An industrial steam automatic regulation system operation anomaly detection method

Abstract

The present application relates to the technical field of anomaly detection, in particular to an industrial steam automatic regulation system operation anomaly detection method. The method comprises: combining steam heat transfer delay characteristics to complete delay alignment, constructing a reaction condition mode cluster, matching to obtain a current reference cluster, obtaining the influence contribution weight value of each operation parameter dimension, weighting and fusing the preprocessed reaction kettle process parameters based on the influence contribution weight value to output a current heat demand index value, calculating a current steam heat supply anomaly index value based on the deviation of the current heat demand index value from the heat demand reference value and the cluster contour coefficient difference before and after the current state feature vector is added to the current reference cluster, and detecting the current steam regulation for anomaly and anomaly type based on the current steam heat supply anomaly index value. The present application can improve the accuracy of steam regulation anomaly detection and accurately identify the steam heat supply deficiency or excess anomaly type.

Description

Technical Field

[0001] This invention relates to the field of anomaly detection technology, and specifically to a method for detecting operational anomalies in an industrial steam automatic regulation system. Background Technology

[0002] In industrial production processes, industrial steam automatic control systems, as core thermal energy control units, are widely used in production stages such as reactor heating, material temperature rise, and process insulation. For example, in continuous phosphate production scenarios, industrial steam automatic control systems typically use steam jackets to precisely regulate the steam input to maintain the temperature conditions required for material reactions, promote raw material conversion, and ensure product quality. Furthermore, due to the continuous and dynamic nature of industrial production processes, the heat load inside the reaction equipment changes with each production stage. This means the heat load inside the reaction equipment is affected by various factors such as material input, material concentration, reaction rate, stirring status, and environmental conditions. Simultaneously, the industrial steam supply process is also affected by factors such as steam pressure, flow rate fluctuations, pipeline transmission losses, and the response characteristics of regulating valves. This results in a time lag and dynamic coupling between the steam input heat and the internal temperature response of the equipment. Therefore, industrial steam automatic control systems need to dynamically adjust the steam supply process based on the current operating status of the production equipment and changes in actual heat demand to maintain stable production operation and improve energy efficiency. Furthermore, since the stable operation of industrial steam automatic regulation systems is the foundation for achieving continuous and refined production, any abnormal heating conditions in the system will directly trigger a chain of problems such as production condition drift (e.g., causing equipment temperature fluctuations, reduced production efficiency, and decreased phosphate product conversion efficiency) and excessive energy consumption. Therefore, it is currently crucial to detect abnormalities in the operation of industrial steam automatic regulation systems.

[0003] In existing technologies, methods for detecting operational anomalies in industrial steam automatic control systems during continuous phosphate production typically use the deviation between the set temperature and the actual equipment temperature as the criterion for anomaly identification. However, existing anomaly detection methods suffer from a series of problems, including low detection accuracy and weak ability to determine the direction of anomalies. Specifically: Firstly, the thermal inertia within industrial reaction equipment, coupled with the combined effects of dynamic changes in steam-side parameters, dynamic changes in the state of materials within the reactor, and the inherent time delay characteristics of heat transfer, creates a dynamic correlation between steam input, heat transfer, and changes in production load. This means that relying solely on real-time temperature deviation data cannot accurately identify the matching state between the current steam input and the actual heat demand of the reactor, directly leading to missed or false anomaly detections. It also fails to promptly capture core anomalies such as insufficient steam supply, excessive steam input, or delayed control response (e.g., when the set temperature is too high, the actual temperature is too low). When changes occur in production load, material status, or process stage, the original regulation strategy may result in insufficient steam supply, excessive steam input, or delayed regulation response, leading to equipment temperature fluctuations, reduced production efficiency, and energy waste. Secondly, existing conventional anomaly detection methods generally determine anomalies by whether the temperature deviation exceeds a threshold, lacking the ability to analyze the changing patterns of heating status during steam regulation. They cannot further distinguish whether the current anomaly is caused by insufficient heat supply or by excess heat overflow, and cannot provide clear guidance on the direction of anomaly deviation for subsequent steam supply adjustments. This makes the detection results completely unsuitable for supporting the closed-loop optimization operation of the regulation system. Therefore, how to adaptively construct a heat demand matching model that adapts to multiple operating conditions to improve the accuracy of anomaly detection and the ability to distinguish the direction of anomaly deviation, thereby reducing the probability of anomalies such as insufficient steam supply and excessive steam input, and improving the control precision of the automatic steam regulation system, has become an urgent problem to be solved. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method for detecting operational anomalies in an industrial steam automatic regulation system. The specific technical solution adopted is as follows: One embodiment of the present invention provides a method for detecting operational anomalies in an industrial steam automatic regulation system, comprising the following steps: In the phosphate production process, the process operating parameters of the pretreatment reactor, the internal temperature of the pretreatment reactor, and the heat of the pretreatment steam input into the reactor are acquired in real time. Based on the pre-tested steam heat transfer delay time, the heat of the pretreatment steam is delayed and aligned to obtain the aligned pretreatment steam heat corresponding to each production moment. The internal temperature sequence of the pretreatment historical reactor is divided into historical temperature segments according to the temperature change trend. Based on the starting temperature of each historical temperature segment and the corresponding average heat value of the aligned pretreatment steam, a historical feature vector corresponding to the historical temperature segment is constructed. Based on the distance between the historical feature vectors, the historical temperature segments are clustered to obtain reaction condition pattern clusters. The current reference cluster is obtained by matching the current state feature vector constructed based on the current internal temperature of the pretreatment reactor and the corresponding aligned pretreatment steam heat. The steam input thermal response coefficient of each segment is calculated based on the initial and final temperatures of the historical temperature segments within the current reference cluster and the aligned pretreatment steam heat. The difference in thermal response coefficients between historical temperature segment pairs formed by pairwise combinations of historical temperature segments within the current reference cluster and the clustering distance are used to calculate the degree of difference in thermal response coefficients between each pair of historical temperature segments. The correlation between the degree of difference in thermal response coefficients formed by all the degree of difference in thermal response coefficients and the degree of change sequence of each operating parameter dimension under the current reference cluster is used to obtain the influence contribution weight value of each operating parameter dimension under the current reference cluster. Based on the influence contribution weight value, the pretreatment reactor process parameters of the corresponding dimension at the current production time are weighted and fused to output the current heat demand index value at the current production time. Based on the deviation of the current heat demand index value from the heat demand reference value and the difference in the cluster profile coefficient before and after the current state feature vector is added to the current reference cluster, the current steam heating anomaly index value is calculated. Based on the current abnormal steam heating index values, detect the abnormality and abnormality type of the current steam regulation.

[0005] Beneficial effects: Based on the pre-tested steam heat transfer delay time, this invention performs delay alignment on the pre-treated steam heat to obtain the aligned pre-treated steam heat corresponding to each production moment. The internal temperature sequence of the pre-treated historical reactor is divided according to the temperature change trend. Based on the historical feature vector distance, the historical temperature segments are clustered to obtain reaction condition pattern clusters. The current reference cluster is matched to obtain the influence contribution weight value of each operating parameter dimension under the current reference cluster. Based on the influence contribution weight value, the corresponding dimension of the pre-treated reactor process parameters at the current production moment is weighted and fused to output the current heat demand index value at the current production moment. Based on the deviation of the current heat demand index value from the heat demand reference value and the difference in cluster profile coefficient before and after the current state feature vector is added to the current reference cluster, the current steam heating anomaly index value is calculated. Based on the current steam heating anomaly index value, the current steam regulation is detected for anomalies and anomaly types. Furthermore, the phosphate production steam regulation anomaly detection method provided by this invention, by combining the steam heat transfer delay characteristics to complete delay alignment, constructing a clustering pattern for each reaction stage, and adaptively obtaining the dynamic influence contribution weight of each operating parameter, can get rid of the limitations of traditional methods that rely solely on temperature deviation judgment, effectively avoid misjudgment problems caused by dynamic changes in operating conditions, significantly improve the accuracy of steam regulation anomaly detection, and can also accurately identify the anomaly type of insufficient or excessive steam heating through abnormal indicators, providing a clear and targeted basis for the dynamic optimization and regulation of steam supply, ensuring stable operation of the reaction process and reducing unnecessary energy consumption. Attached Figure Description

[0006] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or 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 of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0007] Figure 1 This is a flowchart of a method for detecting operational anomalies in an industrial steam automatic regulation system according to the present invention. Detailed Implementation

[0008] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the protection scope of the embodiments of the present invention.

[0009] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art.

[0010] This embodiment provides a method for detecting abnormal operation of an industrial steam automatic regulation system, which is described in detail below: like Figure 1 As shown, the method for detecting abnormal operation of the industrial steam automatic regulation system includes the following steps: Step S001: During the phosphate production process, the process operating parameters of the pretreatment reactor, the internal temperature of the pretreatment reactor, and the heat of the pretreatment steam input into the reactor are acquired in real time. Based on the pre-tested steam heat transfer delay time, the heat of the pretreatment steam is delayed and aligned to obtain the aligned pretreatment steam heat corresponding to each production moment.

[0011] In the phosphate production reaction process, in order to analyze the relationship between the steam input status and the changes in the heat demand of the reactor, it is first necessary to conduct multi-dimensional data monitoring and collection on the steam heating process and the reaction status inside the reactor. The specific process is as follows: Steam temperature, steam flow, and steam pressure sensors are installed at the inlet of the steam delivery pipeline to collect the temperature, mass flow rate, and pressure of the input steam in real time, characterizing the current steam heating capacity. The data directly collected by the sensors is recorded as raw data, thus obtaining the raw temperature, raw mass flow rate, and raw pressure of the input steam. A temperature sensor is installed inside the reactor to collect data on material temperature changes during the reaction process in real time, recording this as the raw reactor internal temperature. Rotation speed detection modules, flow rate detection modules, and concentration detection modules are installed at the reactor's stirring mechanism, feed pipeline, and material detection locations to collect multi-dimensional reactor process operating parameters that affect heat demand, reaction load, and material state during the reaction process in real time, recording these as the raw reactor process operating parameters. In this embodiment, the monitored process operating parameters include, but are not limited to, stirring speed, material feeding rate, pH value, and reactant concentration. In this embodiment, all sensors are synchronously sampled according to a unified time reference during data acquisition. That is, this embodiment synchronously collects parameters such as the temperature, mass flow rate and pressure of the input steam, the internal temperature of the reactor, and multi-dimensional process operating parameters. In specific applications, the implementer can adaptively set the acquisition frequency according to the sensor type, sensor performance parameters, etc. For example, the acquisition frequency of all data is set to once every 0.5 seconds. In this embodiment, the moment of data acquisition is recorded as the production moment.

[0012] In this embodiment, data preprocessing is performed after data acquisition, and the preprocessed data is recorded as preprocessed data. Specifically, the raw temperature, raw mass flow rate, and raw pressure of the input steam acquired by the sensor are preprocessed, and the preprocessed raw data of the input steam are recorded as the preprocessed temperature, preprocessed mass flow rate, and preprocessed pressure, respectively. The data preprocessing of the raw temperature, raw mass flow rate, and raw pressure of the input steam includes, but is not limited to, filtering and missing value detection and processing; however, the data preprocessing of the raw input steam data does not include data standardization or normalization. The raw internal temperature of the reactor acquired by the sensor is preprocessed, and the preprocessed raw internal temperature of the reactor is recorded as the preprocessed internal temperature of the reactor. The data preprocessing of the raw internal temperature of the reactor includes, but is not limited to, filtering, missing value detection and processing, and data standardization or normalization. Data preprocessing is performed on all dimensions of the raw reactor process operating parameters collected by sensors, and the preprocessed raw reactor process operating parameters are recorded as preprocessed reactor process operating parameters. The data preprocessing process includes, but is not limited to, filtering, missing value detection and handling, and data standardization or normalization. In addition, all sensor data needs to be timestamped to ensure consistency across the time dimension. Filtering refers to removing random noise interference during sensor acquisition using median filtering or moving average filtering methods; missing value detection and handling refers to supplementing missing data caused by communication fluctuations or equipment sampling anomalies using linear interpolation; data standardization or normalization refers to normalizing each data point collected by the sensors according to its maximum and minimum values ​​within its range, ensuring the normalized result falls within the 0 to 1 range, eliminating data scale differences caused by different dimensions of the monitoring parameters.

[0013] Therefore, this embodiment can obtain in real time the pretreatment temperature, pretreatment mass flow rate, and pretreatment pressure of the input steam in the phosphate production process, as well as the process operating parameters of the pretreatment reactor and the internal temperature of the pretreatment reactor through the above process.

[0014] In the phosphate production process, steam heating is typically used to heat the materials inside the reactor to maintain the temperature conditions required for the phosphate reaction. However, the reaction process inside the reactor has strong dynamic characteristics. At different production stages, factors such as the amount of material fed, the stirring state, the concentration of reactants, and the acid-base environment all affect the heat demand during the reaction. Therefore, under the same steam input conditions, different temperature responses may occur inside the reactor. At the same time, the transfer of steam from the input pipeline to the inside of the reactor requires a certain heat transfer time, resulting in a time lag between the steam regulation process and the temperature change in the reactor. Therefore, in actual production, relying solely on the internal temperature of the reactor for adjustment not only makes it difficult to promptly determine whether the current steam input matches the actual heat demand of the reaction stage, leading to abnormal situations such as insufficient steam supply causing reduced reaction efficiency or excessive steam input causing energy waste, but also lacks the ability to analyze the changing patterns of the heating state during steam regulation. Consequently, it becomes impossible to further distinguish whether the current anomaly is caused by insufficient heat supply or by excess heat overflow, and it cannot provide clear guidance on the direction of abnormal deviation for subsequent steam supply adjustments. This makes the detection results completely unsuitable for supporting the closed-loop optimization operation of the regulation system. To improve the accuracy and reliability of determining whether the current steam input matches the actual heat demand of the reaction stage, and to achieve reliable and accurate identification of anomaly types or directions of abnormal deviation, this embodiment subsequently analyzes the dynamic relationship between the steam input state and the reactor's thermal response state, establishing a heat demand reference model under different production conditions. This enables the detection of abnormalities in the steam regulation state and their deviation directions.

[0015] To achieve the above objectives, this embodiment, after data acquisition and preprocessing, first determines the specific enthalpy of the steam under the corresponding state based on the preprocessing pressure and temperature of the input steam, and then calculates the steam heat input to the reactor at each production moment in the phosphate production process, combined with the preprocessing mass flow rate of the input steam per unit time. Furthermore, for any phosphate production process, the expression for calculating the steam heat input to the reactor at any moment in that phosphate production process is as follows: ; Where Q0 is the heat of steam input to the reactor at that moment, F0 is the pre-treatment mass flow rate of the input steam at that moment, h0 is the steam enthalpy at that moment (i.e., the steam enthalpy at that production moment can be output by looking up a table or using the equation of state method given the pre-treatment pressure and temperature of the input steam at that moment), and hc is the reference enthalpy after steam condensation. The process of calculating the heat of steam input to the reactor given the temperature, mass flow rate, and pressure of the input steam is a known technique. Through the above process, this embodiment converts the collected steam temperature, pressure, and flow rate data into steam heat time-series data that can reflect the actual heating capacity. Similarly, to eliminate dimensional differences, it is also necessary to perform maximum and minimum value normalization according to the calculated steam heat value range, so that the normalization result is in the range of 0 to 1, and the normalized steam heat is recorded as the pre-treatment steam heat input to the reactor.

[0016] Because the steam heat enters the reactor through the pipeline and undergoes processes such as pipeline transmission, heat exchange, and material absorption, there is a time delay between the change in steam heat and the reactor temperature response. This results in a mismatch between the input heat data and the actual temperature change. In this embodiment, to correlate the input heat data with the actual temperature change, the delay time of heat transfer to the reactor must first be obtained through testing; that is, the steam heat transfer delay time must be determined based on prior testing. The process for determining the steam heat transfer delay time is as follows: Under stable production conditions, the steam input of the steam delivery pipeline is manually adjusted. Timing begins from the moment of adjustment and continues until the temperature sensor of the reactor first detects a recognizable temperature change response. The interval between these two time points is recorded, and this time interval is the single heat transfer delay time under a single test. To avoid result deviations caused by occasional minor fluctuations in operating conditions and sampling errors in a single test, the above steam input adjustment test is repeated multiple times to obtain multiple sets of heat transfer delay times under different steam input test scenarios. Finally, the arithmetic mean of the heat transfer delay times under all tests is taken, and this average is used as the final determined steam heat transfer delay time, which is the steam heat transfer delay time used in the subsequent detection method. In specific applications, the implementer can adaptively set the number of test groups according to the actual situation, but this embodiment requires no less than six groups, which can be set as an empirical value in this embodiment. Stable production conditions refer to a stage in the production process where no production faults, such as non-conforming products or machine malfunctions, are detected by manual or machine monitoring systems, or where all production parameters are within the theoretically stable range (e.g., the temperature inside the reactor is controlled between 30 and 50 degrees Celsius), and no abnormalities are detected in the production process.

[0017] After obtaining the steam heat transfer delay time from the test, the pre-treated steam heat is delayed or time-aligned based on this time to obtain the aligned pre-treated steam heat corresponding to each production moment. This ensures that the data at the same moment represents the reaction state under the same batch of input heat, thereby avoiding analytical bias caused by heat transfer lag and providing an accurate data foundation for subsequent thermal response model construction. The specific alignment process is as follows: For any production time t0 in any phosphate production process, the pretreatment steam heat input into the reactor at the time corresponding to the production time t0 being moved forward by D is taken as the alignment pretreatment steam heat corresponding to the production time t0 in that phosphate production process, where D is the steam heat transfer delay time; moving forward means moving in the past direction on the time axis; in addition, if the production time before the production time t0 is less than D, then the steam heat cannot be effectively aligned at that production time t0. Therefore, in this embodiment, the internal temperature of the reactor at that production time t0 is excluded, and no anomaly detection is performed on that production time t0. That is, this embodiment only analyzes the time when the alignment is effective.

[0018] Therefore, this embodiment can obtain the aligned pre-processed steam heat corresponding to the steam heat transfer delay time and the production time through the above process.

[0019] Step S002: Divide the internal temperature sequence of the pretreatment historical reactor into historical temperature segments according to the temperature change trend. Construct historical feature vectors corresponding to the historical temperature segments based on the starting temperature of each historical temperature segment and the corresponding average heat value of the aligned pretreatment steam. Cluster the historical temperature segments based on the distance between the historical feature vectors to obtain reaction condition pattern clusters.

[0020] To accurately adapt to the thermal demand characteristics of different reaction stages during real-time monitoring, this embodiment will next construct a reaction condition model cluster based on all data recorded during the completed phosphate production process. This reaction condition model cluster serves as a reference for matching the current production status and determining the direction of deviation from abnormal steam heating. The specific process for obtaining the reaction condition model cluster is as follows: Any completed phosphate production batch is selected as the reference process for phosphate production. The time sequence of all pre-treatment reactor internal temperatures obtained during this reference process is defined as the pre-treatment historical reactor internal temperature sequence. The reference process and the subsequent phosphate production process undergoing anomaly detection must have the same production process flow, production sequence, and equipment model. Then, the pre-treatment historical reactor internal temperature sequence is divided into historical temperature segments according to temperature change trends. The segmentation process involves performing first-order difference processing on the obtained pre-treatment historical reactor internal temperature sequence to generate a corresponding first-order difference sequence. Temperature data corresponding to consecutive first-order difference values ​​with the same sign in the first-order difference sequence are grouped into the same temperature segment. This completes the segmentation of the pre-treatment historical reactor internal temperature sequence, resulting in multiple historical temperature segments. In other words, the continuously changing full-volume reactor internal temperature time sequence data is divided into multiple independent historical temperature segments with clear unidirectional thermal change trends to characterize different reaction stages in the phosphate production process.

[0021] Since the conversion efficiency in the phosphate reaction process is affected by temperature conditions, the heat endothermic characteristics, reaction rate, and heat conversion efficiency of the materials may differ at different temperature stages. Furthermore, the impact of changes in parameters such as the material feeding rate, stirring speed, pH value, and reactant concentration on heat consumption may also vary. Therefore, to obtain the influence relationship of various operating parameters on heat changes under different temperature reaction conditions, and thus to determine whether the current steam input can meet future reaction heat requirements, this embodiment needs to quantify the heat input characteristics of different reaction stages to obtain a historical feature vector corresponding to each historical temperature segment. Specifically, this embodiment will construct a historical feature vector corresponding to each historical temperature segment based on the initial temperature of each historical temperature segment and the corresponding average aligned pretreatment steam heat. The first parameter in the historical feature vector corresponding to any historical temperature segment is the initial data in that historical temperature segment, and the second parameter is the average aligned pretreatment steam heat corresponding to all production moments within the time window corresponding to that historical temperature segment. The feature vector is located in a two-dimensional coordinate system where the horizontal axis represents the internal temperature of the pretreatment reactor and the vertical axis represents the pretreatment steam heat. Next, based on the Euclidean distance between historical feature vectors, the K-means++ clustering algorithm was used to cluster all historical temperature segments. Each cluster was recorded as a reaction condition pattern cluster to characterize the reaction condition patterns formed inside the reactor under different initial temperatures and different steam heat input conditions. The number of cluster centers during clustering was obtained using the elbow method.

[0022] Therefore, this embodiment can obtain the historical feature vectors corresponding to each historical temperature segment and each reaction condition mode cluster through the above process.

[0023] Step S003: Based on the current internal temperature of the pretreatment reactor and the corresponding aligned pretreatment steam heat, the current state feature vector is matched to obtain the current reference cluster; based on the initial and final temperatures of historical temperature segments within the current reference cluster and the aligned pretreatment steam heat, the steam input thermal response coefficient of each segment is calculated; based on the difference in steam input thermal response coefficients and clustering distance between pairs of historical temperature segments formed by combining historical temperature segments within the current reference cluster, the degree of difference in thermal response coefficients of each historical temperature segment pair is calculated; based on the correlation between the sequence of thermal response coefficient differences and the sequence of changes in each operating parameter dimension under the current reference cluster, the influence contribution weight value of each operating parameter dimension under the current reference cluster is obtained; based on the influence contribution weight value, the pretreatment reactor process parameters of the corresponding dimension at the current production time are weighted and fused to output the current heat demand index value at the current production time; based on the deviation of the current heat demand index value from the heat demand reference value and the difference in cluster profile coefficient before and after the current state feature vector is added to the current reference cluster, the current steam heating anomaly index value is calculated.

[0024] To determine whether the steam input conditions under the current production status match the actual heat demand of the reactor, it is necessary to first determine the historical reaction mode corresponding to the current reaction stage. Since the historical operating data has been divided into multiple reaction condition modes in the previous stage, the operating status at the current production moment can be matched in the historical reaction mode space to obtain the reference state closest to the current production process. The current production moment belongs to the phosphate production process that is currently under anomaly detection, and the phosphate production process that is currently under anomaly detection is recorded as the current phosphate production process.

[0025] Based on the above description, this embodiment first needs to obtain the current state feature vector at the current production moment. The first parameter in the current state feature vector is the current internal temperature of the pretreatment reactor and the aligned pretreatment steam heat corresponding to the current production moment. The current internal temperature of the pretreatment reactor is the internal temperature of the pretreatment reactor at the current production moment in the current phosphate production process. Then, the current state feature vector is matched with the reaction condition pattern cluster to obtain the current reference cluster, which is also the reference cluster corresponding to the current reaction stage. The specific process of obtaining the current reference cluster is as follows: calculate the Euclidean distance between the current state feature vector and the vector corresponding to the cluster center of each reaction condition pattern cluster. Since the smaller the distance, the closer the temperature state and steam input conditions of the current reaction stage are to the reaction condition pattern cluster, this embodiment selects the reaction condition pattern cluster corresponding to the smallest Euclidean distance value as the current reference cluster.

[0026] Since the historical reaction processes within the current reference cluster share similar temperature conditions and steam input heat levels with the current state, the current reference cluster can reflect the heat conversion patterns under normal conditions during the current production stage. In other words, the current reference cluster can serve as a benchmark for the heat conversion patterns under normal operating conditions during the current production stage. Furthermore, when the distribution of operating parameters directly affecting heat changes—such as the material feeding rate, stirring speed, pH value, and reactant concentration—at the current production moment is consistent with the corresponding parameters in the current reference cluster, it indicates that the current steam input conditions and the heat demand under the current reaction state are relatively matched. The current input heat can meet the heat consumption requirements of the subsequent reaction process. However, if the operating parameters directly affecting heat changes at the current production moment deviate significantly from the corresponding parameters in the current reference cluster, it indicates that the heat consumption trend under the current reaction state has changed. In this case, the original steam input may not be able to meet future heat demands, or there may be steam input redundancy, leading to decreased reaction efficiency or energy waste. Therefore, to obtain the matching relationship between changes in operating parameters across various dimensions and changes in steam thermal response during the current reaction stage, this embodiment needs to calculate the current steam heating anomaly index value based on the matched current reference cluster.

[0027] However, due to differences in reaction rates, material endothermic capacity, and heat transfer efficiency at different temperature stages during phosphate production, even the same operating parameter can have vastly different impacts on reactor heat changes at different reaction stages. Therefore, a uniform, fixed weight value cannot be used to calculate heat demand under all operating conditions. Consequently, a separate parameter influence contribution weight model must be built for the currently matched reference cluster to accurately determine the actual impact of each operating parameter on heat demand changes under the current specific reaction mode. This provides reliable support for accurately calculating real-time heat demand and determining whether steam supply is abnormal. Therefore, this embodiment next needs to specifically deconstruct the currently matched reference cluster and analyze the contribution of each operating parameter to changes in steam thermal response to identify the corresponding matching relationship between parameter changes and thermal response. That is, based on the parameters corresponding to the current reference cluster, the influence contribution weight value of each operating parameter dimension under the current reference cluster will be evaluated. The specific process for obtaining the influence contribution weight value of each operating parameter dimension under the current reference cluster is as follows:

[0028] First, the steam input thermal response coefficient of each historical temperature segment within the current reference cluster is calculated based on the initial and final temperatures of the historical temperature segments and the aligned pre-processed steam heat. To analyze the changes in thermal response capability between different temperature stages within the current reference cluster, the difference in steam input thermal response coefficients and cluster distance between pairs of historical temperature segments formed by pairwise non-repeating combinations within the current reference cluster are used to calculate the degree of difference in thermal response coefficients between each pair of historical temperature segments. Then, a sequence of thermal response coefficient differences under the current reference cluster is formed based on all the degrees of difference in thermal response coefficients. This sequence is derived from the thermal response coefficient differences within the current reference cluster. The thermal response coefficient differences of all historical temperature segments are constructed by pairwise non-repeating combinations of all historical temperature segments. Based on the pretreatment reactor process operating parameters of each dimension corresponding to each historical temperature segment in the current reference cluster, a sequence of changes in each operating parameter dimension under the current reference cluster is constructed. Based on the correlation between the thermal response coefficient difference sequence under the current reference cluster and the change sequence of each operating parameter dimension under the current reference cluster, the influence contribution weight value of each operating parameter dimension under the current reference cluster is obtained. It should be noted that the data at the same position in the change sequence and the thermal response coefficient difference sequence are the data corresponding to the same historical temperature segment pair.

[0029] The specific process for obtaining the steam input thermal response coefficient of the a-th historical temperature segment within the current reference cluster is as follows: The temperature rise corresponding to the a-th historical temperature segment is defined as the difference between the ending temperature and the starting temperature. The cumulative value of the aligned steam heat at all production moments within the time window corresponding to the a-th historical temperature segment is defined as the cumulative steam input heat for the a-th historical temperature segment. The ratio of the temperature rise to the cumulative steam input heat is defined as the steam input thermal response coefficient for the a-th historical temperature segment. The expression for the steam input thermal response coefficient of the a-th historical temperature segment is: ; in, Let be the steam input thermal response coefficient for the a-th historical temperature segment. This refers to the last temperature in the a-th historical temperature segment. Let be the starting temperature in the a-th historical temperature segment. The cumulative heat input for steam corresponding to the a-th historical temperature segment. Used to characterize the temperature response capability produced by a unit steam heat input under the current reaction mode. The higher the value, the stronger the ability of the steam heat to be converted into a temperature rise in the reactor at the current stage. The smaller the value, the greater the heat consumption or heat transfer loss in the current stage.

[0030] The specific process for obtaining the degree of difference in the thermal response coefficient of the r-th historical temperature segment pair in the set formed by all historical temperature segment pairs is as follows: The absolute value of the difference between the steam input thermal response coefficients of the two historical temperature segments in the r-th historical temperature segment pair is denoted as the absolute difference of the thermal response coefficients corresponding to the r-th historical temperature segment pair. The proportional normalized result of the Euclidean distance between the historical eigenvectors corresponding to the two historical temperature segments in the r-th historical temperature segment pair is denoted as the thermal response confidence level of the r-th historical temperature segment pair. The product of the absolute difference of the thermal response coefficients corresponding to the r-th historical temperature segment pair and its corresponding thermal response confidence level is denoted as the degree of difference of the thermal response coefficients of the r-th historical temperature segment pair. The expression for the degree of difference of the thermal response coefficients of the r-th historical temperature segment pair is: ; in, Let R represent the degree of difference in the thermal response coefficients of the r-th historical temperature segment pair, R be the total number of historical temperature segment pairs obtained by pairwise non-repeating combinations of historical temperature segments within the current reference cluster, and e be the natural constant. Let be the absolute value of the difference between the steam input thermal response coefficients of the two historical temperature segments in the r-th historical temperature segment pair. Let be the Euclidean distance between the historical feature vectors corresponding to the two historical temperature segments in the r-th historical temperature segment pair. The smaller, The larger the value, the closer the internal temperature and steam input of the reactor in the two historical temperature segments of the r-th historical temperature segment pair are, and the greater the difference in the calculated thermal response coefficients between them. The more credible it is, the higher its reference value; The larger and When it is larger, The larger, and The larger the value, the greater the difference in thermal response capability between the two historical temperature segments in the r-th historical temperature segment pair. This indicates that under the current reaction mode, the temperature response capability produced by a unit steam heat input in these two stages differs significantly. The smaller the value, the smaller the difference in the steam heat input response coefficient between the two historical temperature segments in the r-th historical temperature segment pair.

[0031] The specific process for obtaining the sequence of changes in the i-th operating parameter dimension under the current reference cluster is as follows: For any pair of historical temperature segments belonging to the current reference cluster, within the time windows corresponding to the two historical temperature segments in the pair, extract all pretreatment reactor process operating parameters belonging to the i-th operating parameter dimension, calculate the arithmetic mean of all pretreatment reactor process operating parameters belonging to the i-th operating parameter dimension extracted within their respective time windows, take the absolute value of the difference between the two means, and record it as the degree of change of the historical temperature segment pair under the i-th operating parameter dimension. The sequence of the degree of change of all historical temperature segment pairs belonging to the current reference cluster under the i-th operating parameter dimension is recorded as the sequence of changes in the i-th operating parameter dimension at the current production time.

[0032] Based on the correlation between the thermal response coefficient difference sequence and the change sequence of each operating parameter dimension under the current reference cluster, the specific process for calculating the influence contribution weight value of each operating parameter dimension under the current reference cluster is as follows: For the i-th operating parameter dimension, the normalized result of the absolute value of the Spearman rank correlation coefficient between the sequence of changes in the i-th operating parameter dimension under the current reference cluster and the sequence of differences in the thermal response coefficients under the current reference cluster is denoted as the influence contribution weight value of the i-th operating parameter dimension under the current reference cluster; the expression for the influence contribution weight value of the i-th operating parameter dimension under the current reference cluster is: ; in, The weight value is assigned to the influence of the i-th operating parameter dimension under the current reference cluster, where X0 is the sequence of differences in thermal response coefficients under the current reference cluster. Z represents the sequence of changes in the i-th operational parameter dimension under the current reference cluster, where Z is the total number of operational parameter dimensions monitored in this embodiment. For X0 and Spearman rank correlation coefficient between them Used to represent the contribution weight of the i-th parameter to the change in thermal response capability under the reaction mode of the reference cluster; The function is to normalize the molecules; since the more easily a change in a certain dimension of operating parameters causes a change in thermal response capability, the more consistent the trend of the change sequence of that operating parameter dimension with the change sequence of thermal response coefficient difference, that is, the stronger the correlation between the two, and the higher the contribution of that dimension to the change in heat demand under the current reaction mode, therefore... The larger the value, the stronger the synchronicity between the change trend of the i-th operating parameter dimension and the change trend of the thermal response coefficient difference sequence under the current reference cluster. The higher the correlation between the two, the stronger the driving effect of parameter fluctuations in the i-th operating parameter dimension on the change of reactor thermal response capability. A smaller value indicates that the fluctuation of the i-th operating parameter dimension has almost no significant driving effect on the change in the reactor's thermal response capability; while The larger The larger, therefore The larger the value, the more dominant the i-th operating parameter dimension is in causing changes in the thermal response state of the reactor corresponding to the current reference cluster. Its contribution to heat demand is higher, making it the main variable affecting heat demand fluctuations under the current reaction mode. Therefore, the i-th operating parameter dimension has a higher weight in subsequent calculations of the current heat demand index and the heat demand reference value. Conversely, a smaller value indicates a lower weighting. The smaller the value, the weaker the influence of the i-th operating parameter dimension on the change of the reactor thermal response state corresponding to the current reference cluster, the lower its contribution to the heat demand, and the more it is a secondary influencing parameter under the current reaction mode. In the subsequent calculation of the current heat demand index value and the heat demand reference value, the weight of the i-th operating parameter dimension is even lower, and the degree of influence on the final heat demand result is even smaller.

[0033] This embodiment obtains the influence contribution weight values ​​of each operating parameter dimension under the current reference cluster through the above process; after obtaining the influence contribution weight values ​​of each operating parameter dimension under the current reference cluster, the current heat demand index value at the current production time and the heat demand reference value corresponding to each historical temperature segment in the current reference cluster are calculated based on the influence contribution weight values ​​of each operating parameter dimension under the current reference cluster.

[0034] The specific process for obtaining the current heat demand index value at the current production time by weighting and fusing the pretreatment reactor process parameters of the corresponding dimension based on the influence contribution weight value is as follows: The pretreatment reactor process operating parameters of each operating parameter dimension at the current production time are multiplied by the influence contribution weight value of the corresponding operating parameter dimension under the current reference cluster, and then directly summed. This summation is the current heat demand index value at the current production time. The expression for the current heat demand index value at the current production time is: ,in, Contribute a weight value to the influence of the i-th running parameter dimension under the current reference cluster. The pretreatment reactor process operating parameters are the i-th operating parameter dimension at the current production time, where t represents the current production time. The current heat demand index value represents the heat demand level estimated at the current production time based on the current reaction state, material state, and reaction stage characteristics. The current heat demand index value can reflect the heat demand required for the reactor to maintain a normal reaction process under continuously changing current production conditions.

[0035] Based on the influence contribution weight value, the pretreatment reactor process parameters of the corresponding dimension at the production time within the historical temperature segment in the current reference cluster are weighted and fused to obtain the heat demand reference value corresponding to each historical temperature segment in the current reference cluster. The process of obtaining the heat demand reference value corresponding to any historical temperature segment in the current reference cluster is as follows: the average value of the heat demand index values ​​at all historical production times within the time window corresponding to the historical temperature segment is the heat demand reference value corresponding to the historical temperature segment. The heat demand index value at any historical production time within the time window corresponding to the historical temperature segment is the pretreatment reactor process operating parameter of each operating parameter dimension at that production time, which is the sum of the results after multiplying by the influence contribution weight value of the corresponding operating parameter dimension in the current reference cluster.

[0036] After obtaining the current heat demand index value at the current production moment and the heat demand reference value corresponding to each historical temperature segment within the current reference cluster, the current steam heating anomaly index value corresponding to the current production moment is calculated based on the deviation of the current heat demand index value from the heat demand reference value and the difference in cluster profile coefficient before and after adding the current state feature vector to the current reference cluster. The cluster profile coefficient calculation method is a publicly available and generally used algorithm in the field of cluster analysis, and belongs to the prior art known in this field. The specific process for obtaining the current steam heating anomaly index value corresponding to the current production moment is as follows:

[0037] The ratio of the average difference between the current heat demand index value and all heat demand reference values ​​to the heat demand fluctuation reference value is denoted as the current heat demand relative deviation. The heat demand fluctuation reference value is obtained as follows: For all historical temperature segments within the current reference cluster, the heat demand reference values ​​are paired without repetition to output heat demand reference value pairs. The absolute value of the difference between the two heat demand reference values ​​in each pair is calculated and denoted as the difference value of the corresponding heat demand reference value pair. The average of the differences of all heat demand reference value pairs is the heat demand fluctuation reference value. The negative exponential mapping result of the absolute value of the difference between the cluster profile coefficient calculated before adding the current state feature vector to the current reference cluster and the cluster profile coefficient calculated after adding the current state feature vector to the current reference cluster is denoted as the current matching confidence. The product of the current heat demand relative deviation and the current matching confidence is denoted as the current steam heating anomaly index value corresponding to the current production time. The expression for the current steam heating anomaly index value corresponding to the current production time is: ; Where DW represents the current steam heating anomaly index value corresponding to the current production moment, and G represents the number of historical temperature segments within the current reference cluster. This represents the current heat demand index value. W0 is the heat demand reference value corresponding to the g-th historical temperature segment within the current reference cluster, and W0 is the heat demand fluctuation reference value. The absolute value of the difference between the cluster profile coefficient calculated before adding the current reference cluster to the current state feature vector and the cluster profile coefficient calculated after adding the current state feature vector to the current reference cluster, where e is a natural constant. This characterizes the average deviation between the current heat demand and the historical normal operating baseline. The further away from 0, the higher the overall difference between the current heat demand under the reaction state and the historical normal reference value, and the more obvious the deviation of the current heat demand from the normal level. If the value is positive and large, it means that the current heat demand is much higher than the historical normal level, and there is a high probability of insufficient steam supply. If the value is negative and the absolute value is large, it means that the current heat demand is much lower than the historical normal level, and the possibility of redundant steam input is higher. The closer W0 is to 0, the better the current heat demand matches the historical normal operating heat demand distribution within the reference cluster. This indicates a better match between current steam heating and reaction heat demand, and the closer the operating condition is to normal operation. W0's function is to... Normalization is performed to make the deviation results under different working conditions comparable; The larger the absolute value, the higher the deviation of the current heat demand from the normal benchmark, and the more severe the mismatch between heat supply and demand. The smaller the absolute value, the lower the deviation of the current heat demand from the normal benchmark, and the better the matching status of heat supply and demand. This represents the change in the silhouette coefficients of the reference cluster before and after the current state feature vector is added to the current reference cluster. It is used to characterize the reliability of the matching calculation results between the current state and historical response patterns. When the value is small, it indicates that the addition of the current state feature vector has not significantly changed the original clustering structure, meaning that the current state feature vector has a high degree of matching with the corresponding response pattern of the reference cluster. In this case, the reliability of the anomaly index calculation results is high. When the value is large, it indicates that the addition of the current state feature vector has caused a significant change in the reference cluster structure, suggesting that the current state may be located in the boundary region between different reaction modes. In this case, by... Reducing the credibility of abnormal indicator results lowers the probability of subsequent false alarms. The larger the absolute value of DW, the higher the mismatch between the current steam heating supply and reaction heat demand, and the higher the confidence level of the judgment result, the higher the probability of triggering an abnormal steam heating warning. A positive number indicates that there may be insufficient heating, while a negative number indicates that there may be excessive heating. The smaller the absolute value of DW, the better the matching status between the current steam heating supply and reaction heat demand, and the more reliable the matching between the operating conditions and historical patterns, the greater the probability that the steam regulation system is in a normal and stable operating state.

[0038] Step S004: Detect anomalies and anomaly types in the current steam regulation based on the current steam heating anomaly index values.

[0039] In this embodiment, after obtaining the current steam heating anomaly index value, the current steam regulation is detected and identified based on the magnitude and sign of the current steam heating anomaly index value; specifically: When DW equals 0, it indicates that the calculated heat demand level at the current production moment is consistent with the average heat demand under the reference reaction mode. In other words, the current heat demand index value is consistent with the reference heat demand value, meaning that the current steam input heat is basically matched with the heat demand under the current reaction state, and the steam regulation process is in normal condition. When DW is greater than 0, it indicates that the heat demand under the current reaction state is higher than the corresponding level under the reference reaction mode. This means that changes in the current material state, reaction stage, or operating parameters have led to an increase in heat demand, and there is a possibility that the current steam input heat cannot meet the heat demand of the subsequent reaction process in the reactor. This indicates a high probability that the current steam supply capacity is insufficient, which may lead to a decrease in the reactor's heating rate, a prolonged reaction time, or a reduction in phosphate conversion efficiency, requiring an increase in steam input. When DW is less than 0, it indicates that the heat demand under the current reaction state is lower than the corresponding level under the reference reaction mode. This means that the current input steam heat exceeds the actual heat required for the reaction process. This indicates a high probability of steam input redundancy, which may lead to energy waste and the reaction temperature exceeding the reasonable range, requiring a reduction in steam supply.

[0040] In this embodiment, to avoid false alarms caused by normal production fluctuations, the existence of an anomaly is not directly determined based on whether DW is greater than or less than 0, nor is the anomaly type or deviation direction identified. Instead, this embodiment subsequently uses a preset anomaly determination threshold V greater than 0 to determine the anomaly and identify the anomaly type or deviation direction. Specifically: when If this occurs, it is determined that there is a significant heating anomaly in the current steam input state, and when... When this happens, a steam shortage warning signal is output, and the steam regulation target value is increased. When an abnormality occurs, an excessive steam warning signal is output to reduce the target steam input value. This detection process not only detects whether there are any abnormalities in the current steam regulation system, but also determines whether the current heating is insufficient or excessive based on the direction of the abnormal deviation. This provides targeted feedback to the automatic steam regulation system. Specifically, if the current abnormal steam supply index value is greater than or equal to the preset abnormality judgment threshold, the steam supply is determined to be insufficient; if the current abnormal steam supply index value is less than or equal to the negative value of the preset abnormality judgment threshold, the steam supply is determined to be redundant; and if the absolute value of the current abnormal steam supply index value is less than the preset abnormality judgment threshold, the steam supply is determined to be in a normal state. In practical applications, the implementer needs to set the value of V according to the actual situation, requiring V to be between 1 and 2, such as 1.2.

[0041] Therefore, this embodiment completes the detection of abnormal operation of the industrial steam automatic regulation system; and the steam regulation abnormality and abnormality type detection strategy provided in this embodiment can accurately identify two types of abnormalities: insufficient steam supply or redundancy, effectively avoid lag misjudgment, adapt to the heat demand patterns of different process stages, ensure stable production process compliance, reduce unnecessary heat energy loss, and simultaneously improve the reliability of steam regulation and energy utilization efficiency.

[0042] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for detecting operational anomalies in an industrial steam automatic regulation system, characterized in that, The method includes the following steps: In the phosphate production process, the process operating parameters of the pretreatment reactor, the internal temperature of the pretreatment reactor, and the heat of the pretreatment steam input into the reactor are acquired in real time. Based on the pre-tested steam heat transfer delay time, the heat of the pretreatment steam is delayed and aligned to obtain the aligned pretreatment steam heat corresponding to each production moment. The internal temperature sequence of the pretreatment historical reactor is divided into historical temperature segments according to the temperature change trend. Based on the starting temperature of each historical temperature segment and the corresponding average heat value of the aligned pretreatment steam, a historical feature vector corresponding to the historical temperature segment is constructed. Based on the distance between the historical feature vectors, the historical temperature segments are clustered to obtain reaction condition pattern clusters. The current reference cluster is obtained by matching the current state feature vector constructed based on the current internal temperature of the pretreatment reactor and the corresponding aligned pretreatment steam heat. Based on the initial and final temperatures of historical temperature segments within the current reference cluster and the aligned pre-processed steam heat, the steam input thermal response coefficient of each segment is calculated. Based on the difference in steam input thermal response coefficients and clustering distance between pairs of historical temperature segments formed by combining historical temperature segments within the current reference cluster, the degree of difference in thermal response coefficients between each pair of historical temperature segments is calculated. Based on the correlation between the sequence of thermal response coefficient differences and the sequence of changes in each operating parameter dimension under the current reference cluster, the influence contribution weight value of each operating parameter dimension under the current reference cluster is obtained. Based on the influence contribution weight value, the pre-processed reactor process parameters of the corresponding dimension at the current production moment are weighted and fused to output the current heat demand index value at the current production moment. Based on the deviation of the current heat demand index value from the heat demand reference value and the difference in cluster profile coefficient before and after adding the current state feature vector to the current reference cluster, the current steam heating anomaly index value is calculated. Based on the current abnormal steam heating index values, detect the abnormality and abnormality type of the current steam regulation.

2. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 1, characterized in that, Methods for obtaining historical temperature fragments include: Temperature data corresponding to consecutive difference values ​​with the same sign in the first-order difference sequence of the pre-processed historical reactor internal temperature sequence are grouped into the same historical temperature segment.

3. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 1, characterized in that, The methods for obtaining the current reference cluster include: Calculate the Euclidean distance between the current state feature vector and the cluster center vector of each reaction condition pattern cluster, and select the reaction condition pattern cluster corresponding to the smallest Euclidean distance value as the current reference cluster.

4. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 1, characterized in that, The method for obtaining the steam input thermal response coefficient for any historical temperature segment within the current reference cluster includes: The temperature rise is taken as the difference between the end temperature and the beginning temperature in the historical temperature segment, and the cumulative value of all aligned pre-processed steam heat corresponding to the historical temperature segment is taken as the cumulative steam input heat. The ratio of the temperature rise to the cumulative steam input heat is taken as the steam input heat response coefficient of the historical temperature segment.

5. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 1, characterized in that, Methods for obtaining the degree of difference in thermal response coefficients for any pair of historical temperature segments include: The absolute difference of the steam input thermal response coefficients of the two historical temperature segments in the historical temperature segment pair is denoted as the absolute difference of thermal response coefficients. The proportional normalization result of the Euclidean distance between the historical feature vectors corresponding to the two historical temperature segments in the historical temperature segment pair is denoted as the thermal response confidence of the historical temperature segment pair. The product of the absolute difference of thermal response coefficients and the thermal response confidence is denoted as the degree of difference of thermal response coefficients of the historical temperature segment pair.

6. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 1, characterized in that, The sequence of changes in any operating parameter dimension under the current reference cluster is composed of the changes in all historical temperature segments under the operating parameter dimension; the change in any historical temperature segment under any operating parameter dimension refers to the absolute difference between the mean values ​​of all pretreatment reactor process operating parameters belonging to the operating parameter dimension obtained within the time window corresponding to each of the two historical temperature segments in the historical temperature segment pair.

7. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 6, characterized in that, The influence contribution weight of any operating parameter dimension under the current reference cluster is the normalized result of the absolute value of the Spearman rank correlation coefficient between the sequence of changes in the operating parameter dimension and the sequence of differences in the thermal response coefficient.

8. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 1, characterized in that, Methods for obtaining current heat demand reference values ​​include: For any historical temperature segment within the current reference cluster, the average value of the heat demand index at all historical production moments within the time window corresponding to the historical temperature segment is recorded as the heat demand reference value corresponding to the historical temperature segment.

9. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 8, characterized in that, The current methods for obtaining abnormal steam heating index values ​​include: The ratio of the average difference between the current heat demand index value and all heat demand reference values ​​to the heat demand fluctuation reference value is denoted as the current heat demand relative deviation. The heat demand fluctuation reference value is the average absolute difference between any two heat demand reference values ​​in all heat demand reference value pairs obtained by pairwise non-repeating combinations of the heat demand reference values ​​corresponding to all historical temperature segments within the current reference cluster. The negative exponential mapping result of the absolute difference between the cluster profile coefficient before and after the current state feature vector is added to the current reference cluster is denoted as the current matching confidence. The product of the current heat demand relative deviation and the current matching confidence is denoted as the current steam heating anomaly index value.

10. The method for detecting operational anomalies in an industrial steam automatic regulation system as described in claim 1, characterized in that, Methods for detecting anomalies in current steam regulation based on current steam heating anomaly index values ​​include: If the current abnormal steam supply index value is greater than or equal to the preset abnormality judgment threshold, the steam supply is determined to be insufficient; if the current abnormal steam supply index value is less than or equal to the negative value of the preset abnormality judgment threshold, the steam supply is determined to be redundant; if the absolute value of the current abnormal steam supply index value is less than the preset abnormality judgment threshold, the steam supply is determined to be in a normal state.

Images