Intelligent smelting plant information management method and system

Abstract

The present invention discloses an intelligent smelting plant information management method and system, which relates to the field of data prediction technology, including analyzing the correlation relationship between information, providing a basis for subsequent multi-feature combination, and ensuring the reliability of multi-factor association calculation. According to the correlation relationship between information, multiple features are combined, and a smelting energy consumption analysis model is trained through feature combination items and trend components. Multiple features are combined for correlation, the correlation between different features is considered, and trend components are added to make the smelting energy consumption analysis model prediction more accurate. The abnormal energy consumption segment is analyzed to trace the cause of the abnormal energy consumption. The deviation of energy consumption is determined through the predicted normal range of energy consumption, thereby determining the cause of the abnormal energy consumption, and subsequent monitoring of these abnormal energy consumption causes and optimization of the smelting process are carried out to ensure the reliability and effectiveness of the information management of the smelting plant.

Description

Technical Field

[0001] The present invention relates to the technical field of data prediction, and in particular to an information management method and system for an intelligent smelting plant. Background Art

[0002] The analysis of abnormal energy consumption in smelting plants mainly relies on advanced technologies such as the Internet of Things, big data, and cloud computing. Traditional smelting plants often have problems such as incomplete data collection, delayed analysis, and insufficient decision-making basis in energy consumption management, resulting in serious energy waste. Intelligent smelting plants collect real-time consumption data of various energy sources such as water, electricity, and gas by installing intelligent sensors, and use the Internet of Things technology to achieve real-time data transmission. The big data platform stores, processes, and analyzes massive energy consumption data, uses machine learning algorithms to identify abnormal energy consumption patterns, predict energy consumption trends, and provide a scientific basis for the formulation of energy-saving strategies. Cloud computing technology provides powerful computing power to support complex data analysis and intelligent decision-making. Through the comprehensive application of these technologies, intelligent smelting plants can achieve real-time monitoring, data analysis, and intelligent management of energy usage, effectively reduce energy consumption, improve energy utilization efficiency, and promote green and sustainable development.

[0003] In the prior art, because the energy consumption in the smelting process is affected by the fluctuations of various factors, the prediction of the normal range of energy consumption is inaccurate and unreliable, which makes it impossible to accurately identify energy consumption anomalies and form effective monitoring and management of the informationization of the smelting plant.

[0004] Therefore, how to improve the accuracy and reliability of predictions within the normal range of energy consumption is a technical problem that needs to be solved. Summary of the invention

[0005] The purpose of the present invention is to solve the problem of poor prediction accuracy and low reliability of normal range of energy consumption in the prior art, and to propose an intelligent smelting plant information management method, which includes:

[0006] Collect the reflected information in the smelting process and analyze the correlation between the reflected information;

[0007] Construct time series data reflecting information, determine the trend components reflecting information, extract the features reflecting information, and combine multiple features according to the correlation between the reflected information to obtain feature combination items;

[0008] Through the training of the smelting energy consumption analysis model by characteristic combination items and trend components, the normal range of future energy consumption is predicted, the actual energy consumption is compared with the normal range of future energy consumption to determine the abnormal energy consumption segment, and the abnormal energy consumption segment is analyzed to trace the cause of abnormal energy consumption;

[0009] Optimize the smelting process by tracking the causes of abnormal energy consumption.

[0010] In some embodiments of the present application, the analysis reflects the correlation between information, including:

[0011] The reflected information includes process parameters, equipment status parameters and energy consumption parameters;

[0012] Calculate the mutual information between process parameters, equipment status parameters and energy consumption parameters, sort the mutual information from large to small, and determine the maximum value, median value and minimum value. Divide the mutual information into several intervals, and determine the frequent intervals in the interval by the number of times the mutual information appears in the interval.

[0013] The mutual information threshold is set according to the positional relationship between the frequent interval and the maximum value, median value and minimum value, and the mutual information is filtered, so that the correlation relationship reflecting the information corresponding to the mutual information is retained.

[0014] In some embodiments of the present application, the mutual information threshold is set according to the positional relationship between the frequent interval and the maximum value, the median value, and the minimum value, including:

[0015] When the frequent interval is between the maximum value and the median value, the median value is adjusted according to the distance between the median value and the frequent interval, and the adjusted median value is used as the mutual information threshold;

[0016] When the median value is within the frequent interval, the lower endpoint of the frequent interval is adjusted according to the distance between the median value and the frequent interval, and the adjusted lower endpoint is used as the mutual information threshold;

[0017] When the frequent interval is between the median and the minimum, the minimum is adjusted according to the distance between the minimum and the frequent interval, and the adjusted minimum is used as the mutual information threshold;

[0018] Among them, the distances between the maximum value, median value, and minimum value and the frequent interval are the differences between the value and the closer endpoint in the frequent interval.

[0019] In some embodiments of the present application, time series data reflecting information is constructed to determine trend components reflecting information, including:

[0020] The data of process parameters, equipment status parameters and energy consumption parameters are time stamped and aligned to obtain time series data, and the time series data is smoothed;

[0021] Decompose the time series data through the time series decomposition algorithm to determine the trend components of process parameters, equipment status parameters and energy consumption parameters;

[0022] Among them, time series decomposition algorithms include STL decomposition and EMD decomposition.

[0023] In some embodiments of the present application, multiple features are combined according to the correlation between the reflected information to obtain feature combination items, including:

[0024] The features of process parameters, equipment status parameters and energy consumption parameters are extracted to obtain process characteristics, equipment status characteristics and energy consumption characteristics. The corresponding features of process characteristics, equipment status characteristics and energy consumption characteristics are cross-featured by reflecting the correlation between the information to form feature combination items. The feature combination items are marked with feature categories and mutual information under the correlation relationship.

[0025] In some embodiments of the present application, the smelting energy consumption analysis model is trained by characteristic combination items and trend components, including:

[0026] Compare the feature categories under all feature combination items, record the feature combination items with overlapping process features or equipment status features between different feature combination items as overlapping feature combination items, and record the overlapping process features or equipment status features as overlapping features;

[0027] Determine the model structure level based on feature combination items, overlapping feature combination items and mutual information;

[0028] ;

[0029] in, is the model structure level, , are the number of feature combination items and the combination weight corresponding to the mutual information, , are the conversion coefficients of feature combination items and overlapping feature combination items respectively, , are the number of feature combinations and overlapping feature combinations respectively, For the The mutual information value of feature combinations, is the number of overlapping features, For the The sum of the mutual information of multiple overlapping feature combinations corresponding to overlapping features, , All are preset constants;

[0030] Different model structure levels correspond to multiple combinations of different model structure parameters. The trend components of the features in each feature combination item are added to the corresponding feature combination item. All feature combination items are divided into training sets and test sets. The smelting energy consumption analysis model is trained using the training sets and test sets.

[0031] In some embodiments of the present application, in the process of training the smelting energy consumption analysis model through the training set and the test set, the method further includes:

[0032] The trend component of each feature in the feature combination item is statistically analyzed to determine the overall change rate of the feature combination item. All feature combination items are grouped according to the overall change rate of the feature combination items. The learning rate is set for the overall change rate of the feature combination items in each group. The smelting energy consumption analysis model is trained by combining the training set and the test set.

[0033] In some embodiments of the present application, the actual energy consumption is compared with the normal range of future energy consumption to determine the abnormal energy consumption segment, and the abnormal energy consumption segment is analyzed to trace the cause of the abnormal energy consumption, including:

[0034] Construct a curve of actual energy consumption changing over time, recorded as the energy consumption curve, mark the normal range of future energy consumption on the energy consumption curve, and the part of the energy consumption curve where the actual energy consumption exceeds the normal range of future energy consumption is regarded as the abnormal energy consumption segment, and determine the timestamp of the abnormal energy consumption segment, recorded as the abnormal energy consumption time;

[0035] Calculate the curve area between the energy consumption abnormality period and the future normal energy consumption range, collect the reflection information of the period before and after the energy consumption abnormality time and during the energy consumption abnormality time, and construct the reflection information set to be tested;

[0036] Calculate the state transition probability of each abnormal cause in the measured reflection information set and the energy consumption abnormal cause library, and determine the comprehensive transition probability by combining the curve area;

[0037] ;

[0038] in, The first The combined transition probability of the abnormal causes is The first The state transition probability of each abnormal cause, The curve area and energy consumption abnormality cause library The deviation of energy consumption abnormality due to abnormal reasons, is the compensation probability, Indicates the curve area and the energy consumption abnormality cause library The compensation probability obtained by mapping the similarity of the deviation of the energy consumption abnormality of each abnormal reason;

[0039] Each abnormal cause in the energy consumption abnormal cause library is screened according to the comprehensive transfer probability, so as to trace the abnormal cause of energy consumption.

[0040] Correspondingly, the present application also provides an intelligent smelting plant information management system, including:

[0041] The analysis module is used to collect the reflected information in the smelting process and analyze the correlation between the reflected information;

[0042] The feature module is used to construct time series data reflecting information, determine the trend components reflecting information, extract the features reflecting information, and combine multiple features according to the correlation relationship between the reflected information to obtain feature combination items;

[0043] The prediction module is used to train the smelting energy consumption analysis model through feature combination items and trend components, predict the normal range of future energy consumption, compare the actual energy consumption with the normal range of future energy consumption to determine the abnormal energy consumption segment, and analyze the abnormal energy consumption segment to trace the cause of abnormal energy consumption;

[0044] Management module, used to optimize the smelting process by tracking the causes of abnormal energy consumption.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. Analyze the correlation between information, provide a basis for subsequent multi-feature combination, and ensure the reliability of multi-factor association calculation. Combine multiple features according to the correlation between information, train the smelting energy consumption analysis model through feature combination items and trend components, combine multiple features for correlation, consider the correlation between different features, and add trend components to make the smelting energy consumption analysis model prediction more accurate.

[0047] 2. Analyze the abnormal energy consumption segments to trace the causes of abnormal energy consumption. Determine the deviation of energy consumption through the predicted normal range of energy consumption, thereby determining the causes of abnormal energy consumption. Follow-up monitoring of these abnormal energy consumption causes and optimization of the smelting process ensure the reliability and effectiveness of the information management of the smelting plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of the process of the intelligent smelting plant information management method proposed by the present invention;

[0049] Figure 2 This is a schematic diagram of the structure of the intelligent smelting plant information management system proposed in the present invention. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0051] Reference Figure 1 , the information management method of the intelligent smelting plant includes the following steps,

[0052] Step S101, collecting the reflected information in the smelting process and analyzing the correlation between the reflected information.

[0053] In this embodiment, an automated data acquisition system, such as a SCADA (supervisory control and data acquisition) system, is used to collect these data in real time or regularly and store them in a database. The reflected information includes process parameters, equipment status parameters and energy consumption parameters. Process parameters refer to parameters that require special attention and control during the smelting process. They directly affect the smooth progress of the smelting process, the quality of the product and the production efficiency. For example, temperature: a key control parameter in the smelting process. Different smelting stages and processes have different requirements for temperature. Pressure: In some smelting processes, such as blast furnace smelting, the control of furnace pressure is crucial to ensure the stability and safety of the smelting process. Flow rate: such as the flow rate of gas, liquid or solid, which directly affects the material transfer and energy transfer during the smelting process. Concentration: The concentration of smelting raw materials or reactants, which determines the chemical reaction rate and the purity of the product during the smelting process. The equipment status parameters reflect the current operating status and performance of the smelting equipment. They are crucial to ensure the smooth progress of the smelting process and the long-term stable operation of the equipment. Temperature: The temperature of the equipment, such as the furnace shell temperature, electrode temperature, etc., too high or too low temperature may cause damage to the equipment. Pressure: the pressure inside or outside the equipment, such as the pressure inside the furnace, the pressure in the gas transmission pipeline, etc. Vibration: the vibration of the equipment. Abnormal vibration may be a precursor to equipment failure. Electrical parameters: such as voltage, current, resistance, etc., which reflect the electrical performance and operating status of the equipment. Energy consumption parameters are important indicators for measuring the energy consumption of the smelting process. They are of great significance for evaluating the energy efficiency of the smelter, formulating energy-saving measures and reducing production costs. For example, power consumption: the energy consumption of power equipment in the smelting process, such as the power consumption of electric arc furnaces, ore-fired furnaces, etc. Fuel consumption: the consumption of fuels such as coke, coal, and natural gas. New water consumption: the amount of new water used in the smelting process.

[0054] In this embodiment, the process parameters, the equipment status parameters and the energy consumption parameters will affect each other, thereby affecting the energy consumption parameters, so it is necessary to analyze the correlation between the three.

[0055] In some embodiments of the present application, the analysis reflects the correlation between information, including:

[0056] The reflected information includes process parameters, equipment status parameters and energy consumption parameters;

[0057] Calculate the mutual information between process parameters, equipment status parameters and energy consumption parameters, sort the mutual information from large to small, and determine the maximum value, median value and minimum value. Divide the mutual information into several intervals, and determine the frequent intervals in the interval by the number of times the mutual information appears in the interval.

[0058] The mutual information threshold is set according to the positional relationship between the frequent interval and the maximum value, median value and minimum value, and the mutual information is filtered, so that the correlation relationship reflecting the information corresponding to the mutual information is retained.

[0059] In this embodiment, mutual information (MI) is an indicator to measure the amount of shared information between two variables. It does not depend on a specific signal distribution model, can process discrete, continuous or mixed distribution variables, and is not limited to describing linear dependencies, but can also detect nonlinear dependencies. The larger the mutual information value, the stronger the dependency between the two variables. The statistical characteristic values ​​(maximum value, median value, minimum value, frequent interval) of all mutual information are combined to screen for relationships with strong correlation.

[0060] It can be understood that the frequent interval refers to the interval where the mutual information value appears most times.

[0061] In some embodiments of the present application, the mutual information threshold is set according to the positional relationship between the frequent interval and the maximum value, the median value, and the minimum value, including:

[0062] When the frequent interval is between the maximum value and the median value, the median value is adjusted according to the distance between the median value and the frequent interval, and the adjusted median value is used as the mutual information threshold;

[0063] When the median value is within the frequent interval, the lower endpoint of the frequent interval is adjusted according to the distance between the median value and the frequent interval, and the adjusted lower endpoint is used as the mutual information threshold;

[0064] When the frequent interval is between the median and the minimum, the minimum is adjusted according to the distance between the minimum and the frequent interval, and the adjusted minimum is used as the mutual information threshold;

[0065] Among them, the distances between the maximum value, median value, and minimum value and the frequent interval are the differences between the value and the closer endpoint in the frequent interval.

[0066] In this embodiment, the maximum value-median value is the upper half, and the median value-minimum value is the lower half. When the frequent interval is between the maximum value and the median value, it means that the mutual information value appears more often in the upper half. At this time, in order to retain more correlation relationships in the lower half, the median value is adjusted. Different median values ​​have different adjustment coefficients corresponding to the distances from the frequent intervals. The adjustment is completed by adjusting the coefficient*median value, and the median value is appropriately lowered to cover some correlation relationships in the lower half.

[0067] In this embodiment, when the median value is within the frequent interval, it means that the frequent interval is partially covered in both the upper and lower halves. In order to retain more correlation relationships in the lower half, the lower endpoint of the frequent interval is adjusted in the same manner as above.

[0068] In this embodiment, when the frequent interval is between the median value and the minimum value, the frequent interval is in the lower half. In order to eliminate some relationships with weaker correlation, the minimum value is adjusted and the relationships with stronger correlation are retained. The adjustment method is the same as above.

[0069] It should be noted that the above-mentioned adjustment of the median value, adjustment of the lower endpoint of the frequent interval, and adjustment of the minimum value do not cover the original median value, lower endpoint, and minimum value, but generate a new threshold for screening based on these three.

[0070] Step S102, construct time series data reflecting information, determine trend components reflecting information, extract features reflecting information, and combine multiple features according to correlation relationships between the reflected information to obtain feature combination items.

[0071] In some embodiments of the present application, time series data reflecting information is constructed to determine trend components reflecting information, including:

[0072] The data of process parameters, equipment status parameters and energy consumption parameters are time stamped and aligned to obtain time series data, and the time series data is smoothed;

[0073] Decompose the time series data through the time series decomposition algorithm to determine the trend components of process parameters, equipment status parameters and energy consumption parameters;

[0074] Among them, time series decomposition algorithms include STL decomposition and EMD decomposition.

[0075] In this embodiment, timestamp alignment: ensure that all reflected information has a unified timestamp so that they can be organized into time series data. Data interpolation: for missing data points, appropriate interpolation methods (such as linear interpolation, mean interpolation) are used to fill in to maintain the continuity of the time series. Data smoothing: smoothing the time series data to reduce the impact of noise and outliers.

[0076] In some embodiments of the present application, multiple features are combined according to the correlation between the reflected information to obtain feature combination items, including:

[0077] The features of process parameters, equipment status parameters and energy consumption parameters are extracted to obtain process characteristics, equipment status characteristics and energy consumption characteristics. The corresponding features of process characteristics, equipment status characteristics and energy consumption characteristics are cross-featured by reflecting the correlation between the information to form feature combination items. The feature combination items are marked with feature categories and mutual information under the correlation relationship.

[0078] In this embodiment, the correlation relationship between information is the relationship between two parameters. The feature combination term here is also a combination of two features. Two or more features are crossed to generate new features. For example, for numerical features, multiplication or addition operations can be performed. Feature crossing helps to capture the nonlinear relationship between features. For example, multiplying temperature and energy consumption to obtain a new feature that reflects the impact of temperature on energy consumption.

[0079] Step S103, training the smelting energy consumption analysis model through feature combination items and trend components, predicting the normal range of future energy consumption, comparing the actual energy consumption with the normal range of future energy consumption to determine the abnormal energy consumption segment, and analyzing the abnormal energy consumption segment to trace the cause of the abnormal energy consumption.

[0080] In this embodiment, the model structure parameters are determined by the attributes of the feature combination items so that the complexity of the model matches the actual smelting characteristics. The model structure parameters include the number of neural network layers, the number of neurons in each layer, the type of activation function, etc. The smelting energy consumption analysis model can be a recurrent neural network (RNN), LSTM (long short-term memory network), etc. The number of feature combinations reflects the complexity of the data and the amount of information that the model needs to process. Mutual information can also help us determine the complexity of the model structure. If the mutual information between features is generally low, it means that the correlation between them is weak, and a model with a simpler structure can be selected at this time. On the contrary, if the mutual information between features is high and complex, a more complex model is required to capture these relationships. The number of feature combinations directly reflects the complexity and correlation between features in the data set. When the number of feature combinations is large, it means that the interactions and dependencies between features in the data set are more complex, which usually requires a more complex model structure to capture these relationships. On the contrary, when the number of feature combinations is small, the relationship between features in the data set may be simpler, and a simpler model structure can be selected.

[0081] In some embodiments of the present application, the smelting energy consumption analysis model is trained by characteristic combination items and trend components, including:

[0082] Compare the feature categories under all feature combination items, record the feature combination items with overlapping process features or equipment status features between different feature combination items as overlapping feature combination items, and record the overlapping process features or equipment status features as overlapping features;

[0083] Determine the model structure level based on feature combination items, overlapping feature combination items and mutual information;

[0084] ;

[0085] in, is the model structure level, , are the number of feature combination items and the combination weight corresponding to the mutual information, , are the conversion coefficients of feature combination items and overlapping feature combination items respectively, , are the number of feature combinations and overlapping feature combinations respectively, For the The mutual information value of feature combinations, is the number of overlapping features, For the The sum of the mutual information of multiple overlapping feature combinations corresponding to overlapping features, , All are preset constants;

[0086] Different model structure levels correspond to multiple combinations of different model structure parameters. The trend components of the features in each feature combination item are added to the corresponding feature combination item. All feature combination items are divided into training sets and test sets. The smelting energy consumption analysis model is trained using the training sets and test sets.

[0087] In this embodiment, feature overlap means that the same features are included in different feature combinations. For example, if feature combination 1 is "process parameter 1 and equipment state parameter 1", and feature combination 2 is "process parameter 1 and energy consumption parameter 1", then "process parameter 1" overlaps in these two feature combinations. The feature combination items, the number of overlapping feature combination items, and the mutual information jointly determine the model structure level.

[0088] In this embodiment, the model structure level is determined by the number of feature combination items, overlapping feature combination items and mutual information. It indicates the correction of the mutual information average of overlapping feature combinations to the mutual information average of feature combinations. The higher the model structure level, the larger the corresponding model structure parameter.

[0089] In some embodiments of the present application, in the process of training the smelting energy consumption analysis model through the training set and the test set, the method further includes:

[0090] The trend component of each feature in the feature combination item is statistically analyzed to determine the overall change rate of the feature combination item. All feature combination items are grouped according to the overall change rate of the feature combination items. The learning rate is set for the overall change rate of the feature combination items in each group. The smelting energy consumption analysis model is trained by combining the training set and the test set.

[0091] In this embodiment, the trend component is the speed or trend of change of each feature. For feature combinations that change faster, selecting a smaller learning rate can help the model capture these changes more promptly, and avoid the model "not keeping up" with the changes in data due to excessive learning rates. For feature combinations that change slower, the learning rate can be appropriately increased to speed up the convergence of the model. The trend components of the two features in the statistical feature combination are combined to obtain the overall rate of change (weighted sum or average). Grouped learning rate refers to dividing features or parameters into different groups during the training process of the model and setting different learning rates for each group. This strategy is often used to deal with features with different importance or different speeds of change. By setting different learning rates for each group of features, the training speed and performance of the model can be better balanced. Features of different groups of models can be trained using different learning rates. This strategy helps to adapt to the learning speed and complexity of different layers or features.

[0092] In some embodiments of the present application, the actual energy consumption is compared with the normal range of future energy consumption to determine the abnormal energy consumption segment, and the abnormal energy consumption segment is analyzed to trace the cause of the abnormal energy consumption, including:

[0093] Construct a curve of actual energy consumption changing over time, recorded as the energy consumption curve, mark the normal range of future energy consumption on the energy consumption curve, and the part of the energy consumption curve where the actual energy consumption exceeds the normal range of future energy consumption is regarded as the abnormal energy consumption segment, and determine the timestamp of the abnormal energy consumption segment, recorded as the abnormal energy consumption time;

[0094] Calculate the curve area between the energy consumption abnormality period and the future normal energy consumption range, collect the reflection information of the period before and after the energy consumption abnormality time and during the energy consumption abnormality time, and construct the reflection information set to be tested;

[0095] Calculate the state transition probability of each abnormal cause in the measured reflection information set and the energy consumption abnormal cause library, and determine the comprehensive transition probability by combining the curve area;

[0096] ;

[0097] in, The first The combined transition probability of the abnormal causes is The first The state transition probability of each abnormal cause, The curve area and energy consumption abnormality cause library The deviation of energy consumption abnormality due to abnormal reasons, is the compensation probability, Indicates the curve area and the energy consumption abnormality cause library The compensation probability obtained by mapping the similarity of the deviation of the energy consumption abnormality of each abnormal reason;

[0098] Each abnormal cause in the energy consumption abnormal cause library is screened according to the comprehensive transfer probability, so as to trace the abnormal cause of energy consumption.

[0099] In this embodiment, process parameters and equipment state parameters are input into the smelting energy consumption analysis model, and target variable energy consumption parameters are output. The curve area between the energy consumption abnormality segment and the future normal range of energy consumption indicates the degree of energy consumption abnormality, indicating that energy consumption is abnormal. These factors may include equipment failure, improper adjustment of process parameters, etc. Each abnormal cause in the energy consumption abnormality cause library is provided with a corresponding energy consumption abnormality degree and abnormal state. The state transition probability of each abnormal cause in the energy consumption abnormality cause library is calculated. Here, the state transition probability is realized by using a Markov chain. First, the abnormal state of each abnormal cause in the energy consumption abnormality cause library is configured, and the state transition probability of the current state (the reflected information set to be tested) and the abnormal state of the abnormal cause is calculated to describe the degree of abnormal matching.

[0100] In this embodiment, is the similarity between the current curve area (current energy consumption abnormality) and the energy consumption abnormality of the abnormal cause, It is the Euclidean distance or Manhattan distance, which is used to describe the difference between the two. A compensation probability is determined by similarity, and the state transition probability is adjusted to obtain a comprehensive transition probability. The larger the probability, the more it matches the cause of the anomaly.

[0101] Step S104, optimizing the smelting process by tracking the causes of abnormal energy consumption.

[0102] In this embodiment, the causes of abnormal energy consumption are continuously tracked and monitored, and targeted optimization suggestions are put forward, such as adjusting process parameters, improving equipment maintenance plans, optimizing raw material ratios, etc.

[0103] Compared with the prior art, the present invention has the following beneficial effects:

[0104] 1. Analyze the correlation between information, provide a basis for subsequent multi-feature combination, and ensure the reliability of multi-factor association calculation. Combine multiple features according to the correlation between information, train the smelting energy consumption analysis model through feature combination items and trend components, combine multiple features for correlation, consider the correlation between different features, and add trend components to make the smelting energy consumption analysis model prediction more accurate.

[0105] 2. Analyze the abnormal energy consumption segments to trace the causes of abnormal energy consumption. Determine the deviation of energy consumption through the predicted normal range of energy consumption, thereby determining the causes of abnormal energy consumption. Follow-up monitoring of these abnormal energy consumption causes and optimization of the smelting process ensure the reliability and effectiveness of the information management of the smelting plant.

[0106] Correspondingly, this application also provides an intelligent smelting plant information management system, such as Figure 2 As shown, including,

[0107] The analysis module is used to collect the reflected information in the smelting process and analyze the correlation between the reflected information;

[0108] The feature module is used to construct time series data reflecting information, determine the trend components reflecting information, extract the features reflecting information, and combine multiple features according to the correlation relationship between the reflected information to obtain feature combination items;

[0109] The prediction module is used to train the smelting energy consumption analysis model through feature combination items and trend components, predict the normal range of future energy consumption, compare the actual energy consumption with the normal range of future energy consumption to determine the abnormal energy consumption segment, and analyze the abnormal energy consumption segment to trace the cause of abnormal energy consumption;

[0110] Management module, used to optimize the smelting process by tracking the causes of abnormal energy consumption.

[0111] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present invention can be implemented by hardware, or by software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present invention.

[0112] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required for implementing the present invention.

[0113] Those skilled in the art will appreciate that the modules in the system in the implementation scenario can be distributed in the system of the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more systems different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0114] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An information management method for an intelligent smelting plant, characterized in that: include, Collect the reflected information in the smelting process and analyze the correlation between the reflected information; Construct time series data reflecting information, determine the trend components reflecting information, extract the features reflecting information, and combine multiple features according to the correlation between the reflected information to obtain feature combination items; Through the training of the smelting energy consumption analysis model by characteristic combination items and trend components, the normal range of future energy consumption is predicted, the actual energy consumption is compared with the normal range of future energy consumption to determine the abnormal energy consumption segment, and the abnormal energy consumption segment is analyzed to trace the cause of abnormal energy consumption; Optimize the smelting process by tracking the causes of abnormal energy consumption; in, Analyze the correlation between information, including: The reflected information includes process parameters, equipment status parameters and energy consumption parameters; Calculate the mutual information between process parameters, equipment status parameters and energy consumption parameters, sort the mutual information from large to small, and determine the maximum value, median value and minimum value. Divide the mutual information into several intervals, and determine the frequent intervals in the interval by the number of times the mutual information appears in the interval. The mutual information threshold is set according to the positional relationship between the frequent interval and the maximum value, median value, and minimum value, and the mutual information is filtered, so that the correlation relationship reflecting the information corresponding to the mutual information is retained; And according to the correlation between the information, multiple features are combined to obtain feature combination items, including: Extract features from process parameters, equipment status parameters and energy consumption parameters to obtain process features, equipment status features and energy consumption features. By reflecting the correlation between the information, cross-feature the corresponding features in the process features, equipment status features and energy consumption features to form feature combination items. The feature combination items are marked with feature categories and mutual information under the correlation relationship. The smelting energy consumption analysis model is trained through characteristic combination items and trend components, including: Compare the feature categories under all feature combination items, record the feature combination items with overlapping process features or equipment status features between different feature combination items as overlapping feature combination items, and record the overlapping process features or equipment status features as overlapping features; Determine the model structure level based on feature combination items, overlapping feature combination items and mutual information; ； in, is the model structure level, , are the number of feature combination items and the combination weights corresponding to the mutual information, respectively. , are the conversion coefficients of feature combination items and overlapping feature combination items respectively, , are the number of feature combinations and overlapping feature combinations respectively, For the The mutual information value of feature combinations, is the number of overlapping features, For the The sum of the mutual information of multiple overlapping feature combinations corresponding to overlapping features, , All are preset constants; Different model structure levels correspond to multiple combinations of different model structure parameters. The trend components of the features in each feature combination item are added to the corresponding feature combination item. All feature combination items are divided into training sets and test sets. The smelting energy consumption analysis model is trained using the training sets and test sets.

2. The information management method for an intelligent smelting plant according to claim 1 is characterized in that: The mutual information threshold is set according to the positional relationship between the frequent interval and the maximum value, median value, and minimum value, including: When the frequent interval is between the maximum value and the median value, the median value is adjusted according to the distance between the median value and the frequent interval, and the adjusted median value is used as the mutual information threshold; When the median value is within the frequent interval, the lower endpoint of the frequent interval is adjusted according to the distance between the median value and the frequent interval, and the adjusted lower endpoint is used as the mutual information threshold; When the frequent interval is between the median and the minimum, the minimum is adjusted according to the distance between the minimum and the frequent interval, and the adjusted minimum is used as the mutual information threshold; Among them, the distances between the maximum value, median value, and minimum value and the frequent interval are the differences between the value and the closer endpoint in the frequent interval.

3. The information management method for an intelligent smelting plant according to claim 1 is characterized in that: Construct time series data that reflects information and determine the trend components that reflect information, including: The data of process parameters, equipment status parameters and energy consumption parameters are time stamped and aligned to obtain time series data, and the time series data is smoothed; Decompose the time series data through the time series decomposition algorithm to determine the trend components of process parameters, equipment status parameters and energy consumption parameters; Among them, time series decomposition algorithms include STL decomposition and EMD decomposition.

4. The intelligent smelting plant information management method according to claim 1 is characterized in that: In the process of training the smelting energy consumption analysis model through the training set and the test set, the method further includes: The trend component of each feature in the feature combination item is statistically analyzed to determine the overall change rate of the feature combination item. All feature combination items are grouped according to the overall change rate of the feature combination items. The learning rate is set for the overall change rate of the feature combination items in each group. The smelting energy consumption analysis model is trained by combining the training set and the test set.

5. The intelligent smelting plant information management method according to claim 1 is characterized in that: Compare the actual energy consumption with the normal range of future energy consumption to determine the abnormal energy consumption segment, and analyze the abnormal energy consumption segment to trace the cause of abnormal energy consumption. include, Construct a curve of actual energy consumption changing over time, recorded as the energy consumption curve, mark the normal range of future energy consumption on the energy consumption curve, and the part of the energy consumption curve where the actual energy consumption exceeds the normal range of future energy consumption is regarded as the abnormal energy consumption segment, and determine the timestamp of the abnormal energy consumption segment, recorded as the abnormal energy consumption time; Calculate the curve area between the energy consumption abnormality period and the future normal energy consumption range, collect the reflection information of the period before and after the energy consumption abnormality time and during the energy consumption abnormality time, and construct the reflection information set to be tested; Calculate the state transition probability of each abnormal cause in the measured reflection information set and the energy consumption abnormal cause library, and determine the comprehensive transition probability by combining the curve area; ； in, The first The combined transition probability of the abnormal causes is The first The state transition probability of each abnormal cause, The curve area and energy consumption abnormality cause library The deviation of energy consumption abnormality due to abnormal reasons, is the compensation probability, Indicates the curve area and the energy consumption abnormality cause library The compensation probability obtained by mapping the similarity of the deviation of the energy consumption abnormality of each abnormal reason; Each abnormal cause in the energy consumption abnormal cause library is screened according to the comprehensive transfer probability, so as to trace the abnormal cause of energy consumption.

6. Intelligent smelting plant information management system, characterized by: For implementing the method according to any one of claims 1 to 5, the system comprises: The analysis module is used to collect the reflected information in the smelting process and analyze the correlation between the reflected information; The feature module is used to construct time series data reflecting information, determine the trend components reflecting information, extract the features reflecting information, and combine multiple features according to the correlation relationship between the reflected information to obtain feature combination items; The prediction module is used to train the smelting energy consumption analysis model through feature combination items and trend components, predict the normal range of future energy consumption, compare the actual energy consumption with the normal range of future energy consumption to determine the abnormal energy consumption segment, and analyze the abnormal energy consumption segment to trace the cause of abnormal energy consumption; Management module, used to optimize the smelting process by tracking the causes of abnormal energy consumption.

Images