An energy collection and monitoring system based on big data

Abstract

The present invention discloses an energy collection and monitoring system based on big data, which specifically relates to the field of energy monitoring and is used to solve the problem of energy consumption anomaly identification and management optimization in steel production. The system collects energy parameters through the production process, and then constructs an energy signature reference model of multi-dimensional energy consumption characteristics based on historical data and equipment operation characteristics, providing an accurate benchmark for dynamic comparison and anomaly detection; utilizes a comprehensive analysis of energy consumption deviation distribution and load characteristics to accurately identify potential anomalies, and deeply mines key influencing factors and root causes through a multivariate machine learning algorithm; combines a predictive scheduling engine and an online optimization algorithm to dynamically adjust process parameters, equipment operation status and task scheduling, effectively eliminate anomalies and restore energy consumption balance; thereby realizing real-time monitoring of process energy consumption and intelligent anomaly detection, improving the level of refinement of energy management, improving the energy utilization efficiency and production stability of steel enterprises, and reducing potential production risks.

Description

Technical Field

[0001] The present invention relates to the field of energy monitoring, and more specifically, to an energy acquisition and monitoring system based on big data. Background Art

[0002] During the steel production process, each process link, such as ironmaking, steelmaking, and rolling, exhibits significant energy consumption characteristics, forming a unique energy signature. Traditional energy monitoring systems rely primarily on aggregate energy consumption statistics or simple threshold alarms, making it difficult to conduct in-depth analysis and identify potential energy consumption anomalies within each process link. This monitoring approach fails to promptly detect hidden process anomalies, leading to energy waste and reduced production efficiency. With the development of big data technology and IoT devices, in-depth analysis based on multi-source data has become possible, facilitating the construction of a refined energy monitoring system, enabling real-time monitoring of process energy consumption and intelligent anomaly detection, thereby improving energy management and production optimization capabilities.

[0003] Existing energy monitoring systems are unable to accurately identify and analyze energy consumption anomalies across various steel production processes. This lack of in-depth analysis and dynamic comparison of energy signatures results in delayed anomaly detection and ineffective guidance for energy optimization and troubleshooting. This limitation limits the intelligent and refined nature of energy management, making it difficult to meet the demands of modern steel companies for efficient and accurate energy monitoring. Therefore, there is an urgent need for a big data-based energy acquisition and monitoring system that can accurately monitor process energy consumption and detect anomalies through multi-source data fusion and in-depth analysis, thereby improving energy efficiency and reducing production risks.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an energy collection and monitoring system based on big data, which collects energy parameters through the production process, and then constructs an energy signature reference model of multi-dimensional energy consumption characteristics based on historical data and equipment operation characteristics, providing an accurate benchmark for dynamic comparison and anomaly detection; utilizes a comprehensive analysis of energy consumption deviation distribution and load characteristics to accurately identify potential anomalies, and deeply mines key influencing factors and root causes through multivariate machine learning algorithms; combines a predictive scheduling engine and an online optimization algorithm to dynamically adjust process parameters, equipment operating status and task scheduling, effectively eliminate anomalies and restore energy consumption balance; thereby realizing real-time monitoring of process energy consumption and intelligent anomaly detection, improving the level of refinement of energy management, improving the energy utilization efficiency and production stability of steel enterprises, and reducing potential production risks, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An energy collection and monitoring system based on big data, comprising: a data collection module, a data processing module, a feature modeling module, an anomaly detection module, a correlation analysis module and a scheduling optimization module;

[0008] Data acquisition module: Utilizes pre-placed sensors to collect energy-related parameters in real time and upload them to the big data processing platform. The output data set is then passed to the data processing module.

[0009] Data processing module: This module converts the uploaded data into a unified format and removes noise signals to ensure data accuracy and consistency. It also indexes and stores the data by timestamp and process number, and outputs a standardized data set that is passed to the feature modeling module.

[0010] Feature modeling module: Based on historical data and equipment operating characteristics, it extracts the multi-dimensional energy consumption characteristics of each process link, constructs an energy signature reference model as a benchmark for normal energy consumption status, and transmits the output energy signature reference model to the anomaly detection module;

[0011] Anomaly Detection Module: Continuously receives real-time energy consumption data, comprehensively evaluates the deviation distribution of multi-dimensional energy consumption states and load concentration characteristics, dynamically compares them with the energy signature reference model, identifies abnormal data points that exceed the normal energy consumption range, and transmits the detected abnormal data to the correlation analysis module;

[0012] Correlation Analysis Module: This module conducts in-depth correlation analysis on detected abnormal data, combines equipment status and production parameters, and uses multivariate machine learning algorithms to determine the root cause of the abnormality and classify and archive it. The output abnormality classification archive and root cause analysis results are then passed to the scheduling optimization module.

[0013] Scheduling Optimization Module: Utilizes a predictive scheduling engine and online energy optimization algorithms to evaluate the effectiveness of adjustment plans. By adjusting key process temperatures, switching backup fuel supply lines, and issuing production line operation sequence adjustment instructions, it can immediately eliminate energy consumption anomalies and ensure production and furnace temperature balance.

[0014] In a preferred embodiment, the data acquisition module includes the following:

[0015] Using multiple types of sensors arranged at key process nodes, multiple energy parameters are collected in real time at fixed sampling intervals to construct an energy state vector to characterize the comprehensive energy consumption status at each sampling moment; through a dynamic partition scheduling strategy, the data collected by multiple types of sensors are classified and stored in distributed data nodes, and the timestamp is used as the primary key to generate an original data set with a unique identifier; the cleaned original data set is marked according to the timestamp and process number to generate a structured data set.

[0016] In a preferred embodiment, the data processing module includes the following:

[0017] Extract multi-dimensional energy parameters from the data set uploaded by the data acquisition module, and preliminarily analyze the type of source sensor and data encoding format;

[0018] Based on the rule-based dynamic mapping table, the data encoding formats uploaded by sensors from different sources are converted and an initial data set with a unified format is generated;

[0019] The noise signal in the initial data set is removed using the wavelet transform method;

[0020] An index matrix is ​​constructed by combining the timestamp and process number, and the cleaned data is assigned to the time series storage unit of the corresponding process, ensuring that the data is continuous in the time dimension and independent in the process dimension;

[0021] Generate a standardized dataset that has been format converted, denoised, and indexed.

[0022] In a preferred embodiment, the feature modeling module includes the following:

[0023] From the standardized data set, the data series is divided by process number to form an independent time series data set for each process link, so as to conduct group analysis on the energy consumption characteristics of different processes;

[0024] For each process step, the time series data is decomposed into basic consumption trend and periodic fluctuation components by applying the adaptive weighted curve decomposition method.

[0025] Based on the basic consumption trend, key inflection point characteristics are calculated, including the maximum change rate and the time position of the energy consumption inflection point, to capture the critical moments of energy demand;

[0026] For periodic fluctuation components, short-time Fourier transform is used to extract the main frequency components, and the energy concentration is calculated to quantify the fluctuation intensity.

[0027] The trend features and fluctuation features are combined to form a multi-dimensional energy consumption feature vector, and the high-dimensional features are embedded into a low-dimensional space through a dimensionality reduction algorithm to form a visual feature distribution map;

[0028] The density clustering algorithm is used to perform density clustering analysis on the multidimensional energy consumption feature vectors in the feature distribution map. By setting the core point density threshold and the minimum number of samples, the dense areas of energy consumption characteristics are identified and divided into clusters with similar energy consumption patterns. The center point of the cluster with the highest density is extracted as the typical energy consumption state of each process link, and the upper and lower limits of the feature range are calculated to form an energy signature reference model with boundary conditions, which serves as the normal energy consumption benchmark for real-time comparison.

[0029] In a preferred embodiment, the anomaly detection module includes the following:

[0030] The typical energy consumption state center value and energy consumption characteristic boundary of each process link are extracted from the energy signature reference model as the comparison benchmark; the real-time energy consumption data set is loaded into the comparison module point by point, and the energy consumption deviation symmetry index and instantaneous load energy consumption compression rate index are calculated and processed for the real-time energy consumption state vector at each time point.

[0031] In a preferred embodiment, the energy consumption deviation symmetry index measures whether the deviation of energy consumption data in each dimension is evenly distributed by performing a symmetry analysis on the deviation of the real-time energy consumption state vector; specifically, the energy consumption deviation symmetry index decomposes the energy consumption deviation into two dimensions: absolute value and square value during the calculation process: the average absolute deviation reflects the overall level of the deviation amplitude, while the square root of the mean square deviation emphasizes the contribution of the dimension with a large deviation; the absolute value of the difference between the two is used to characterize the distribution characteristics of the deviation of each dimension, that is, the energy consumption deviation symmetry index.

[0032] In a preferred embodiment, the instantaneous load energy consumption compression rate index is intended to analyze the concentration of real-time energy consumption data in the load distribution and reveal whether there is abnormal load aggregation in the energy consumption state vector. During the calculation, the real-time energy consumption data is first converted into a standard load value with relative weight significance through normalization processing to ensure that the data of each dimension are compared under a unified dimension. Subsequently, the instantaneous load energy consumption compression rate index quantifies the load concentration by calculating the ratio of the range of the load value to the total load. At the same time, a logarithmic function is introduced to enhance the sensitivity to the load extreme value ratio, thereby identifying abnormal characteristics caused by extreme load distribution.

[0033] In a preferred embodiment, the energy consumption deviation symmetry index and the instantaneous load energy consumption compression rate index are weighted and summed to form a comprehensive comparison parameter for comprehensively evaluating the degree of abnormality of real-time energy consumption data; a comprehensive comparison threshold is set, and when the comprehensive comparison parameter is greater than the comprehensive comparison threshold, it is determined that there is an energy consumption abnormality at the corresponding time point; the information of the abnormal data point is recorded and stored in the abnormal data log.

[0034] In a preferred embodiment, the association analysis module includes the following contents:

[0035] Extract deviation values, corresponding process links, and time points from abnormal data logs as initial input for deep correlation analysis;

[0036] By combining the equipment status dataset and the process parameter dataset, we conduct correlation modeling on the abnormality occurrence time point and its adjacent time window, and preliminarily screen the key influencing factors.

[0037] A multivariate time series regression model is applied, with energy consumption deviation as the target variable and all variables in equipment status data and process parameter data as input factors. By calculating the product of the partial derivative coefficient of each factor and the actual value, the contribution of each factor to energy consumption deviation is quantified.

[0038] Sort the factors in descending order according to their impact contribution, select the top N factors with the largest contribution values ​​as key factors, form an abnormal feature factor set, and record their corresponding feature deviation values ​​and time periods;

[0039] Based on the abnormal feature factor set, the clustering algorithm is used to classify abnormal events, classify abnormalities with similar patterns into the same category, and mark the classification results and root cause types to form an abnormal classification archive.

[0040] In a preferred embodiment, the scheduling optimization module includes the following:

[0041] Extract anomaly types, key factor sets, and root cause analysis results from anomaly classification archives as input for scheduling and optimization;

[0042] Based on the predictive scheduling model, a time series prediction algorithm is used to make short-term predictions of the future status of key factors, a prediction function is constructed, and the prediction is completed by combining a multivariate regression model with other relevant factors;

[0043] Match the prediction results with the energy signature reference model, simulate future energy consumption deviations, calculate the dynamic response effects of different adjustment schemes, and select the adjustment strategy with the shortest response time and the best recovery effect to form the optimal adjustment plan;

[0044] After determining the adjustment plan, the key process parameters are revised through the online optimization algorithm. The adjustment plan includes:

[0045] Update key process temperature setting values ​​to ensure that temperature parameters gradually approach the energy signature reference model;

[0046] Switch backup fuel supply lines and adjust supply priorities based on gas flow demand;

[0047] Adjust the production line operation sequence by optimizing the task allocation matrix, dynamically adjusting task priorities and scheduling sequences based on the root cause analysis results of energy consumption anomalies and real-time process load status;

[0048] The adjusted energy consumption data is dynamically compared with the energy signature reference model. If the deviation value gradually decreases and tends to the predetermined range, the adjustment effect is recorded and the abnormal state is marked as recovered.

[0049] The technical effects and advantages of the energy collection and monitoring system based on big data of the present invention are as follows:

[0050] The present invention combines multiple types of sensors to collect energy-related parameters in real time. It uses a unified format conversion, noise signal removal, and a preprocessing method with timestamp and process number indexing to ensure data accuracy and consistency. Furthermore, it constructs an energy signature reference model with multidimensional energy consumption characteristics based on historical data and equipment operating characteristics, providing an accurate benchmark for real-time energy consumption monitoring and dynamic comparison. Through a comprehensive assessment of energy consumption state deviation distribution and load concentration characteristics, it accurately identifies abnormal data points. Using a multivariate machine learning algorithm, it conducts in-depth correlation analysis combining equipment status and process parameters to effectively locate the root cause of abnormalities and key influencing factors. Furthermore, through a predictive scheduling engine and an online energy consumption optimization algorithm, it generates refined adjustment plans, including process parameter revisions, equipment switching, and task scheduling adjustments, thereby achieving immediate elimination of energy consumption anomalies. Ultimately, it dynamically updates the energy signature reference model, further enhancing the model's adaptability and monitoring accuracy. Furthermore, through real-time collection, dynamic comparison, in-depth analysis, and closed-loop optimization of energy consumption data, it achieves intelligent, refined, and efficient energy management, significantly improving energy utilization efficiency and production stability in steel production, and reducing energy waste and potential equipment operation risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a structural diagram of an energy collection and monitoring system based on big data in the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] Example 1: Figure 1 The present invention provides an energy acquisition and monitoring system based on big data, comprising: a data acquisition module, a data processing module, a feature modeling module, an anomaly detection module, a correlation analysis module and a scheduling optimization module;

[0054] Data acquisition module: Utilizes pre-placed sensors to collect energy-related parameters in real time and upload them to the big data processing platform. The output data set is then passed to the data processing module.

[0055] Data processing module: This module converts the uploaded data into a unified format and removes noise signals to ensure data accuracy and consistency. It also indexes and stores the data by timestamp and process number, and outputs a standardized data set that is passed to the feature modeling module.

[0056] Feature modeling module: Based on historical data and equipment operating characteristics, it extracts the multi-dimensional energy consumption characteristics of each process link, constructs an energy signature reference model as a benchmark for normal energy consumption status, and transmits the output energy signature reference model to the anomaly detection module;

[0057] Anomaly Detection Module: Continuously receives real-time energy consumption data, comprehensively evaluates the deviation distribution of multi-dimensional energy consumption states and load concentration characteristics, dynamically compares them with the energy signature reference model, identifies abnormal data points that exceed the normal energy consumption range, and transmits the detected abnormal data to the correlation analysis module;

[0058] Correlation Analysis Module: This module conducts in-depth correlation analysis on detected abnormal data, combines equipment status and production parameters, and uses multivariate machine learning algorithms to determine the root cause of the abnormality and classify and archive it. The output abnormality classification archive and root cause analysis results are then passed to the scheduling optimization module.

[0059] Scheduling Optimization Module: Utilizes a predictive scheduling engine and online energy optimization algorithms to evaluate the effectiveness of adjustment plans. By adjusting key process temperatures, switching backup fuel supply lines, and issuing production line operation sequence adjustment instructions, it can immediately eliminate energy consumption anomalies and ensure production and furnace temperature balance.

[0060] In the steel production process, energy consumption at each stage often exhibits a complex alternation between transient and steady-state states. Real-time collection of high-dimensional, multi-category energy parameters is fundamental to building a precise energy monitoring system. Traditional data collection methods, lacking effective screening and correlation of multi-dimensional features, struggle to meet the stringent requirements of modern energy management for data accuracy, real-time performance, and system compatibility. By rationally deploying multiple sensor types and employing efficient data collection and transmission mechanisms, high-quality input can be provided for subsequent data processing and anomaly analysis.

[0061] The data acquisition module includes the following:

[0062] Using multiple types of sensors (including power monitoring modules, gas flow meters, temperature sensors, and pressure sensors) deployed at key process nodes such as ironmaking, steelmaking, and steel rolling, multiple energy parameters such as power consumption P(t), gas flow F(t), temperature T(t), and pressure p(t) are collected in real time at fixed sampling intervals to construct the energy state vector X(t) = [P(t), F(t), T(t), p(t)]. T , to characterize the comprehensive energy consumption state at each sampling moment.

[0063] Through the dynamic partition scheduling strategy, the data collected by multiple types of sensors are classified and stored in distributed data nodes, and the timestamp t is used as the primary key to generate the original data set with a unique identifier Ensure that the data collection process has temporal continuity and spatial distribution integrity;

[0064] During the data upload process, edge computing devices are used to perform preliminary cleaning of the original data set, and physically unacceptable abnormal points are eliminated based on the multivariate nonlinear filtering algorithm. For example, when detecting gas flow fluctuations, the fluctuation identification formula based on empirical mode decomposition is applied: ΔF=F(t)-F trend (t), if |ΔF|>∈, F(t) is eliminated; where F trend (t) represents the trend value of the gas flow rate corresponding to time t. It is the long-term trend of flow rate changes extracted by data decomposition or trend analysis algorithms (such as empirical mode decomposition (EMD) or sliding average method), and is used to represent the flow rate benchmark under normal conditions.

[0065] ΔF is the flow deviation, which is calculated as the difference between the original flow value and the trend flow value and is used to detect abnormal fluctuation amplitude of flow.

[0066] ∈ represents the abnormal detection threshold of gas flow fluctuation, and its value needs to be set according to the analysis results of historical data to ensure that abnormal points can be accurately identified.

[0067] Mark the cleaned raw data set by timestamp and process number to generate a structured data set Provide standardized input for subsequent format unification and multi-dimensional feature extraction.

[0068] D clean Represents the cleaned dataset after format conversion and noise removal.

[0069] X ′ (t i ) represents time t i The energy state vector after cleaning corresponding to the moment.

[0070] t i Represents time point i, which is a valid time point included in the cleaned dataset.

[0071] A time index set containing all timestamps that are considered valid. The raw data corresponding to these time points has been cleaned to exclude unreliable data.

[0072] The cleaned data vector is a subset of the original dataset, containing only data points that have passed the validity filter.

[0073] The cleaned and classified energy consumption data is uploaded to the big data processing platform through industrial protocols (such as MQTT or OPCUA), and the energy state vector at each moment is indexed within the platform to ensure efficient transmission of data from real-time collection to final storage, forming a time-continuous and logically rigorous energy consumption data stream to meet the high-quality input requirements of subsequent modeling and comparison.

[0074] Through this approach, the data acquisition module not only completes the real-time collection and cleaning of high-dimensional energy parameters, but also utilizes structured data streams and dynamic partitioning storage strategies to address the poor temporal continuity and insufficient data accuracy inherent in traditional data collection methods. This lays a solid foundation for subsequent format conversion, multidimensional feature extraction, and energy signature modeling. This approach boasts high precision and efficiency, and its multivariate filtering and distributed storage design demonstrate its technical depth.

[0075] During energy collection and monitoring, real-time data uploaded by sensors, due to complex sources, may suffer from inconsistent formats, noise interference, and timestamp misalignment. These flaws directly impact the accuracy and consistency of subsequent feature extraction and energy signature modeling. To ensure efficient data processing and accurate analysis, targeted data format conversion and noise removal technology must be designed to ensure that the data structure and content are highly reliable and available at both the physical and logical levels.

[0076] The data processing module includes the following:

[0077] Datasets uploaded from the data collection module Extract multi-dimensional energy parameters, including power consumption, gas flow, temperature and pressure, and preliminarily analyze the type of source sensor and data encoding format;

[0078] Design a rule-based dynamic mapping table to convert the encoding format of data uploaded by sensors from different sources, such as unifying different units of power data into kilowatt-hours and unifying different flow meter units into cubic meters per hour for flow data, and generate an initial data set with a unified format.

[0079] The wavelet transform method is used to remove the noise signal in the initial data set. The specific method is as follows:

[0080] For each time point, the energy state vector X fmt (t) Perform wavelet decomposition to obtain the low-frequency component L and the high-frequency component H, and only retain the components that meet the threshold condition |H|<λ H Components to filter high-frequency noise and reconstruct noise-free energy consumption data

[0081] Combined with timestamp ti Construct an index matrix with process number k Assign the cleaned data to the time series storage unit of the corresponding process to ensure the continuity of the data in the time dimension and independence in the process dimension;

[0082] Generate a standardized dataset after format conversion, noise removal and indexing The standardized dataset has high consistency and accuracy, providing input for the multi-dimensional feature extraction and energy signature modeling of the feature modeling module.

[0083] Through this process, the data processing module completes the comprehensive processing from raw data to standardized data. Utilizing dynamic format mapping, wavelet noise reduction, and index construction, it resolves format inconsistencies and noise interference issues in multi-source data integration, providing high-quality, standardized data input for subsequent energy signature modeling. This logically connects the data acquisition module's collection content and lays a precise foundation for the feature modeling module's modeling work, demonstrating its processing logic and data quality assurance capabilities.

[0084] In energy monitoring and management, extracting multidimensional energy consumption characteristics is a core step in achieving accurate modeling and anomaly detection. In-depth analysis of energy consumption characteristics across various process steps, based on historical data and equipment operating characteristics, not only reveals the patterns of equipment operation but also provides a clear benchmark for anomaly identification. However, energy consumption characteristics involve complex relationships across multiple dimensions and levels, and traditional univariate analysis methods are unable to effectively capture these high-dimensional correlations. Therefore, it is necessary to mine potential patterns from historical data and construct an energy signature reference model that can be used for real-time monitoring, laying the foundation for dynamic comparison and anomaly detection.

[0085] The feature modeling module includes the following:

[0086] Standardized datasets generated from the data processing module In the process, the data sequence is divided by process number k to form an independent time series data set D for each process step. k ={X std (t i ,k)|k∈K}, so as to conduct group analysis on the energy consumption characteristics of different processes;

[0087] The adaptive weighted curve decomposition method is applied to the time series data of each process link to decompose it into the basic consumption trend and the periodic fluctuation component W k (t), the formula is as follows: X std (t i ,k)=T k (t i )+W k (ti )+ε k (t i ); where T k (t i ) reflects the long-term energy consumption trend, W k (t i ) represents the periodic fluctuation characteristics, ε k (t i ) is random interference.

[0088] For basic consumption trend T k (t), calculate key inflection point characteristics, including the maximum rate of change and the time position t of the energy consumption inflection point peak , to capture critical moments of energy demand;

[0089] For the periodic fluctuation component W k (t), short-time Fourier transform is used to extract the main frequency component ν k , and calculate the energy concentration Quantify the intensity of fluctuations;

[0090] Combine trend features and fluctuation features to form a multi-dimensional energy consumption feature vector F k =

[0091] [Δ max ,t peak ,ν k ,E k ] and embed high-dimensional features into low-dimensional space through dimensionality reduction algorithms (such as t-SNE) to form a visual feature distribution map;

[0092] The density clustering algorithm is used to perform density clustering analysis on the multidimensional energy consumption feature vectors in the feature distribution map. By setting the core point density threshold ε and the minimum sample number MinPts, the dense areas of energy consumption features are identified and divided into clusters with similar energy consumption patterns; the center point of the cluster with the highest density is extracted. As the typical energy consumption state of each process link, and the upper and lower limits of its calculation characteristic range, an energy signature reference model with boundary conditions is formed Serves as the normal energy consumption benchmark for subsequent real-time comparison.

[0093] The feature modeling module uses a density clustering algorithm to accurately divide the multidimensional distribution of energy consumption characteristics, successfully extracts the typical energy consumption status of each process link, and constructs an energy signature reference model containing central values ​​and boundary conditions. This model can clearly reflect the normal energy consumption range of each process link, which not only provides a clear judgment standard for subsequent real-time comparison, but also significantly improves the accuracy and reliability of energy consumption anomaly detection, avoiding the waste of resources and inefficiency caused by feature overlap or misjudgment in traditional methods. Through this process, the establishment of the energy signature model closely matches the background technology's need for in-depth exploration of multidimensional energy consumption patterns, helping to upgrade the steel industry's energy management to be intelligent and refined.

[0094] The anomaly detection module includes the following:

[0095] Extract the typical energy consumption state center value of each process link from the energy signature reference model generated by the feature modeling module and energy consumption characteristic boundaries As a comparison benchmark.

[0096] The real-time energy consumption data set processed by the data processing module is loaded into the comparison module point by point. The real-time energy consumption state vector X at each time point is real (t i )=[x1(t i ),x2(t i ),...,x n (t i )] T , perform the following calculation and processing of the energy consumption deviation symmetry index and the instantaneous load energy consumption compression rate index.

[0097] Calculate the energy consumption deviation symmetry index EDSI:

[0098] The energy consumption deviation symmetry index is used to measure whether the deviation of real-time energy consumption data in various dimensions is symmetrical, reflecting the balance of the energy consumption state vector in multiple dimensions.

[0099] The Energy Consumption Deviation Symmetry Index measures whether the deviations of energy consumption data in each dimension are evenly distributed by performing a symmetry analysis on the deviations of the real-time energy consumption state vector. Specifically, during the calculation process, the Energy Consumption Deviation Symmetry Index decomposes the energy consumption deviations into two dimensions: absolute value and square value. The average absolute deviation reflects the overall level of deviation, while the square root of the mean square deviation emphasizes the contribution of dimensions with larger deviations. The absolute value of the difference between the two is used to characterize the distribution characteristics of the deviations in each dimension. If the difference is large, it means that the deviations are mainly concentrated in certain dimensions, indicating that the asymmetry of the energy consumption state between the dimensions is increasing, which may indicate a potential abnormal state. The Energy Consumption Deviation Symmetry Index quantifies the symmetry of multidimensional energy consumption data and provides a theoretical basis for identifying abnormal deviation characteristics.

[0100] The calculation process is as follows: Where Δx j (t i ) represents the jth dimension at time point t i The energy consumption deviation.

[0101] Energy consumption deviation symmetry index EDSI: in, Represents the average absolute value of the deviation of each dimension, Represents the root mean square of the deviation in each dimension. A larger absolute value of the energy consumption deviation symmetry index indicates a more significant asymmetry in the energy consumption deviation in each dimension, which may indicate an anomaly.

[0102] Calculate the instantaneous load energy consumption compression rate index ILECRI:

[0103] The instantaneous load energy consumption compression rate index is used to evaluate the concentration of real-time energy consumption data in the load distribution and reflect whether there is abnormal load concentration in the energy consumption state vector.

[0104] The instantaneous load energy consumption compression rate index is designed to analyze the concentration of real-time energy consumption data in the load distribution and reveal whether there is abnormal load aggregation in the energy consumption state vector. During the calculation, the real-time energy consumption data is first converted into a standard load value with relative weight significance through normalization to ensure that the data of each dimension is compared under a unified dimension. Subsequently, the instantaneous load energy consumption compression rate index quantifies the degree of load concentration by calculating the ratio of the extreme difference of the load value to the total load. At the same time, a logarithmic function is introduced to enhance the sensitivity to the ratio of load extreme values, thereby more accurately identifying abnormal characteristics caused by extreme load distribution. By quantifying the compression and concentration of the load distribution, this index provides a strong basis for detecting abnormal load behavior in the energy consumption state.

[0105] The calculation process is as follows: Among them, w jis the energy consumption weight of the j-th dimension, is the normalized energy consumption value.

[0106] Instantaneous load energy consumption compression index: in, and are the maximum and minimum normalized energy consumption values, respectively. The first term measures the relative difference in energy consumption, while the second enhances sensitivity to extreme load concentrations using a logarithmic function. Higher values ​​of the instantaneous load energy consumption compression index indicate a more uneven load distribution, potentially indicating abnormal energy consumption.

[0107] The energy consumption deviation symmetry index and the instantaneous load energy consumption compression rate index are weighted and summed to form a comprehensive comparison parameter CPI, which is used to comprehensively evaluate the abnormality of real-time energy consumption data.

[0108] Set the comprehensive comparison threshold θ CPI , when CPI(t i )>θ CPI (comprehensive comparison parameter is greater than the comprehensive comparison threshold), the corresponding time point t i There is an abnormal energy consumption. Record the information of abnormal data points, including the deviation value ΔX(t i ), corresponding to process step k and time point t i and save it in the abnormal data log

[0109] The anomaly detection module implements a multi-dimensional, complex comparison of real-time energy consumption data by introducing an energy consumption deviation symmetry index and an instantaneous load energy consumption compression ratio index. This process not only improves the accuracy and sensitivity of anomaly detection by quantifying deviation symmetry and load concentration, but also effectively integrates anomaly characteristics from different dimensions through comprehensive comparison parameters, ensuring the comprehensiveness and reliability of anomaly determination. Through precise anomaly identification and pattern tagging, the anomaly detection module provides high-quality data input for subsequent in-depth correlation analysis and intelligent anomaly handling, significantly enhancing the intelligence and responsiveness of the energy monitoring system.

[0110] Accurately pinpointing energy consumption anomalies is a key component of optimizing energy management and production efficiency. Identifying anomalies is only the first step; further investigation into potential causes of anomalies requires combining the operating status of production equipment with process parameters. Because energy consumption anomalies can be caused by a combination of multiple factors, multivariate correlation analysis techniques, through modeling the linkage between historical data and real-time parameters, can pinpoint the root cause of anomalies and provide a reliable basis for subsequent energy optimization.

[0111] The association analysis module includes the following:

[0112] Extract the deviation value, corresponding process link and time point from the abnormal data log generated by the anomaly detection module as the initial input for deep correlation analysis;

[0113] Combined with the equipment status dataset S k (t)={s1(t),s2(t),...,s m (t)} and process parameter data set P k (t)={p1(t),p2(t),...,p l (t)}, the abnormality occurs at time t i and its neighborhood time window [t i -Δt,t i +Δt] to conduct association modeling and preliminarily screen possible key influencing factors.

[0114] The equipment status dataset refers to a collection of multidimensional time series data collected during the operation of each process link to characterize the operating status of the equipment. This dataset covers the operating status parameters of key equipment, including but not limited to equipment vibration signals, bearing temperature, speed, operating current, lubricating oil pressure, hydraulic system pressure, etc. For example, in the steel rolling process, equipment status data may include the real-time speed of the rolls, the temperature rise curve of the bearings, and the fluctuation range of the rolling pressure; in the steelmaking process, the operating current of the electric furnace, the water flow of the cooling system, and the real-time temperature of the main transformer may be recorded in the equipment status dataset. This data is collected in real time by on-site sensors and indexed by timestamps to form time series data, which is used to describe the dynamic operating status of the equipment in a specific time period.

[0115] A process parameter dataset refers to a collection of control and environmental parameters directly related to the production process during the operation of each process link. This dataset reflects the real-time changes in the process's operational settings and environmental conditions, and typically includes production load, temperature setpoints, gas flow, oxygen ratio, steel specifications, raw material consumption, and so on. For example, in ironmaking, process parameters may include the temperature distribution within the blast furnace, coke charge, and blast pressure; in steel rolling, parameters may include rolling temperature settings, target steel plate thickness, and cooling water spray volume. Together with equipment status data, these data form the basis for a complete energy consumption analysis, used to establish correlations between energy consumption anomalies and process operations and environmental conditions.

[0116] A multivariate time series regression model is applied, with energy consumption deviation as the target variable and all variables in the equipment status data and process parameter data as input factors. By calculating the product of the partial derivative coefficient of each factor and the actual value, the contribution of each factor to the energy consumption deviation is quantified. The formula is as follows: in, Representation factor Cumulative contribution to energy consumption deviation, is the partial derivative coefficient in the regression model, is the factor value within the time window;

[0117] Sort the factors in descending order according to their impact contribution, and select the top N factors with the largest contribution values ​​as key factors to form the abnormal characteristic factor set. And record the corresponding characteristic deviation value and time period;

[0118] Based on the abnormal feature factor set, clustering algorithms (such as Gaussian mixture model) are used to classify abnormal events, classify abnormalities with similar patterns into the same category, and mark the classification results and root cause types to form an abnormal classification archive in Infer the result from the root cause;

[0119] Combined with anomaly classification archiving, the energy signature reference model is updated, the characteristic distribution of abnormal events is used as boundary condition correction, and the sensitivity of the reference model to similar events in the future is optimized to provide an enhanced comparison benchmark for subsequent processing.

[0120] The Correlation Analysis Module uses multidimensional correlation modeling and regression analysis of abnormal data, equipment status, and process parameters to accurately identify the key factors influencing energy consumption anomalies. It then uses clustering and classification techniques to categorize and archive anomaly patterns. This process not only provides a scientific basis for clearly identifying the root cause of anomalies, but also dynamically optimizes the energy signature reference model, laying the data foundation for intelligent scheduling and energy consumption optimization in the Scheduling Optimization Module. Through systematic multivariate analysis, the Correlation Analysis Module significantly improves the accuracy and decision-making efficiency of anomaly handling, further enhancing the level of intelligent energy management.

[0121] Promptly addressing energy consumption anomalies and adjusting processes are key to improving energy efficiency and production stability. After detecting anomalies and analyzing their root causes, intelligent scheduling and optimization algorithms are used to develop refined adjustment strategies tailored to the specific anomaly type. Predictive scheduling and online optimization enable dynamic adjustment of key process parameters and equipment status, helping to minimize the impact of anomalies on production and energy consumption while ensuring the stable operation of the overall energy system.

[0122] The scheduling optimization module includes the following:

[0123] Extract the anomaly type, key factor set, and root cause analysis results from the anomaly classification archive output by the correlation analysis module as input for scheduling and optimization;

[0124] Based on the predictive scheduling model, a time series prediction algorithm is used to make short-term predictions of the future status of key factors and construct a prediction function: in, Factor The predicted value at the future time t+Δt, and They are the rate of change and acceleration at the current moment, respectively, and the prediction is completed through a multivariate regression model combined with other relevant factors;

[0125] The prediction results are matched with the energy signature reference model, possible energy consumption deviations in the future are simulated, the dynamic response effects of different adjustment schemes are calculated, and the adjustment strategy with the shortest response time and the best recovery effect is selected to form the optimal adjustment scheme τ opt ={ParamAdjust,DeviceSwitch,ScheduleChange};

[0126] ParamAdjust corrects key process parameters that cause abnormal energy consumption to achieve energy recovery and production optimization. This includes adjusting process temperature setpoints, gas flow ratios, oxygen concentration settings, and pressure control parameters. For example, in blast furnace ironmaking, if excessive oxygen concentration causes energy consumption to deviate, normal conditions can be restored by lowering the oxygen concentration setpoint.

[0127] Device Switching involves switching the operating status of backup equipment or different devices based on the results of an anomaly root cause analysis to optimize resource utilization or minimize the impact of an anomaly. This includes switching backup fuel supply lines, activating backup pump stations, switching backup motors, or switching redundant equipment. For example, if the main gas pipeline is insufficient, a backup line can be activated to maintain system operation.

[0128] ScheduleChange adjusts the order of production line tasks or the load distribution to balance the load across nodes and avoid energy consumption fluctuations caused by the concentration of high-energy-consuming tasks. This includes rescheduling the execution of high-energy-consuming tasks, deferring the execution of low-priority tasks, and distributing high-energy-consuming tasks to different process nodes. For example, in the steel rolling process, lower-temperature steel plate rolling tasks can be prioritized, while high-temperature rolling tasks can be postponed to off-peak hours.

[0129] After determining the adjustment plan, the key process parameters are revised through the online optimization algorithm. The adjustment plan includes:

[0130] Update the key process temperature setting value T new , ensuring that the temperature parameters gradually approach the reference model: Among them, T current is the current temperature setting value, E is the energy consumption deviation objective function, and η is the optimization step size;

[0131] Switch the backup fuel supply line, adjust the supply priority according to the gas flow demand, and use the gas supply balance equation: Q in =Q primary +Q secondary ; Ensure ΔQ = Q demand -Q in →0; where Q in Represents the total input of gas or energy actually received by the system, usually the sum of multiple energy sources, to meet process demands.

[0132] Q primary Refers to the flow of the main energy supply source, such as the input flow of the main gas pipeline or the main power line, which is the main energy supply part of the system operation.

[0133] Q secondary Refers to the flow of backup or auxiliary energy supply sources, such as the input flow of backup gas pipelines or backup power lines, which are usually activated to supplement the energy supply when the main energy source is insufficient.

[0134] Q demand It indicates the actual energy or gas flow required in the process of production, which is the demand calculated based on the current production tasks and equipment operating status.

[0135] ΔQ represents the difference between the energy supply and demand of a system and is used to measure whether the current energy supply and demand are balanced. When ΔQ is close to zero, it means that the supply just meets the demand; if ΔQ is positive or negative, it indicates an energy surplus or shortage.

[0136] Adjust the production line operation sequence by optimizing the task allocation matrix Where [m ov ] represents the task arrangement of the oth process node in time period v, and dynamically adjusts the task priority and scheduling order based on the root cause analysis results of energy consumption anomalies and the real-time process load status. First, the current load level of each process node is calculated by combining the equipment status data and process parameter data. where c ov is the energy consumption coefficient of the corresponding task. Then, according to the load level and the key factors of the abnormal area, the task priority is reordered, the tasks with lower energy consumption and less impact on the overall load are scheduled first, and the execution of high energy consumption tasks is postponed. By introducing the load balancing objective function Φ=max(L o )-min(L o ), where Φ represents the load balancing index, max(L o ) indicates the maximum load, min(L o) represents the minimum load. A genetic algorithm or linear programming optimization method is used to iteratively adjust the task allocation matrix until the load balancing index Φ converges to a set threshold. Finally, the production schedule is updated based on the optimized task allocation plan, and the adjusted task sequence instructions are issued to each process node, achieving dynamic load balancing on the production line while reducing the sudden impact of high-energy-consuming tasks.

[0137] The feedback control mechanism is used to monitor the adjustment effect in real time, and the adjusted energy consumption data is dynamically compared with the energy signature reference model. If the deviation value gradually decreases and tends to the predetermined range, the adjustment effect is recorded and the abnormal state is marked as recovered.

[0138] Finally, the energy signature reference model is updated to incorporate changes in key parameters during the adjustment process into the model boundary range, thereby improving the model's adaptability to similar abnormal events in the future.

[0139] The scheduling optimization module combines a predictive scheduling model with an online optimization algorithm to dynamically adjust and precisely control energy consumption anomalies. This includes multi-dimensional operations such as revising key process parameters, switching to backup equipment, and optimizing task loads. Logically linking the anomaly classification and root cause analysis results of the correlation analysis module, the module verifies the effectiveness of adjustments through a feedback mechanism and optimizes the energy signature reference model, providing intelligent, closed-loop technical support for long-term energy management optimization. This process not only eliminates energy consumption anomalies immediately, but also ensures production stability and the efficient operation of the energy system.

[0140] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0141] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0142] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article or device. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0143] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. An energy collection and monitoring system based on big data, characterized in that: include: Data acquisition module, data processing module, feature modeling module, anomaly detection module, correlation analysis module and scheduling optimization module; Data acquisition module: Utilizes pre-placed sensors to collect energy-related parameters in real time and upload them to the big data processing platform. The output data set is then passed to the data processing module. Data processing module: This module converts the uploaded data into a unified format and removes noise signals to ensure data accuracy and consistency. It also indexes and stores the data by time point and process number, and outputs standardized data sets that are passed to the feature modeling module. Feature modeling module: Based on historical data and equipment operating characteristics, it extracts the multi-dimensional energy consumption characteristics of each process link, constructs an energy signature reference model as a benchmark for normal energy consumption status, and transmits the output energy signature reference model to the anomaly detection module; specifically: The trend and fluctuation characteristics are combined to form a multidimensional energy consumption feature vector. The multidimensional distribution of the multidimensional energy consumption feature vector is divided using a density clustering algorithm. The center point of the cluster with the highest density is extracted as the typical energy consumption state of each process link. The upper and lower limits of the characteristic range are calculated to form an energy signature reference model with boundary conditions, which serves as a normal energy consumption benchmark for real-time comparison. Anomaly Detection Module: Continuously receives real-time energy consumption data, comprehensively evaluates the deviation distribution of multi-dimensional energy consumption states and load concentration characteristics, dynamically compares them with the energy signature reference model, identifies abnormal data points that exceed the normal energy consumption range, and transmits the detected abnormal data to the correlation analysis module; The anomaly detection module includes the following: The typical energy consumption state center value and energy consumption characteristic boundary of each process link are extracted from the energy signature reference model as the comparison benchmark. The real-time energy consumption data set is loaded point by point into the comparison module. For the real-time energy consumption state vector at each time point, the energy consumption deviation symmetry index and the instantaneous load energy consumption compression rate index are calculated and processed. The energy consumption deviation symmetry index measures whether the deviations of energy consumption data in each dimension are evenly distributed by performing a symmetry analysis on the deviations of the real-time energy consumption state vector. Specifically, the energy consumption deviation symmetry index decomposes the energy consumption deviation into two dimensions: absolute value and square value. The average absolute deviation reflects the overall level of deviation, while the square root of the mean square deviation emphasizes the contribution of dimensions with large deviations. The absolute value of the difference between the two is used to characterize the distribution characteristics of the deviations in each dimension, namely the energy consumption deviation symmetry index. The instantaneous load energy consumption compression ratio index is designed to analyze the concentration of real-time energy consumption data in the load distribution and reveal whether there is any abnormal load concentration in the energy consumption state vector. During the calculation, the real-time energy consumption data is first converted into standard load values ​​with relative weights through normalization, ensuring that data of various dimensions are compared under a unified dimension. The instantaneous load energy consumption compression ratio index then quantifies the degree of load concentration by calculating the ratio of the load range to the total load. A logarithmic function is also introduced to enhance sensitivity to load extreme value ratios, thereby identifying abnormal characteristics caused by extreme load distribution. The energy consumption deviation symmetry index and the instantaneous load energy consumption compression rate index are weighted and summed to form a comprehensive comparison parameter, which is used to comprehensively evaluate the degree of abnormality of real-time energy consumption data; a comprehensive comparison threshold is set. When the comprehensive comparison parameter is greater than the comprehensive comparison threshold, it is determined that energy consumption abnormality exists at the corresponding time point; information about the abnormal data point is recorded and stored in the abnormal data log; Correlation Analysis Module: This module conducts in-depth correlation analysis on detected abnormal data, combines equipment status and production parameters, and uses multivariate machine learning algorithms to determine the root cause of the abnormality and classify and archive it. The output abnormality classification archive and root cause analysis results are then passed to the scheduling optimization module. Scheduling Optimization Module: Utilizes a predictive scheduling engine and online energy optimization algorithms to evaluate the effectiveness of adjustment plans. By adjusting key process temperatures, switching backup fuel supply lines, and issuing production line operation sequence adjustment instructions, it can immediately eliminate energy consumption anomalies and ensure production and furnace temperature balance.

2. The energy collection and monitoring system based on big data according to claim 1 is characterized in that: The data acquisition module includes the following: Using multiple types of sensors placed at key process nodes, multiple energy parameters are collected in real time at fixed sampling intervals to construct an energy state vector to represent the comprehensive energy consumption state at each sampling moment. Through a dynamic partition scheduling strategy, the data collected by multiple types of sensors are classified and stored in distributed data nodes, and the time point is used as the primary key to generate an original data set with a unique identifier; the cleaned original data set is marked according to the time point and process number to generate a structured data set.

3. The energy collection and monitoring system based on big data according to claim 2 is characterized in that: The data processing module includes the following: Extract multi-dimensional energy parameters from the data set uploaded by the data acquisition module, and preliminarily analyze the type of source sensor and data encoding format; Based on the rule-based dynamic mapping table, the data encoding formats uploaded by sensors from different sources are converted and an initial data set with a unified format is generated; The noise signal in the initial data set is removed using the wavelet transform method; An index matrix is ​​constructed by combining time points and process numbers, and the cleaned data is assigned to the time series storage unit of the corresponding process, ensuring that the data is continuous in the time dimension and independent in the process dimension; Generate a standardized dataset that has been format converted, denoised, and indexed.

4. The energy collection and monitoring system based on big data according to claim 3 is characterized in that: The feature modeling module includes the following: From the standardized data set, the data series is divided by process number to form an independent time series data set for each process link, so as to conduct group analysis on the energy consumption characteristics of different processes; For each process step, the time series data is decomposed into basic consumption trend and periodic fluctuation components by applying the adaptive weighted curve decomposition method. Based on the basic consumption trend, key inflection point characteristics are calculated, including the maximum change rate and the time position of the energy consumption inflection point, to capture the critical moments of energy demand; For periodic fluctuation components, short-time Fourier transform is used to extract the main frequency components, and the energy concentration is calculated to quantify the fluctuation intensity. The trend features and fluctuation features are combined to form a multi-dimensional energy consumption feature vector, and the high-dimensional features are embedded into a low-dimensional space through a dimensionality reduction algorithm to form a visual feature distribution map; The density clustering algorithm is used to perform density clustering analysis on the multidimensional energy consumption feature vectors in the feature distribution map. By setting the core point density threshold and the minimum number of samples, the dense areas of energy consumption characteristics are identified and divided into clusters with similar energy consumption patterns. The center point of the cluster with the highest density is extracted as the typical energy consumption state of each process link, and the upper and lower limits of the feature range are calculated to form an energy signature reference model with boundary conditions, which serves as the normal energy consumption benchmark for real-time comparison.

5. The energy collection and monitoring system based on big data according to claim 1 is characterized in that: The association analysis module includes the following: Extract deviation values, corresponding process links, and time points from abnormal data logs as initial input for deep correlation analysis; By combining the equipment status dataset and the process parameter dataset, we conduct correlation modeling on the abnormality occurrence time point and its adjacent time window, and preliminarily screen the key influencing factors. A multivariate time series regression model is applied, with energy consumption deviation as the target variable and all variables in equipment status data and process parameter data as input factors. By calculating the product of the partial derivative coefficient of each factor and the actual value, the contribution of each factor to energy consumption deviation is quantified. Sort the factors in descending order according to their impact contribution, and select the top The factors with the largest contribution value are taken as key factors to form the abnormal characteristic factor set, and their corresponding characteristic deviation values ​​and time periods are recorded; Based on the abnormal feature factor set, the clustering algorithm is used to classify abnormal events, classify abnormalities with similar patterns into the same category, and mark the classification results and root cause types to form an abnormal classification archive.

6. The energy collection and monitoring system based on big data according to claim 5 is characterized in that: The scheduling optimization module includes the following: Extract anomaly types, key factor sets, and root cause analysis results from anomaly classification archives as input for scheduling and optimization; Based on the predictive scheduling model, a time series prediction algorithm is used to make short-term predictions of the future status of key factors, a prediction function is constructed, and the prediction is completed by combining a multivariate regression model with other relevant factors; Match the prediction results with the energy signature reference model, simulate future energy consumption deviations, calculate the dynamic response effects of different adjustment schemes, and select the adjustment strategy with the shortest response time and the best recovery effect to form the optimal adjustment plan; After determining the adjustment plan, the key process parameters are revised through the online optimization algorithm. The adjustment plan includes: Update key process temperature setting values ​​to ensure that temperature parameters gradually approach the energy signature reference model; Switch backup fuel supply lines and adjust supply priorities based on gas flow demand; Adjusting the production line operation sequence is achieved by optimizing the task allocation matrix, dynamically adjusting the task priority and scheduling order based on the root cause analysis results of energy consumption anomalies and the real-time process load status; The adjusted energy consumption data is dynamically compared with the energy signature reference model. If the deviation value gradually decreases and tends to the predetermined range, the adjustment effect is recorded and the abnormal state is marked as recovered.

Images