An intelligent classification and collaborative processing data management system for oil sludge based on big data

Abstract

The present application relates to the technical field of data-driven industrial waste intelligent disposal and process control, in particular to an intelligent classification and collaborative processing data management system for oil sludge based on big data, which comprises a data acquisition module for acquiring multi-source heterogeneous data of the whole life cycle of oil sludge; a first processing module for constructing an oil sludge graph structure model and generating an oil sludge disposal potential vector; a second processing module for calculating a compatibility index and determining a compatibility evaluation result according to the oil sludge disposal potential vector and a preset ideal disposal feature vector; a third processing module for collecting process state time series data and predicting key performance indicators in response to the disposal process selected by the compatibility evaluation result; and a feedback control module for determining a comprehensive benefit deviation factor and outputting a hierarchical correction strategy. The present application overcomes the defects of traditional methods relying on artificial experience and subjective decision-making, significantly improving the scientificity and accuracy of disposal scheme recommendation.

Description

Technical Field

[0001] This invention relates to the field of data-driven intelligent disposal and process control technology for industrial waste, specifically a data management system for intelligent classification and collaborative processing of oil sludge based on big data. Background Technology

[0002] With the continuous development of industrial production, oil sludge, as a common hazardous waste, is crucial for environmental protection and sustainable development through its safe disposal and resource utilization. Oil sludge has diverse sources, complex composition, and fluctuating properties, which makes it difficult to treat and results in high disposal costs. How to scientifically classify oil sludge with different characteristics and match it with the most suitable treatment process, while carrying out refined control during the treatment process to maximize benefits, is a major challenge currently facing the industry.

[0003] Traditional methods of oil sludge disposal mainly rely on manual experience and offline laboratory analysis, which have the following limitations:

[0004] Due to the lack of in-depth and quantitative characterization methods for the intrinsic properties of oil sludge, the selection of treatment processes often relies on the experience of operators, which is highly subjective. This approach makes it difficult to ensure the scientific and optimal nature of decision-making, often leading to problems such as low resource recovery rate, high energy consumption, or secondary pollution.

[0005] Traditional techniques typically analyze the various physicochemical parameters of oil sludge as independent variables, ignoring the complex interactions and higher-order dependencies between the chemical components. This approach fails to reveal the true treatment potential of oil sludge and limits the refinement and personalization of subsequent treatment solutions.

[0006] In the process of handling, there is a general lack of effective dynamic and forward-looking monitoring methods; the control system is mostly passive, that is, it only makes adjustments after the key performance indicators are detected to deviate; this lagging control mode is difficult to cope with real-time changes in operating conditions, and cannot provide early warning and avoid potential risks, resulting in process instability and large fluctuations in benefits;

[0007] When a single batch of sludge is not fully suitable for a specific process, blending it with other batches of sludge to optimize feed characteristics is a potentially efficient disposal strategy. However, existing technologies lack data-driven collaborative evaluation and batching optimization methods, making it difficult to scientifically identify sludge batches with complementary characteristics and determine the optimal blending scheme from a large sludge inventory.

[0008] In summary, existing technologies have not yet formed a comprehensive intelligent management system covering the entire process from in-depth characterization of oil sludge properties, intelligent evaluation of treatment plans, dynamic prediction of production processes to closed-loop control of risk deviations. As a result, it is difficult to achieve a significant improvement in resource recovery rate, energy utilization efficiency and environmental safety in oil sludge treatment.

[0009] The information disclosed in the background section is provided only to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0010] To address the aforementioned technical problems, this invention discloses a data management system for intelligent classification and collaborative processing of oil sludge based on big data. Specifically, the technical solution of this invention is as follows:

[0011] A big data-based intelligent classification and collaborative processing data management system for oil sludge includes:

[0012] The data acquisition module is used to acquire multi-source heterogeneous data of the entire life cycle of oil sludge, including chemical parameters, physical parameters and history information;

[0013] The first processing module is used to construct a sludge map structure model and generate a sludge disposal potential vector based on multi-source heterogeneous data.

[0014] The second processing module is used to calculate the compatibility index and determine the compatibility assessment results based on the sludge treatment potential vector and the preset ideal treatment characteristic vector.

[0015] The third processing module is used to collect process status time series data and predict key performance indicators in response to the treatment process selected by the compatibility assessment results.

[0016] The feedback control module is used to determine the comprehensive benefit deviation factor and output a graded correction strategy based on the deviation between the predicted value and the preset target value of the key performance indicator.

[0017] Preferably, the first processing module generates a sludge disposal potential vector, including:

[0018] Chemical components in multi-source heterogeneous data are defined as nodes in a graph, and interactions between components are defined as edges, in order to establish a sludge graph structure model.

[0019] A graph neural network model is used to iteratively update the node features in the sludge diagram structure model to generate the final node feature vector.

[0020] Global average pooling is performed on the feature vectors of all final nodes to obtain the sludge disposal potential vector.

[0021] Preferably, the second processing module determines the compatibility assessment result, including:

[0022] Extract oil sludge samples from historically high-efficiency treatments and calculate the centroid of the oil sludge treatment potential vector to define the ideal treatment feature vector;

[0023] Cosine similarity is used to calculate the directional consistency between the sludge treatment potential vector and the ideal treatment feature vector of each treatment process to obtain the compatibility index.

[0024] Compare the compatibility index with the first recommended threshold and the second synergistic threshold;

[0025] When the compatibility index is greater than the first recommendation threshold, the compatibility evaluation result is determined to be the priority recommendation;

[0026] When the compatibility index is less than or equal to the first recommended threshold and greater than the second collaboration threshold, the compatibility evaluation result is determined to consider collaboration.

[0027] When the compatibility index is less than or equal to the second collaboration threshold, the compatibility assessment result is determined to be not recommended.

[0028] Preferably, the third processing module predicts key performance indicators, including:

[0029] Based on the selected processing technology according to the compatibility assessment results, the corresponding long short-term memory network model is invoked;

[0030] The sludge disposal potential vector is used as the initial hidden state of the long short-term memory network model;

[0031] Using process state time series data as sequence input, a long short-term memory network model is driven to predict future states and output key performance indicators.

[0032] Preferably, the feedback control module determines the comprehensive benefit deviation factor, including:

[0033] Obtain the predicted values ​​of key performance indicators, which include expected resource productivity, expected energy consumption index and environmental risk index;

[0034] The predicted values ​​of key performance indicators are compared with their respective preset target values ​​and normalized to obtain the deviation of each indicator;

[0035] The deviations of each indicator are weighted and summed according to the preset strategy weights to obtain the comprehensive benefit deviation factor.

[0036] Preferably, the feedback control module outputs a graded correction strategy, including:

[0037] Compare the comprehensive benefit deviation factor with the first correction threshold and the second correction threshold;

[0038] When the comprehensive benefit deviation factor is less than or equal to the first correction threshold, the strategy of maintaining the current treatment parameters is output.

[0039] When the comprehensive benefit deviation factor is greater than the first correction threshold and less than or equal to the second correction threshold, the first-level correction strategy is output.

[0040] When the comprehensive benefit deviation factor is greater than the second correction threshold, a secondary correction strategy is output.

[0041] Preferably, the first-level correction strategy is:

[0042] Fine-tuning of individual operating parameters in the treatment process, including reactor temperature or catalyst injection rate.

[0043] Preferably, the secondary correction strategy is:

[0044] Perform a multi-objective optimization search to find complementary sludge batches and mixing ratios in the sludge pool;

[0045] Based on complementary sludge batches and mixing ratios, a synergistic batching scheme is generated.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1. This invention constructs a graph structure model of the chemical components of oil sludge, deeply explores the high-order dependencies between the components, transforms high-dimensional heterogeneous data into feature vectors that can accurately characterize their treatment potential, achieves a deep characterization of oil sludge properties, and objectively evaluates its matching degree with different treatment processes through vector similarity calculation. This overcomes the shortcomings of traditional methods that rely on human experience and subjective decision-making, and significantly improves the scientificity and accuracy of treatment plan recommendations.

[0048] 2. This invention uses the disposal potential vector, which characterizes the intrinsic properties of oil sludge, as the initial information input for the process prediction model, providing the model with high-value prior knowledge. Combined with real-time process data, it realizes dynamic and forward-looking prediction of key performance indicators such as resource yield and energy consumption index, transforming traditional lagging passive control into proactive predictive management, effectively avoiding potential process risks, and improving the stability and efficiency of the disposal process.

[0049] 3. This invention integrates multiple predictive performance indicators with different dimensions into a single deviation factor that comprehensively reflects overall benefits through weighted fusion. This factor serves as a unified decision-making basis, driving a tiered correction strategy and ensuring that the cost and intensity of intervention measures match the degree of deviation from the prediction, thereby achieving precise and economical control decisions.

[0050] 4. This invention constructs a closed-loop control system from single-parameter fine-tuning to cross-batch collaborative batching. In particular, when the prediction deviation is large, the system can automatically initiate a multi-objective optimization search to find oil sludge with complementary characteristics in the sludge pool and generate the optimal collaborative batching scheme, optimizing the feed characteristics to solve disposal problems, and providing strong data support for realizing the refined and resource-oriented collaborative processing of oil sludge. Attached Figure Description

[0051] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0052] Figure 1 This is a flowchart of a data management system for intelligent classification and collaborative processing of oil sludge based on big data, according to the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. Example 1:

[0054] Please see Figure 1 A data management system for intelligent classification and collaborative processing of oil sludge based on big data, comprising:

[0055] The data acquisition module is used to acquire multi-source heterogeneous data of the entire life cycle of oil sludge, including chemical parameters, physical parameters and history information;

[0056] The first processing module is used to construct a sludge map structure model and generate a sludge disposal potential vector based on multi-source heterogeneous data.

[0057] The second processing module is used to calculate the compatibility index and determine the compatibility assessment results based on the sludge treatment potential vector and the preset ideal treatment characteristic vector.

[0058] The third processing module is used to collect process status time series data and predict key performance indicators in response to the treatment process selected by the compatibility assessment results.

[0059] The feedback control module is used to determine the comprehensive benefit deviation factor and output a graded correction strategy based on the deviation between the predicted value and the preset target value of the key performance indicator.

[0060] This embodiment provides a data management system for intelligent classification and collaborative processing of oily sludge based on big data. The system aims to solve the problems of strong subjectivity in the selection of oily sludge disposal methods, lagging process control, and lack of data support for collaborative processing in the existing technology. The system constructs an integrated technical closed loop covering data acquisition, characteristic characterization, scheme evaluation, process prediction to closed-loop control.

[0061] The system includes: a data acquisition module, a first processing module, a second processing module, a third processing module, and a feedback control module;

[0062] The data acquisition module aims to comprehensively and multidimensionally acquire basic data reflecting the characteristics of oil sludge throughout its entire lifecycle. In this embodiment, this module is implemented through an integrated data interface and sensor network. Specifically, it collects multi-source heterogeneous data of oil sludge throughout its generation, storage, and processing lifecycle. Multi-source heterogeneous data is a data set with diverse sources and structures, serving to provide comprehensive information input for subsequent deep modeling. In this embodiment, multi-source heterogeneous data specifically includes three subcategories:

[0063] Chemical parameters: obtained by laboratory analytical equipment such as gas chromatography-mass spectrometry, including but not limited to hydrocarbon components and heavy metal content, to reveal the material composition of oil sludge;

[0064] Physical parameters: These are monitored in real time by physical sensors deployed on storage tanks or pipelines, including viscosity, density, water content, and solids content, to characterize the macroscopic physical state of the sludge.

[0065] History information: Records are entered through the production execution system or manually, including the source of the sludge, its formation process, storage duration, etc., to trace the formation history and environmental background of the sludge;

[0066] The first processing module aims to transform high-dimensional, heterogeneous raw data into a low-dimensional, dense mathematical expression that can deeply characterize the intrinsic chemical properties and treatment potential of oil sludge. In this embodiment, this module constructs an oil sludge diagram structure model based on the collected multi-source heterogeneous data, and ultimately generates an oil sludge treatment potential vector through this model. The oil sludge treatment potential vector is defined as a real-number vector with a fixed dimension. Its position and orientation in vector space contain the most suitable disposal method and potential resource value of the sludge sample. It is the core information carrier connecting the original sludge data with all subsequent decision-making, assessment and control steps.

[0067] The second processing module aims to quantitatively and objectively assess the compatibility of sludge with different treatment processes based on its inherent characteristics. In this embodiment, this module calculates a compatibility index and ultimately determines a graded compatibility assessment result based on the sludge treatment potential vector generated by the first processing module and a preset ideal treatment feature vector. The ideal treatment feature vector is defined as the feature vector representing the most ideal treatment object for a certain treatment process. The data is derived from statistical analysis and centroid calculation of sludge samples from historical databases that have used this process and achieved high efficiency.

[0068] The third processing module aims to achieve dynamic and forward-looking monitoring of the disposal process, thereby transforming from passive response to proactive prediction. In this embodiment, in response to the disposal process selected by the compatibility assessment results determined by the second processing module, this module collects real-time process status time series data through process sensors, and then predicts the future trend and key performance indicators of the disposal process based on these data. Key performance indicators refer to the core quantitative indicators used to measure the effectiveness of the disposal process, such as expected resource yield, expected energy consumption index, and environmental risk index.

[0069] The feedback control module aims to transform predicted future risks or deviations into specific, executable control instructions, thereby achieving closed-loop optimization of the system. In this embodiment, the module determines a comprehensive benefit deviation factor based on the deviation between the predicted values ​​of key performance indicators output by the third processing module and the preset target values, and outputs a graded correction strategy according to the severity of this factor. The comprehensive benefit deviation factor is defined as a quantitative indicator. Its value directly reflects the degree to which the current handling process deviates from the overall optimal goal, and is the core decision-making basis for driving subsequent graded correction strategies;

[0070] In addition, to ensure the stability and reliability of the system in practical applications, it is necessary to conduct comprehensive robustness tests on the established graph neural network model and long short-term memory network model, including stress tests using outlier and boundary value data to verify the model’s performance when handling atypical working condition data, and to set up corresponding early warning or anomaly handling mechanisms.

[0071] This embodiment constructs a complete data-driven intelligent management closed loop for oil sludge by organically combining the above modules. It overcomes the shortcomings of traditional disposal methods that rely on human experience and subjective decision-making, and realizes full-process intelligentization from in-depth characterization of oil sludge characteristics, intelligent recommendation of disposal solutions, dynamic prediction of production process to closed-loop control of risk deviation. It significantly improves the resource recovery rate, energy utilization efficiency and environmental safety of oil sludge disposal, and provides strong technical support for the scientific and refined management of oil sludge. Example 2:

[0072] The first processing module generates a sludge disposal potential vector, including:

[0073] Chemical components in multi-source heterogeneous data are defined as nodes in a graph, and interactions between components are defined as edges, in order to establish a sludge graph structure model.

[0074] A graph neural network model is used to iteratively update the node features in the sludge diagram structure model to generate the final node feature vector.

[0075] Global average pooling is performed on the feature vectors of all final nodes to obtain the sludge disposal potential vector.

[0076] This embodiment further explains the specific implementation of the first processing module generating the sludge disposal potential vector based on Embodiment 1. One technical contribution of this embodiment is that it does not treat the various parameters of the sludge as mutually independent variables, but innovatively uses a graph neural network model to deeply explore the intrinsic interactions and higher-order dependencies between various chemical components.

[0077] Chemical components in multi-source heterogeneous data are defined as nodes in a graph, and interactions between components are defined as edges, to establish a sludge graph structure model. The sludge graph structure model refers to a technique for abstracting sludge samples into a graph data structure, which transforms unstructured chemical relationships into structured inputs that can be processed by machine learning models. In this embodiment, each node of the graph... Representing a chemical component, the initial characteristics assigned to the node. For example, it could be the concentration value of the component, which comes from the data acquisition module; the edges of the graph represent the interactions between components, and the correlation weights serve as the characteristics of the edges. These weights can be set based on the chemical reaction kinetics knowledge base or determined by analyzing the statistical correlations in historical data.

[0078] A graph neural network model is used to iteratively update the node features in the sludge graph structure model, generating the final node feature vector. To accurately simulate the information transmission between nodes in the graph and effectively learn higher-order dependencies, this embodiment specifically uses a graph convolutional network to update the node state. The aim is to generate feature expressions that better reflect the intrinsic chemical nature of the sludge, rather than a simple linear combination. The node state update equation is as follows:

[0079] in: :No. Nodes in a layered network The eigenvectors of , with initial values The source is multi-source heterogeneous data;

[0080] :node The set of neighboring nodes;

[0081] The number of neighboring nodes, used as a normalization factor;

[0082] Weight matrix and These are the core adjustable parameters of the model; among them, Used to perform a linear transformation on the aggregation of neighbor node information, and Used to transform the information of the node itself; in this embodiment, its initial value is set using the Xavier initialization method, and during the model training phase, historical sludge disposal data is used, the input is sludge map data, and the label is a quantitative indicator representing the actual disposal effect, such as resource recovery rate. Through the backpropagation algorithm and Adam optimizer, with the goal of minimizing the mean square error loss function between the predicted disposal effect and the actual disposal effect, these weights are iteratively optimized.

[0083] Non-linear activation functions, such as the ReLU function;

[0084] Calculation logic and effect: This formula is executed iteratively. Second-rate( (where is the number of network layers). In each iteration, the feature vector of a component is fused with the feature information of its neighboring components; after... Layer propagation, the final output is the feature vector of the top-level node. It contains rich high-order coupling information, which goes beyond the original concentration or physical property parameters;

[0085] Global average pooling is performed on the feature vectors of all final nodes to obtain the sludge disposal potential vector; the final feature representation of each node in the graph is then obtained after refinement. Subsequently, a global feature extraction operation is needed to aggregate the information from all nodes into a graph-level vector of fixed dimensions. This embodiment uses global average pooling as the implementation method for this operation; its calculation method is as follows:

[0086] in: : The final generated oil sludge disposal potential vector;

[0087] The set of all nodes in the graph;

[0088] Total number of nodes;

[0089] : The nodes output by the iterative update step of the preorder graph neural network. The final eigenvector.

[0090] Example 3:

[0091] The second processing module determines the compatibility assessment results, including:

[0092] Extract oil sludge samples from historically high-efficiency treatments and calculate the centroid of the oil sludge treatment potential vector to define the ideal treatment feature vector;

[0093] Cosine similarity is used to calculate the directional consistency between the sludge treatment potential vector and the ideal treatment feature vector of each treatment process to obtain the compatibility index.

[0094] Compare the compatibility index with the first recommended threshold and the second synergistic threshold;

[0095] When the compatibility index is greater than the first recommendation threshold, the compatibility evaluation result is determined to be the priority recommendation;

[0096] When the compatibility index is less than or equal to the first recommended threshold and greater than the second collaboration threshold, the compatibility evaluation result is determined to consider collaboration.

[0097] When the compatibility index is less than or equal to the second collaboration threshold, the compatibility assessment result is determined to be not recommended.

[0098] This embodiment, based on Embodiment 1, further explains the specific implementation method of the second processing module determining the compatibility assessment results; this embodiment aims to establish a standardized and quantifiable assessment process to overcome the subjectivity and uncertainty brought about by the reliance on human experience in the selection of traditional disposal solutions;

[0099] To establish an evaluation benchmark, samples of historically high-efficiency oil sludge treatment were extracted, and the centroid of the oil sludge treatment potential vector was calculated to define the ideal treatment characteristic vector. The settings are based on sufficient objective evidence and are entirely data-driven. The specific method is as follows: from the historical database, all sludge samples that have achieved high efficiency using specific treatment processes are selected. Here, high efficiency is defined as quantifiable, such as a resource recovery rate greater than 90% and unit energy consumption lower than the industry benchmark by 10%. Then, using the graph neural network model of the first processing module, the treatment potential vectors of these selected samples are calculated. The centroids of these vectors are calculated, and the result is set as the ideal treatment feature vector for this process. ;

[0100] Cosine similarity is used to calculate the directional consistency between the sludge treatment potential vector and the ideal treatment feature vector of each treatment process, thus obtaining a compatibility index. The technical reason for using this calculation method is that whether sludge characteristics match the process depends on whether the directions of their inherent feature combinations are consistent, rather than the absolute magnitude of the feature values. Cosine similarity is specifically used to measure the directional consistency between vectors, making it an ideal choice for evaluating the degree of matching. The compatibility index is calculated as follows:

[0101] in: : Sludge to be evaluated and the first The compatibility index of this treatment process is a value ranging from... dimensional scalar;

[0102] : The disposal potential vector of the sludge to be evaluated, which is derived from the calculation output of the first treatment module;

[0103] :No. The ideal processing feature vector for this process is generated by the initialization steps described above in this section;

[0104] Based on the calculated compatibility index, a tiered assessment is conducted; the compatibility index is compared with the first recommendation threshold and the second synergy threshold; to ensure the scientific and reliable nature of the tiered decision, the specific values ​​of the first recommendation threshold and the second synergy threshold, for example, can be set to 0.9 and 0.7, are the optimal cut-off points determined by analyzing receiver operating characteristic (ROC) curves of a large amount of historical treatment data, under the principle of maximizing recommendation accuracy while constraining the error recommendation rate;

[0105] When the compatibility index is greater than the first recommendation threshold, that is In this case, the compatibility assessment results should be given priority.

[0106] When the compatibility index is less than or equal to the first recommendation threshold and greater than the second collaboration threshold, i.e. At that time, the compatibility assessment results were determined to be in consideration of collaboration;

[0107] When the compatibility index is less than or equal to the second collaboration threshold, i.e. At that time, the compatibility assessment result was determined to be "not recommended".

[0108] Example 4:

[0109] The third processing module predicts key performance indicators, including:

[0110] Based on the selected processing technology according to the compatibility assessment results, the corresponding long short-term memory network model is invoked;

[0111] The sludge disposal potential vector is used as the initial hidden state of the long short-term memory network model;

[0112] Using process state time series data as sequence input, a long short-term memory network model is driven to predict future states and output key performance indicators.

[0113] This embodiment further explains the specific implementation method of the third processing module predicting key performance indicators based on embodiment 1; its core technology lies in using long short-term memory networks to model and predict the typical time series scenario of the disposal process.

[0114] To improve the targeting and accuracy of predictions, the system pre-trains and stores a dedicated long short-term memory network model for each specific treatment process. The long short-term memory network model is a recurrent neural network suitable for processing and predicting time series events. When the second processing module outputs the compatibility evaluation results, the system will automatically call the pre-trained long short-term memory network model corresponding to the selected treatment process.

[0115] Using the sludge disposal potential vector as the initial hidden state of a long short-term memory network model is a key technical feature of this invention; its technical value lies in providing the prediction model with high-value prior knowledge about the intrinsic properties of the material to be treated; specifically, the initial hidden state of the model... The vector set as the disposal potential of the sludge to be treated ,Right now Therefore, the model's prediction process is no longer without prior information, but incorporates a deep understanding of the chemical nature and potential behavioral tendencies of the batch of sludge from the initial moment, thus enabling more accurate long-term predictions.

[0116] Based on the above settings, process state time series data is used as sequence input to drive a long short-term memory network model to predict future states and output key performance indicators. After the treatment process begins, the data acquisition module collects process parameters in real time, forming a process state time series dataset. This sequence data is used as the sequence input to a Long Short-Term Memory (LSTM) network model, driving the model to predict future key performance indicators; for example:

[0117] Expected resource yield:

[0118] :arrive The expected total resource yield at any given moment;

[0119] : A trained long short-term memory network model for predicting yield for a specific process;

[0120] The source is the initial sludge potential vector output by the first processing module;

[0121] The source is real-time monitoring data from time 1 to... The process state data sequence;

[0122] Expected energy consumption index:

[0123] :arrive The comprehensive energy consumption index of expected unit resource output at any given time;

[0124] Long Short-Term Memory Network Model for Predicting Energy Consumption.

[0125] Example 5:

[0126] The feedback control module determines the overall benefit deviation factor, including:

[0127] Obtain the predicted values ​​of key performance indicators, which include expected resource productivity, expected energy consumption index and environmental risk index;

[0128] The predicted values ​​of key performance indicators are compared with their respective preset target values ​​and normalized to obtain the deviation of each indicator;

[0129] The deviations of each indicator are weighted and summed according to the preset strategy weights to obtain the comprehensive benefit deviation factor;

[0130] This embodiment, based on Embodiment 1, further explains the specific implementation method of the feedback control module determining the comprehensive benefit deviation factor; it aims to integrate multiple performance indicators with different dimensions into a single comprehensive evaluation factor that can directly guide control decisions.

[0131] The predicted values ​​of key performance indicators (KPIs) are obtained, including expected resource productivity, expected energy consumption index, and environmental risk index. These predicted values ​​are output in real time by the long short-term memory network model of the third processing module. Among them, the environmental risk index It is itself a comprehensive indicator, calculated by weighting and summing the predicted concentrations of multiple pollutants. Its purpose is to transform multi-dimensional pollutant emission data into a single, standardized risk measure. The calculation formula is as follows:

[0132] : The comprehensive environmental risk index at any given moment;

[0133] Number of pollutant types monitored;

[0134] : is the output of a long short-term memory network model specifically designed for predicting pollutant concentrations. Pollutants in Real-time predicted concentration at any given moment;

[0135] The statutory emission standard limit for this pollutant is derived from national or industry environmental protection regulations;

[0136] :No. The risk weight of a pollutant is determined based on a comprehensive assessment of the substance's toxicological data and environmental regulations, and must meet certain conditions. ;

[0137] The predicted values ​​of key performance indicators are compared with their respective preset target values ​​and normalized to obtain the deviation of each indicator; preset target values This refers to the performance benchmark set based on the production task book or historical best batch data;

[0138] The deviations of each indicator are weighted and summed according to the preset strategy weights to obtain the comprehensive benefit deviation factor; strategy weights It can be dynamically set by production managers based on the current production strategy, and meets the following requirements. The calculation method for the comprehensive benefit deviation factor is as follows: Structure and Logical Derivation:

[0139] The comprehensive benefit deviation factor at any given time is a dimensionless value;

[0140] The source is the prediction output of the third processing module;

[0141] :A baseline value set based on the production task sheet or the best historical batch;

[0142] : These are the weights set by production managers based on the current production strategy;

[0143] The settings ensure that deviation penalties are only applied when productivity falls below the target, while exceeding the target results in a negative penalty. It will not produce a negative bias to offset other deficiencies; for the energy consumption index and environmental risk index For these two cost indicators, the deviation term is directly expressed as a ratio. When the predicted value exceeds the target value, this ratio is greater than 1, directly constituting a penalty term. This design ensures the overall deviation factor... Logical consistency as a unified cost function.

[0144] Example 6:

[0145] The feedback control module outputs a tiered correction strategy, including:

[0146] Compare the comprehensive benefit deviation factor with the first correction threshold and the second correction threshold;

[0147] When the comprehensive benefit deviation factor is less than or equal to the first correction threshold, the strategy of maintaining the current treatment parameters is output.

[0148] When the comprehensive benefit deviation factor is greater than the first correction threshold and less than or equal to the second correction threshold, the first-level correction strategy is output.

[0149] When the comprehensive benefit deviation factor is greater than the second correction threshold, a secondary correction strategy is output.

[0150] The first-level correction strategy is:

[0151] Fine-tuning of individual operating parameters in the treatment process, including reactor temperature or catalyst injection rate;

[0152] The secondary correction strategy is as follows:

[0153] Perform a multi-objective optimization search to find complementary sludge batches and mixing ratios in the sludge pool;

[0154] Based on complementary sludge batches and mixing ratios, a synergistic batching scheme is generated;

[0155] Based on Example 1 and following Example 5, this embodiment provides a detailed explanation of the output hierarchical correction strategy in the feedback control module and its specific first-level and second-level correction strategies. Its core idea is to adopt different cost and intensity intervention measures for different degrees of deviation based on the principle of matching risk level with intervention intensity.

[0156] To initiate tiered correction, the comprehensive benefit deviation factor is compared with the first correction threshold and the second correction threshold. To ensure the economy and effectiveness of the correction strategy, the first correction threshold and the second correction threshold are set, for example, 0.05 and 0.2, based on statistical correlation analysis of the deviation factor values ​​in historical data and the resulting final economic or safety losses.

[0157] When the comprehensive benefit deviation factor is less than or equal to the first correction threshold, i.e. When necessary, output a strategy to maintain the current processing parameters;

[0158] When the comprehensive benefit deviation factor is greater than the first correction threshold and less than or equal to the second correction threshold, that is... When this happens, output the first-level correction strategy;

[0159] When the comprehensive benefit deviation factor is greater than the second correction threshold, that is When this happens, output the secondary correction strategy;

[0160] The primary correction strategy involves fine-tuning a single operating parameter in the treatment process. This single operating parameter is preferably one that directly affects the state of the treatment process, responds quickly, and has low adjustment costs. The aim of this strategy is to quickly suppress the expansion of deviations at minimal cost. In this embodiment, the single operating parameter includes the reactor temperature or the catalyst injection rate. For example, the system can automatically generate instructions to increase or decrease the reactor temperature by 2%-5%, or adjust the catalyst injection rate by 5%-10%.

[0161] The secondary correction strategy involves performing a multi-objective optimization search to find complementary sludge batches and mixing ratios in the sludge pool, and generating a synergistic batching scheme based on this. Unlike the primary correction strategy, the core idea of ​​the secondary correction strategy is to shift from adjusting process parameters to optimizing raw material treatment, aiming to solve the problem.

[0162] Perform a multi-objective optimization search: The system initiates a multi-objective optimization algorithm with the objective function... as follows:

[0163] in: : Vector of disposal potential for complementary sludge batches searched in the sludge pool;

[0164] The mixing ratio to be optimized;

[0165] A first-order linear approximation model of the theoretical disposal potential vector of mixed sludge, wherein... This represents the disposal potential vector of the current batch of sludge. It should be noted that this linear approximation aims to provide a computationally efficient representation for multi-objective optimization search. In future system iterations, a nonlinear mixing model based on chemical mechanisms can be introduced to further improve the accuracy of the synergistic batching scheme.

[0166] Expected compatibility of the mixed sludge with the target process;

[0167] It utilizes the graph neural network model of the first processing module and the long short-term memory network model of the third processing module to process mixed sludge. Feedforward prediction of the expected comprehensive benefit deviation factor for the disposal;

[0168] : Weighting coefficients used to balance compatibility and expected correction effects;

[0169] Generating a co-feeding scheme: The above optimization process aims to find a suitable batch of sludge. and mixing ratio The optimal solution minimizes the deviation of the estimated future comprehensive benefits of the mixture while ensuring basic compatibility; based on this optimal solution, the system will generate a clear collaborative batching scheme.

Claims

1. A data management system for intelligent classification and collaborative processing of oil sludge based on big data, characterized in that, include: The data acquisition module is used to acquire multi-source heterogeneous data of the entire life cycle of oil sludge, including chemical parameters, physical parameters and history information; The first processing module is used to construct a sludge map structure model and generate a sludge disposal potential vector based on multi-source heterogeneous data. The second processing module is used to calculate the compatibility index and determine the compatibility assessment results based on the sludge treatment potential vector and the preset ideal treatment characteristic vector. The third processing module is used to collect process status time series data and predict key performance indicators in response to the treatment process selected by the compatibility assessment results. The feedback control module is used to determine the comprehensive benefit deviation factor and output a graded correction strategy based on the deviation between the predicted value and the preset target value of the key performance indicator. The feedback control module determines the overall benefit deviation factor, including: Obtain the predicted values ​​of key performance indicators, which include expected resource productivity, expected energy consumption index and environmental risk index; The predicted values ​​of key performance indicators are compared with their respective preset target values ​​and normalized to obtain the deviation of each indicator; The deviations of each indicator are weighted and summed according to the preset strategy weights to obtain the comprehensive benefit deviation factor; The calculation method for the comprehensive benefit deviation factor is as follows: Structure and Logical Derivation: The comprehensive benefit deviation factor at any given time is a dimensionless value; The source is the prediction output of the third processing module; The baseline value is set based on the production task sheet or the best historical batch. These are weights set by production managers based on the current production strategy. The feedback control module outputs a graded correction strategy, including: comparing the comprehensive benefit deviation factor with a first correction threshold and a second correction threshold; When the comprehensive benefit deviation factor is less than or equal to the first correction threshold, the strategy of maintaining the current treatment parameters is output. When the comprehensive benefit deviation factor is greater than the first correction threshold and less than or equal to the second correction threshold, the first-level correction strategy is output. When the comprehensive benefit deviation factor is greater than the second correction threshold, a secondary correction strategy is output. The secondary correction strategy is as follows: Perform a multi-objective optimization search to find complementary sludge batches and mixing ratios in the sludge pool; Based on complementary sludge batches and mixing ratios, a synergistic batching scheme is generated; Its objective function as follows: in: Disposal potential vector of complementary sludge batches searched in the sludge pool; The mixing ratio to be optimized; A first-order linear approximation model of the theoretical treatment potential vector of mixed sludge, wherein... This represents the disposal potential vector for the current batch of oil sludge; Expected compatibility of the mixed sludge with the target process; It utilizes the graph neural network model of the first processing module and the long short-term memory network model of the third processing module to process mixed sludge. Feedforward prediction of the expected comprehensive benefit deviation factor for the disposal; Weighting coefficients used to balance compatibility and expected correction effects.

2. The data management system for intelligent classification and collaborative processing of oil sludge based on big data as described in claim 1, characterized in that, The first processing module generates a sludge disposal potential vector, including: Chemical components in multi-source heterogeneous data are defined as nodes in a graph, and interactions between components are defined as edges, in order to establish a sludge graph structure model. A graph neural network model is used to iteratively update the node features in the sludge diagram structure model to generate the final node feature vector. Global average pooling is performed on the feature vectors of all final nodes to obtain the sludge disposal potential vector.

3. The data management system for intelligent classification and collaborative processing of oil sludge based on big data as described in claim 1, characterized in that, The second processing module determines the compatibility assessment results, including: Extract oil sludge samples from historically high-efficiency treatments and calculate the centroid of the oil sludge treatment potential vector to define the ideal treatment feature vector; Cosine similarity is used to calculate the directional consistency between the sludge treatment potential vector and the ideal treatment feature vector of each treatment process to obtain the compatibility index. Compare the compatibility index with the first recommended threshold and the second synergistic threshold; When the compatibility index is greater than the first recommendation threshold, the compatibility evaluation result is determined to be the priority recommendation; When the compatibility index is less than or equal to the first recommended threshold and greater than the second collaboration threshold, the compatibility evaluation result is determined to consider collaboration. When the compatibility index is less than or equal to the second collaboration threshold, the compatibility assessment result is determined to be not recommended.

4. The data management system for intelligent classification and collaborative processing of oil sludge based on big data as described in claim 1, characterized in that, The third processing module predicts key performance indicators, including: Based on the selected processing technology according to the compatibility assessment results, the corresponding long short-term memory network model is invoked; The sludge disposal potential vector is used as the initial hidden state of the long short-term memory network model; Using process state time series data as sequence input, a long short-term memory network model is driven to predict future states and output key performance indicators.

5. A data management system for intelligent classification and collaborative processing of oil sludge based on big data as described in claim 4, characterized in that, The first-level correction strategy is as follows: Fine-tuning of individual operating parameters in the treatment process, including reactor temperature or catalyst injection rate.

Images