generation enterprise operating indicators based on dynamic boundaries is provided. Figure 1 This is a schematic diagram of an adaptive early warning system for power generation enterprise operating indicators based on dynamic boundaries, according to an embodiment of the present invention. Figure 1 As shown, the adaptive early warning system for power generation enterprise operating indicators based on dynamic boundaries according to an embodiment of the present invention specifically includes: The data acquisition module 10 is used to acquire the operating parameter data of the generator set in real time, and to preprocess the operating parameter data to extract a multi-dimensional feature vector representing the current operating condition. The multi-dimensional feature vector includes one or more of the following: unit load, main steam temperature, flue gas temperature, calorific value of coal fed into the furnace, coal mill operating combination status, main steam pressure, reheat steam temperature, condenser vacuum, and ambient temperature.

[0018] The operating condition identification module 12, connected to the data acquisition module, is used to receive the multi-dimensional feature vector, classify the current operating state into a target operating condition category based on the matching degree between the multi-dimensional feature vector and each operating condition category in a preset historical operating condition database, dynamically update the category center of the target operating condition category, and output the identification result of the current operating condition category; specifically used for: Obtain the category center of each operating condition category in the historical operating condition database; wherein, the historical operating condition database includes multiple operating condition categories, and each operating condition category has a corresponding category center vector; Calculate the distance between the multidimensional feature vector at the current time and the center of each working condition category, and determine the working condition category with the smallest distance as the candidate working condition category; The minimum distance is compared with a preset distance threshold. When the minimum distance is less than or equal to the distance threshold, the current operating state is classified into the candidate operating condition category, and the category center of the candidate operating condition category is dynamically updated using the multidimensional feature vector at the current moment. When the minimum distance is greater than the distance threshold, it is determined that the current operating state does not belong to any existing operating condition category in the historical operating condition library, and a new operating condition category is added to the historical operating condition library, with the multidimensional feature vector at the current moment as the category center of the newly added operating condition category.

[0019] Boundary calculation module 14, connected to the working condition identification module, is used to receive the identification result, retrieve corresponding historical operating data from the historical database according to the current working condition category, construct the operational feasibility domain boundary under the current working condition category based on the historical operating data, and output the corresponding dynamic boundary parameters; specifically used for: Obtain the current working condition category output by the working condition identification module; Based on the current operating condition category, all historical operating indicator data belonging to that operating condition category are extracted from the historical database to form a historical sample set under that operating condition category; wherein, each historical sample is an indicator vector at a certain moment under that operating condition category; Based on the historical sample set, a hypersphere is constructed using the Support Vector Data Description (SVDD) algorithm, and the boundary of the hypersphere is used as the boundary of the feasible operating domain under the current working condition category. When constructing the hypersphere, some historical samples are allowed to be located outside the hypersphere, and a kernel function is used to map the historical samples from the original space to a high-dimensional feature space. The center and radius of the hypersphere are calculated, and dynamic boundary parameters containing the center and radius are output. The dynamic boundary parameters include safety boundary parameters and economic boundary parameters. The safety boundary parameters are trained with a first penalty parameter based on safety indicators in historical samples, and the economic boundary parameters are trained with a second penalty parameter based on economic indicators in historical samples. The first penalty parameter is smaller than the second penalty parameter.

[0020] The early warning analysis module 16, connected to the data acquisition module and the boundary calculation module, is used to receive the dynamic boundary parameters, acquire current operating status data in real time, and generate graded early warning signals based on the degree of deviation of the current operating status data from the dynamic boundary parameters and the time trend of the degree of deviation; specifically used for: Obtain the dynamic boundary parameters output by the boundary calculation module; Obtain the current operating status data output by the data acquisition module; wherein, the current operating status data includes the indicator vector at the current moment; Calculate the distance from the index vector at the current moment to the center of the sphere, and determine the degree of deviation of the current running state from the boundary of the running feasible domain based on the relationship between the distance and the radius; Calculate the rate of change of the position deviation over time; wherein the rate of change is used to characterize the speed at which the current operating state evolves toward or away from the boundary of the feasible operating domain; A graded early warning signal is generated based on the degree of positional deviation and the rate of change.

[0021] The early warning push module 18, connected to the early warning analysis module, is used to receive the graded early warning signals and push the graded early warning signals to the user terminal; specifically, it is used for: Obtain the graded early warning signals output by the early warning analysis module; The corresponding warning level and color code are determined based on the tiered warning signals; wherein, the warning level includes security level, attention level, warning level and alarm level; Obtain the relative position information between the current running status data and the boundary of the feasible running domain; The graded warning signal, the color code corresponding to the warning level, the relative position information, and the estimated time to reach the boundary or the duration of time that has exceeded the boundary are assembled into warning information. The warning information will be pushed to the user terminal.

[0022] The system also includes a boundary smoothing update module, specifically used for: Monitor the update status of the working condition identification module to the working condition category center; When preset triggering conditions are met, a recalculation command is sent to the boundary calculation module to trigger the boundary calculation module to recalculate the operational feasible domain boundary of the current working condition category and obtain the updated boundary parameters output after recalculation. The triggering conditions include: the number of newly added historical samples in the current working condition category reaches a preset number threshold, the cumulative offset of the category center of the current working condition category exceeds a preset offset threshold, and the time interval between the current moment and the last recalculation of the operational feasible domain boundary exceeds a preset time threshold. The updated boundary parameters and the currently used boundary parameters are weighted and fused according to a preset weight ratio using a weighted smoothing method, and the fused parameters are used as the final boundary parameters to replace the currently used boundary parameters.

[0023] The following describes in detail the above-mentioned technical solution of the present invention with reference to the specific circumstances of the adaptive early warning system for power generation enterprise operation indicators based on dynamic boundaries in the embodiments of the present invention.

[0024] The key to the embodiments of the present invention lies in: 1. By using online operating condition identification and SVDD dynamic boundary calculation, the warning thresholds of each key indicator are automatically adapted to the current operating conditions, fundamentally eliminating the problem of false alarms and missed alarms caused by fixed thresholds under changing operating conditions.

[0025] 2. By calculating the safety boundary and economic boundary separately, the safety early warning and economic early warning are decoupled, providing operators with more differentiated decision-making information.

[0026] 3. Through a two-dimensional hierarchical early warning mechanism based on location and trend, it provides progressive early warning from safety to alarm and calculates the estimated time to reach the boundary, enabling operators to anticipate risks in advance and respond calmly, rather than responding passively when the standard is exceeded.

[0027] This invention proposes an adaptive early warning system for power generation enterprise operating indicators based on dynamic identification of operating boundaries. The system includes a data acquisition and feature extraction module, an online operating condition identification module, a dynamic boundary calculation module, a multi-level early warning module, and an early warning push module. Details are as follows: I. Data Acquisition and Feature Extraction Module This module is used to collect real-time operating parameter data, environmental data, and coal quality data from the generator set's DCS system, and extract a multi-dimensional feature vector to characterize the current operating conditions. The multi-dimensional feature vector includes the following dimensions: unit load, main steam temperature, flue gas temperature, received lower heating value of coal, and the operating combination status of the coal mills (which coal mills are operating can be represented by binary encoding). Preferably, it may also include dimensions such as main steam pressure, reheat steam temperature, condenser vacuum, and ambient temperature.

[0028] The selection of the above-mentioned feature dimensions mainly takes into account two aspects: first, it can fully distinguish different operating conditions (such as different load sections, different coal quality conditions, and different equipment combinations); second, each feature dimension has a reliable data source in the DCS system and has good engineering feasibility.

[0029] II. Online Operating Condition Recognition Module This module is used to classify the current operating status into a certain operating condition category in the historical operating condition database based on the extracted multidimensional feature vector using an online classification algorithm based on density clustering, and dynamically update the operating condition category center.

[0030] Unlike traditional offline clustering (which involves collecting a large amount of historical data, performing clustering once, and obtaining fixed operating condition categories), this invention adopts an online classification mechanism that can adaptively identify new operating conditions that have not yet appeared and update the center position of the operating condition categories in real time, so that the operating condition database can be continuously enriched and improved as the unit operates for a longer period of time.

[0031] Specifically, suppose the historical operating condition database contains K operating condition categories, and each operating condition category... It consists of the multidimensional feature vectors of all historical samples in this category. For the multidimensional feature vector collected at the current moment... Calculate the Mahalanobis distance between each of the K working condition categories and its center: ; in, Let be the center vector of the k-th working condition category. Let be the covariance matrix of the k-th work condition category. The advantage of Mahalanobis distance over Euclidean distance is that it considers the correlation between feature dimensions and the dimensional differences between different dimensions, avoiding the bias caused by load (order of magnitude). ) and coal calorific value (order of magnitude) This addresses the issue of weight imbalance in distance calculations for features of different magnitudes, such as (e.g., features of different magnitudes).

[0032] like Then, the current working condition is assigned to the category with the smallest distance, and the center of that category is updated using an exponentially weighted moving average: ; in, The learning rate is set to a small value to ensure the stability of the working condition category center and avoid a large shift in the category center caused by a single abnormal sampling point. The value that is closest to the current working condition among all K historical working condition categories. This is the preset distance threshold.

[0033] like If a condition is identified as a new operating condition, a new operating condition category is added to the operating condition database, and the current feature vector is used as the initial center of this category. This mechanism enables the operating condition database of this embodiment to have self-expanding capabilities, allowing it to adapt to long-term changes such as unit equipment upgrades and coal type switching.

[0034] III. Dynamic Boundary Calculation Module This module is used to extract historical operating data of key indicators under the current operating condition category output by the real-time identification module from the historical database, and calculate the dynamic feasible region boundary of each key indicator under the operating condition using the Support Vector Data Description (SVDD) algorithm. The basic principle of the SVDD algorithm is: for a given sample set, the samples are mapped to a high-dimensional feature space through a nonlinear mapping, and a hypersphere with the smallest volume is found within it, such that as many samples as possible are contained within the hypersphere. The boundary of this hypersphere is the boundary of the normal range; samples falling outside the boundary are considered abnormal or outliers.

[0035] Specifically, for N historical samples under the kth working condition category Each of them Let be the index vector at a certain moment under this operating condition, containing the values ​​of m key indicators, such as main steam temperature, flue gas temperature, and coal consumption for power generation. The optimization problem of SVDD is: ; Constraints: ; Where R is the radius of the hypersphere, and o is the center of the hypersphere. C is a slack variable (allowing a small number of samples outside the hypersphere to enhance the algorithm's robustness to noise), and C is a penalty parameter (controlling the penalty for samples outside the boundary). It is a non-linear mapping function.

[0036] The inner product operation is transformed into a kernel function calculation using the Gaussian radial basis kernel function: ; in, This refers to the kernel width parameter. The kernel width parameter varies depending on the operating condition category. The determination is made adaptively based on the distribution density of historical data under this working condition category: ; Right now The average distance from all samples in the k-th working condition category to their nearest neighbor is taken. This adaptive setting results in a smaller kernel width (fine boundaries) in data-dense regions and a larger kernel width (smooth boundaries) in data-sparse regions, avoiding the boundary distortion problem caused by a globally uniform kernel width when the data distribution is uneven.

[0037] The SVDD optimization problem is transformed into a dual problem using the Lagrange multiplier method and then solved to obtain the center of the hypersphere. and radius .

[0038] For the index vector at the current moment Calculate its high-dimensional spatial distance to the center of the hypersphere: ; when When the current indicator is within the normal range, it is determined that the current indicator is within the normal range; when When the current indicator is determined to exceed the dynamic boundary, an early warning is triggered.

[0039] Furthermore, embodiments of the present invention further distinguish dynamic boundaries into two different types: safety boundaries and economic boundaries. For the safety boundary, safety-related parameters from historical samples are selected to construct a safety index vector. This includes, but is not limited to, parameters directly related to equipment safety such as main steam temperature, boiler wall temperature, turbine bearing vibration, and cylinder expansion difference. Smaller penalty parameters are used. The SVDD model is trained to have a large hypersphere radius and relatively loose boundaries, ensuring that any potential security risk points are contained within the boundaries, and it is better to have false positives than false negatives.

[0040] For the economic boundary, economic indicators from historical samples are selected to construct an economic indicator vector. This includes, but is not limited to, indicators characterizing the unit's economic performance such as coal consumption for power supply, plant power consumption rate, and turbine heat consumption. Larger penalty parameters are used. The SVDD model is trained to have a smaller hypersphere radius and a more compact boundary, including only the subset of samples with the best historical economic performance within the economic boundary.

[0041] In other words, the safety boundary answers the question of whether the current operation is safe, while the economic boundary answers the question of whether the current operation is economical. For operators, simultaneously knowing information from both dimensions—for example, being within the safety boundary but exceeding the economic boundary—means that the current operating condition is safe but not economically efficient, requiring optimization and adjustment but not immediate action; while exceeding the safety boundary means that immediate intervention is needed. This level of information differentiation cannot be provided by traditional fixed threshold early warning systems.

[0042] IV. Multi-level early warning module This module is used to generate graded early warning signals based on the positional relationship and rate of change between the current operating status and the dynamic operating boundary.

[0043] First, calculate the current indicator vector. Normalized distance to the center o of the dynamic boundary hypersphere, and relative distance to the surface of the hypersphere: ; in, This indicates that the current state has exceeded the dynamic boundary. This indicates that the current state is inside the boundary; The larger the value, the further it deviates from the normal range, and the more serious the deviation.

[0044] Secondly, calculate the rate of change of the current indicator, that is, the rate of change from the boundary: ; in The sampling interval is... This indicates that the current situation is deteriorating towards the outer boundary. This indicates that the current state is improving towards the inside of the boundary.

[0045] According to the embodiments of the present invention and Combinatorial logic generates a four-level warning: 1. Safety (Green): and The current state is far from the boundary, and the trend is positive or stable, requiring no intervention.

[0046] 2. Pay attention (yellow): or and The current status is near the boundary, or although far from the boundary, it is rapidly approaching the boundary, requiring attention from operators but not immediate action.

[0047] 3. Warning (Orange): The current state has just exceeded the boundary, which is a minor boundary violation. The operators should take measures in the near future.

[0048] 4. Alarm (Red): or The acceleration remains consistently above the preset threshold. The current state is severely out of control or rapidly deteriorating, and operators must take immediate action.

[0049] Among them, "Safe" and "Concern" correspond to the state within the safety boundary, while "Warning" and "Alarm" correspond to the state outside the safety boundary. When the current state exceeds the economic boundary but is still within the safety boundary, the warning level is further superimposed with an economic concern indicator. The color of the indicator remains consistent with the safety level but includes additional economic information.

[0050] In addition, when and At present (not yet crossed the boundary but deteriorating), the estimated time to reach the boundary is: ; The estimated time is directly pushed to the operations staff, so that they know exactly how much time window is left to take intervention measures, thereby rationally prioritizing operations.

[0051] when When (out of bounds), output the duration that has exceeded the boundary. This helps operators assess the urgency of a problem.

[0052] Preferably, when generating graded early warning signals, the multi-level early warning module also considers the historical early warning frequency factor of the current operating condition category. : ; when Higher than the preset threshold When this condition is classified as a high-risk condition, it is subject to the same conditions. Under certain conditions, the warning level can be raised by one level, or the warning trigger threshold can be lowered.

[0053] V. Early Warning Push Module This module is used to push graded early warning signals and auxiliary information to the user terminals of on-duty personnel, shift leaders, and production management personnel. The pushed content includes at least: the name and characteristic description of the current operating condition category, a visual diagram of the relative position of each key indicator value with the dynamic safety boundary and economic boundary, the graded early warning level and its corresponding color code, and the estimated time to reach the boundary or the duration of the boundary violation.

[0054] VI. Boundary Smoothing Update Module To avoid frequent and significant changes in dynamic boundaries due to frequent updates of the work condition category center, this embodiment of the invention also includes a boundary smoothing update module. When the work condition category center is updated, the SVDD boundary of that work condition category is not immediately recalculated. Instead, boundary recalculation is triggered only when any of the following conditions are met: 1. The number of new samples in this working condition category has reached the preset threshold. ; 2. The cumulative offset of the center in this working condition exceeds the preset threshold. ; 3. The current time is more than the time since the last boundary calculation, exceeding a preset time threshold. .

[0055] When recalculating the boundary, a weighted smoothing method is used to update the boundary parameters: ; in, The final boundary radius adopted after updating for the k-th working condition category. The boundary radius for the k-th working condition category is recalculated based on the latest historical data. The boundary radius currently in use before updating the k-th working condition category. This is the smoothing coefficient.

[0056] Method Implementation Examples According to embodiments of the present invention, an adaptive early warning method for power generation enterprise operating indicators based on dynamic boundaries is provided. Figure 2 This is a flowchart of the adaptive early warning method for power generation enterprise operation indicators based on dynamic boundaries, according to an embodiment of the present invention. Figure 2 As shown, the adaptive early warning method for power generation enterprise operation indicators based on dynamic boundaries according to an embodiment of the present invention specifically includes: Step S201: The operating parameter data of the generator set is acquired in real time through the data acquisition module, and the operating parameter data is preprocessed to extract a multi-dimensional feature vector representing the current operating condition. Step S202: Receive the multi-dimensional feature vector through the working condition recognition module; classify the current operating state into the target working condition category based on the matching degree between the multi-dimensional feature vector and each working condition category in the preset historical working condition database; dynamically update the category center of the target working condition category; and output the recognition result of the current working condition category. Step S203: Receive the identification result through the boundary calculation module, retrieve the corresponding historical operation data from the historical database according to the current working condition category, construct the operational feasibility domain boundary under the current working condition category based on the historical operation data, and output the corresponding dynamic boundary parameters. Step S204: Receive the dynamic boundary parameter through the early warning analysis module, acquire the current operating status data in real time, and generate a graded early warning signal based on the degree of deviation of the current operating status data from the dynamic boundary parameter and the time trend of the degree of deviation. Step S205: Receive the graded early warning signal through the early warning push module and push the graded early warning signal to the user terminal.

[0057] The following describes in detail the above-mentioned technical solution of the present invention with reference to the specific circumstances of the adaptive early warning method for power generation enterprise operation indicators based on dynamic boundaries in the embodiments of the present invention.

[0058] This invention proposes an adaptive early warning method for power generation enterprise operating indicators based on dynamic identification of operating boundaries, comprising the following steps: S1: Data Acquisition and Feature Extraction. Real-time acquisition of operating parameter data, environmental data, and coal quality data from the generator set's DCS system; extraction of multi-dimensional feature vectors to characterize the current operating conditions, including at least unit load, main steam temperature, flue gas temperature, calorific value of coal fed into the furnace, and the operating combination status of the coal mill.

[0059] S2: Online working condition identification. Calculate the Mahalanobis distance between the current multidimensional feature vector and the centers of each category in the historical working condition database, assign the current state to the working condition category with the smallest distance, and dynamically update the center of that category; if the distance between the current state and all category centers exceeds a preset threshold, then add a new working condition category to the working condition database.

[0060] S3: Dynamic Boundary Calculation. For the current operating condition category, historical operating data of key indicators under this category are extracted from the historical database. Support Vector Data Description (SVM) algorithm is used to calculate the dynamic safety boundary and dynamic economic boundary for this operating condition. The safety boundary is based on penalty parameters. The training yields relatively loose boundaries; the economic boundary is based on penalty parameters. The training yielded relatively compact boundaries.

[0061] S4: Graded early warning judgment. Calculate the normalized distance from the current indicator status to the center of the dynamic safety boundary hypersphere. and the speed of change ,according to and The combinational logic generates four levels of tiered early warning signals: safety, attention, warning, and alarm.

[0062] At the same time, based on the normalized distance of the current state relative to the economic boundary Additional economic information is provided.

[0063] S5: Early Warning Information Push. Pushes tiered early warning signals, along with corresponding dynamic boundary information, current indicator values, and estimated arrival times at the boundary to the user terminal.

[0064] The embodiments of the present invention are method embodiments corresponding to the system embodiments described above. The specific operations of each step can be understood by referring to the description of the system embodiments, and will not be repeated here.

[0065] In summary, compared with the prior art, the embodiments of the present invention have the following beneficial effects: 1. Fundamentally solves the problem of false alarms and missed alarms under varying operating conditions with fixed thresholds: Existing technologies use fixed thresholds or preset level thresholds, which cannot adaptively adjust under dynamic operating conditions such as deep peak shaving and coal quality fluctuations. This invention, through online operating condition identification and SVDD dynamic boundary calculation, enables the warning boundary to adaptively adjust with the operating conditions. Each operating condition category has its own exclusive boundary, completely eliminating the interference of operating condition changes on the accuracy of warnings.

[0066] 2. For the first time, the decoupling of safety boundary and economic boundary has been achieved in the field of power generation indicator early warning: The embodiments of this invention separate the judgment of whether it is safe and whether it is economic by training the safety SVDD model and the economic SVDD model respectively. Operators can simultaneously know different combinations of states such as whether the current operation is safe but not economical or whether the current operation is both safe and economical, which provides an information basis for accurate decision-making. This is completely impossible to achieve in traditional fixed threshold systems.

[0067] 3. The dual-dimensional hierarchical early warning mechanism based on location and trend provides a progressive early warning from mild to urgent: Traditional systems can only trigger alarms the instant the standard is exceeded, which is a reactive measure. This embodiment of the invention calculates the distance (location) from the current state to the boundary and the rate of change (trend), and can predict the time to reach the boundary before the standard is exceeded, thus realizing the early warning X minutes before reaching the boundary. This lead time has important practical value for operators to reasonably arrange intervention measures.

[0068] 4. The self-expanding working condition database and the smooth update mechanism of the boundary ensure the long-term adaptability of the system: The embodiments of the present invention not only identify existing working condition categories, but also automatically identify new working conditions and add new categories; at the same time, through delayed update and smooth switching mechanisms, frequent and large fluctuations in the boundary are avoided. These designs enable the system to remain effective under conditions such as unit equipment modification, coal type change, and long-term operation aging, without the need for manual reconfiguration of any parameters.

[0069] Device Example 1 This invention provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, performs the steps described in the method embodiment.

[0070] Device Example 2 This invention provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor, performs the steps described in the method embodiment.

[0071] The computer-readable storage media described in this embodiment include, but are not limited to, ROM, RAM, disk, or optical disk.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive early warning system for power generation enterprise operating indicators based on dynamic boundaries, characterized in that, include: The data acquisition module is used to acquire the operating parameter data of the generator set in real time, and to preprocess the operating parameter data to extract a multi-dimensional feature vector that represents the current operating condition. The working condition identification module is connected to the data acquisition module and is used to receive the multi-dimensional feature vector, classify the current operating state into the target working condition category according to the matching degree between the multi-dimensional feature vector and each working condition category in the preset historical working condition database, dynamically update the category center of the target working condition category, and output the identification result of the current working condition category. The boundary calculation module, connected to the working condition identification module, is used to receive the identification result, retrieve the corresponding historical operation data from the historical database according to the current working condition category, construct the operational feasibility domain boundary under the current working condition category based on the historical operation data, and output the corresponding dynamic boundary parameters. The early warning analysis module, connected to the data acquisition module and the boundary calculation module, is used to receive the dynamic boundary parameters, acquire the current operating status data in real time, and generate a graded early warning signal based on the degree of deviation of the current operating status data from the dynamic boundary parameters and the time trend of the degree of deviation. The early warning push module, connected to the early warning analysis module, is used to receive the graded early warning signals and push the graded early warning signals to the user terminal.

2. The system according to claim 1, characterized in that, The multidimensional feature vector includes one or more of the following: unit load, main steam temperature, flue gas temperature, calorific value of coal fed into the furnace, coal mill operating combination status, main steam pressure, reheat steam temperature, condenser vacuum, and ambient temperature.

3. The system according to claim 1, characterized in that, The operating condition identification module is specifically used for: Obtain the category center of each operating condition category in the historical operating condition database; wherein, the historical operating condition database includes multiple operating condition categories, and each operating condition category has a corresponding category center vector; Calculate the distance between the multidimensional feature vector at the current time and the center of each working condition category, and determine the working condition category with the smallest distance as the candidate working condition category; The minimum distance is compared with a preset distance threshold. When the minimum distance is less than or equal to the distance threshold, the current operating state is classified into the candidate operating condition category, and the category center of the candidate operating condition category is dynamically updated using the multidimensional feature vector at the current moment. When the minimum distance is greater than the distance threshold, it is determined that the current operating state does not belong to any existing operating condition category in the historical operating condition library, and a new operating condition category is added to the historical operating condition library, with the multidimensional feature vector at the current moment as the category center of the newly added operating condition category.

4. The system according to claim 1, characterized in that, The boundary calculation module is specifically used for: Obtain the current working condition category output by the working condition identification module; Based on the current operating condition category, all historical operating indicator data belonging to that operating condition category are extracted from the historical database to form a historical sample set under that operating condition category; wherein, each historical sample is an indicator vector at a certain moment under that operating condition category; Based on the historical sample set, a hypersphere is constructed using the Support Vector Data Description (SVDD) algorithm, and the boundary of the hypersphere is used as the boundary of the feasible operating domain under the current working condition category. When constructing the hypersphere, some historical samples are allowed to be located outside the hypersphere, and a kernel function is used to map the historical samples from the original space to a high-dimensional feature space. The center and radius of the hypersphere are calculated, and dynamic boundary parameters containing the center and radius are output. The dynamic boundary parameters include safety boundary parameters and economic boundary parameters. The safety boundary parameters are trained with a first penalty parameter based on safety indicators in historical samples, and the economic boundary parameters are trained with a second penalty parameter based on economic indicators in historical samples. The first penalty parameter is smaller than the second penalty parameter.

5. The system according to claim 4, characterized in that, The early warning analysis module is specifically used for: Obtain the dynamic boundary parameters output by the boundary calculation module; Obtain the current operating status data output by the data acquisition module; wherein, the current operating status data includes the indicator vector at the current moment; Calculate the distance from the index vector at the current moment to the center of the sphere, and determine the degree of deviation of the current running state from the boundary of the running feasible domain based on the relationship between the distance and the radius; Calculate the rate of change of the position deviation over time; wherein the rate of change is used to characterize the speed at which the current operating state evolves toward or away from the boundary of the feasible operating domain; A graded early warning signal is generated based on the degree of positional deviation and the rate of change.

6. The system according to claim 1, characterized in that, The early warning push module is specifically used for: Obtain the graded early warning signals output by the early warning analysis module; The corresponding warning level and color code are determined based on the tiered warning signals; wherein, the warning level includes security level, attention level, warning level and alarm level; Obtain the relative position information between the current running status data and the boundary of the feasible running domain; The graded warning signal, the color code corresponding to the warning level, the relative position information, and the estimated time to reach the boundary or the duration of time that has exceeded the boundary are assembled into warning information. The warning information will be pushed to the user terminal.

7. The system according to claim 1, characterized in that, The system also includes a boundary smoothing update module, specifically used for: Monitor the update status of the working condition identification module to the working condition category center; When preset triggering conditions are met, a recalculation command is sent to the boundary calculation module to trigger the boundary calculation module to recalculate the operational feasible domain boundary of the current working condition category and obtain the updated boundary parameters output after recalculation. The triggering conditions include: the number of newly added historical samples in the current working condition category reaches a preset number threshold, the cumulative offset of the category center of the current working condition category exceeds a preset offset threshold, and the time interval between the current moment and the last recalculation of the operational feasible domain boundary exceeds a preset time threshold. The updated boundary parameters and the currently used boundary parameters are weighted and fused according to a preset weight ratio using a weighted smoothing method, and the fused parameters are used as the final boundary parameters to replace the currently used boundary parameters.

8. An adaptive early warning method for power generation enterprise operation indicators based on dynamic boundaries, characterized in that, include: The data acquisition module acquires the generator set's operating parameter data in real time, and preprocesses the operating parameter data to extract a multi-dimensional feature vector that represents the current operating condition. The multi-dimensional feature vector is received by the working condition recognition module. Based on the matching degree between the multi-dimensional feature vector and each working condition category in the preset historical working condition database, the current operating state is classified into the target working condition category, and the category center of the target working condition category is dynamically updated, and the recognition result of the current working condition category is output. The boundary calculation module receives the identification result and retrieves the corresponding historical operation data from the historical database according to the current working condition category. Based on the historical operation data, it constructs the operational feasibility domain boundary under the current working condition category and outputs the corresponding dynamic boundary parameters. The early warning analysis module receives the dynamic boundary parameters and acquires the current operating status data in real time. It generates a graded early warning signal based on the degree of deviation of the current operating status data from the dynamic boundary parameters and the time trend of the degree of deviation. The warning push module receives the graded warning signals and pushes them to the user terminal.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the adaptive early warning method for power generation enterprise operating indicators based on dynamic boundaries as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an information transmission implementation program, which, when executed by a processor, implements the steps of the adaptive early warning method for power generation enterprise operating indicators based on dynamic boundaries as described in claim 8.

Images