Coal preparation plant scheduling method and system based on multi-time-series data analysis

Abstract

The invention provides a coal preparation plant scheduling method and system based on multi-time-series data analysis. The method comprises the steps of collecting real-time production data of a coal preparation plant based on Internet of Things equipment; generating a target washing strategy based on the real-time production data and a pre-constructed polymorphic uncertainty full-time sequence simulation model; wherein the polymorphic uncertainty full-time-sequence simulation model predicts real-time production data based on a dynamic knowledge graph of time sequence perception and a time sequence prediction model, and the dynamic knowledge graph of time sequence perception is obtained by performing clustering analysis on historical production data through a dynamic knowledge graph algorithm of time sequence perception; and distributing the target washing strategy to the corresponding production equipment. According to the method, the dependence on artificial experience is reduced, and the scientificity of decision making is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent management of coal preparation plants, and in particular to a coal preparation plant scheduling method and system based on multi-time series data analysis. BACKGROUND

[0002] Traditional coal preparation plant scheduling mainly relies on manual experience and scattered automation systems, including the following modes: (1) Manual experience dominant scheduling: production reports, equipment status and other information are manually summarized, and production plans are formulated based on experience.

[0003] (2) Local automation system: some enterprises use centralized control systems (such as PLC control) or video monitoring systems to monitor the running state of the equipment, but the data islands between systems are serious, and there is a lack of global coordination capability.

[0004] (3) Static task allocation: existing scheduling methods are mostly based on fixed priority or rules, lacking dynamic response to historical data patterns and future trends. SUMMARY

[0005] Therefore, the present application aims to provide a coal preparation plant scheduling method and system based on multi-time series data analysis to reduce the dependence on manual experience and improve the scientific nature of decision-making.

[0006] To achieve the above-mentioned purpose, the technical solutions adopted by the present application are as follows: In a first aspect, the present application provides a coal preparation plant scheduling method based on multi-time series data analysis, comprising: collecting real-time production data of the coal preparation plant based on Internet of Things devices; generating a target washing strategy based on real-time production data and a pre-constructed multi-state uncertainty full-time series simulation model; wherein the multi-state uncertainty full-time series simulation model predicts real-time production data based on a time series perception dynamic knowledge graph and a time series prediction model, and the time series perception dynamic knowledge graph is obtained by clustering analysis of historical production data through a time series perception dynamic knowledge graph algorithm; and distributing the target washing strategy to the corresponding production equipment.

[0007] Optionally, generating a target washing strategy based on real-time production data and a pre-constructed multi-state uncertainty full-time series simulation model comprises: predicting and simulating real-time production data based on a time series perception dynamic knowledge graph and a time series prediction model to obtain a prediction result of the equipment health state and production data; and generating a target washing strategy based on the prediction result of the equipment health state and production data.

[0008] Optionally, real-time production data is predicted and simulated based on a time-series-aware dynamic knowledge graph and a time-series prediction model to obtain prediction results of equipment health status and production data, including: data cleaning of real-time production data to obtain a real-time time-series production data set; wherein the real-time time-series production data set includes uncertain coal preparation parameters; feature extraction of the real-time time-series production data set to obtain characteristic values ​​of the real-time production data; prediction based on the time-series-aware dynamic knowledge graph and the characteristic values ​​of the real-time production data to obtain equipment health status; prediction based on the time-series prediction model and the real-time time-series production data set to obtain the predicted value and probability distribution of the production data; wherein, a Monte Carlo simulation engine is used to generate the probability distribution of uncertain coal preparation parameters.

[0009] Optionally, generating a target cleaning strategy based on the predicted results of the equipment health status and production data includes: generating multiple cleaning strategies based on the predicted values ​​and probability distribution of the equipment health status and production data; and determining the target cleaning strategy from the multiple cleaning strategies based on a preset multi-objective optimization function; wherein the multi-objective optimization function is:

[0010] in, MTBF is the average trouble-free working time of production equipment, α 、 β 、 gamma is the weight coefficient.

[0011] Optionally, after distributing the target washing strategy to the corresponding production equipment, it also includes: obtaining the actual production results after the target washing strategy is distributed, and feeding back the actual production results to the historical production database; based on the actual production results and the target washing strategy, optimizing the time-aware dynamic knowledge graph.

[0012] Optionally, before collecting real-time production data of a coal preparation plant based on IoT devices, the method also includes: obtaining historical production data, and performing data cleaning on the historical production data to obtain historical time series production data; performing feature extraction of entity quadruple on the historical time series production data based on a time series-aware encoder to obtain an entity embedding vector; wherein, the time series-aware encoder performs feature extraction of entity quadruple based on an improved convolutional neural network, and the improved convolutional neural network performs feature extraction of entity quadruple in spatial and time dimensions based on the time sequence of the historical time series production data, and the entity quadruple includes: entity, relationship, event and time; decoding the entity embedding vector based on a time series convolution decoder to obtain a new entity quadruple, and calculating the similarity between the entity quadruple and the new entity quadruple; if the similarity exceeds a pre-set similarity threshold, constructing a time series-aware dynamic knowledge graph based on the entity quadruple.

[0013] Optionally, the entity quadruple is subjected to feature extraction based on the time sequence perception encoder to obtain an entity embedding vector, including: the entity quadruple is subjected to feature extraction based on the time sequence perception encoder on the historical time sequence production data to obtain a feature vector of the entity quadruple; and the feature vector of the entity quadruple is subjected to feature fusion through an attention mechanism to obtain the entity embedding vector.

[0014] In a second aspect, the present application provides a coal preparation plant scheduling system based on multi-time sequence data analysis, comprising: a data acquisition module for acquiring real-time production data of the coal preparation plant based on Internet of Things devices; a prediction module for generating a target washing and separation strategy based on the real-time production data and a pre-constructed multi-state uncertainty full-time sequence simulation model; wherein the multi-state uncertainty full-time sequence simulation model predicts the real-time production data based on a time sequence perception dynamic knowledge graph and a time sequence prediction model, and the time sequence perception dynamic knowledge graph is obtained through clustering analysis of historical production data by a time sequence perception dynamic knowledge graph algorithm; and a distribution module for distributing the target washing and separation strategy to corresponding production equipment.

[0015] In a third aspect, the present application provides an electronic device, comprising a processor and a memory, the memory stores computer executable instructions capable of being executed by the processor, and the processor executes the computer executable instructions to implement the steps of the method of any one of the first aspect.

[0016] In a fourth aspect, the present application provides a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program is run by a processor to execute the steps of the method of any one of the first aspect.

[0017] The present application has the following beneficial effects: The coal preparation plant scheduling method and system based on multi-time sequence data analysis provided by the application first collects real-time production data of the coal preparation plant based on Internet of Things equipment; then generates a target washing and sorting strategy based on real-time production data and a pre-constructed multi-state uncertainty full-time sequence simulation model; wherein the multi-state uncertainty full-time sequence simulation model predicts real-time production data based on a time sequence perception dynamic knowledge graph and a time sequence prediction model, and the time sequence perception dynamic knowledge graph is obtained by clustering analysis of historical production data through a time sequence perception dynamic knowledge graph algorithm; and finally the target washing and sorting strategy is distributed to corresponding production equipment. In the above method, the time sequence perception dynamic knowledge graph is constructed by clustering analysis of historical production data through the time sequence perception dynamic knowledge graph algorithm, which can mine potential laws or abnormal patterns between production equipment and process flow, and then realize multi-objective collaborative optimization through fusion analysis of real-time production data by the multi-state uncertainty full-time sequence simulation model to generate a target washing and sorting decision, thereby reducing the dependence on artificial experience and improving the scientific nature of the decision. At the same time, the above method realizes closed-loop decision optimization of the whole process of coal preparation plant production scheduling through fusion of historical data clustering analysis, real-time data monitoring and prediction suggestions, improves the equipment utilization rate, and reduces the energy consumption cost.

[0018] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and achieved by the structure particularly pointed out in the description, claims and drawings.

[0019] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are described in detail below, and the accompanying drawings are used for reference. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0021] Figure 1 A flowchart of a coal preparation plant scheduling method based on multi-time sequence data analysis provided by an embodiment of the present application; Figure 2 A flowchart of a multi-state uncertainty full-time sequence simulation model provided by an embodiment of the present application; Figure 3 A flowchart of a clustering analysis provided by an embodiment of the present application; Figure 4A full-process schematic diagram of production scheduling of a coal preparation plant is provided for an embodiment of the present application. Figure 5 A structural schematic diagram of a coal preparation plant scheduling system based on multi-time series data analysis is provided for an embodiment of the present application. Figure 6 A structural schematic diagram of an electronic device is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be described below in detail with reference to the drawings. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0023] At present, the traditional scheduling of coal preparation plants mainly relies on manual experience and scattered automation systems, which has the following problems: (1) Insufficient data integration: historical data is not effectively utilized, and there is a lack of clustering analysis of long-term regularities such as equipment failure and coal quality fluctuations, resulting in a lack of predictability in scheduling decisions.

[0024] (2) Real-time response lag: Although existing monitoring systems (such as video monitoring and sensor networks) can collect real-time data, they cannot quickly generate dynamic adjustment schemes. For example, due to the aging of the centralized control system in some coal preparation plants, process adjustment requires shutdown operation, resulting in resource waste.

[0025] (3) Weak prediction ability: There is a lack of multi-dimensional prediction of production demand, equipment failure and energy consumption trends, making it difficult to respond to market fluctuations and sudden working conditions. For example, although sensors are deployed in some coal preparation plants, predictive maintenance models have not been built.

[0026] (4) Lack of decision-making closed loop: scheduling instructions and execution feedback do not form a closed loop, resulting in low resource allocation efficiency. For example, although some coal preparation plants have achieved centralized management of data, they have not established an optimization mechanism based on real-time feedback.

[0027] Based on this, the coal preparation plant scheduling method and system based on multi-time series data analysis provided by the embodiments of the present application can reduce the dependence on manual experience and improve the scientific nature of decision-making.

[0028] To facilitate understanding of the present embodiment, first, a coal preparation plant scheduling method based on multi-time series data analysis disclosed by the present embodiment will be described in detail. The method is applied to a coal preparation plant scheduling system based on multi-time series data analysis and can be executed by an electronic device, such as a smart phone, a computer, a tablet computer, etc. Referring to Figure 1A flowchart of a coal preparation plant scheduling method based on multi-time series data analysis is shown, which shows that the method mainly includes the following steps S011 to step S103: Step S101: Collect real-time production data of the coal preparation plant based on Internet of Things devices.

[0029] In an embodiment, the coal preparation plant integrates Internet of Things devices (such as sensors, video monitoring), collects real-time production data through Internet of Things devices, and constructs a virtual mapping model of the coal preparation plant by combining digital twin technology, synchronizes real-time production data to the virtual mapping model for display, and realizes second-level response of device state and process flow. Real-time production data includes but is not limited to: device operation data (such as device failure condition, device processing capacity, washing and separation efficiency, current, temperature, vibration data), coal quality parameters (ash content, moisture content, coal quality fluctuation), flow rate, reagent dosage, storage capacity, etc.

[0030] Step S102: Generate target washing and separation strategy based on real-time production data and pre-constructed multi-state uncertainty full-time series simulation model.

[0031] In an embodiment, the multi-state uncertainty full-time series simulation model predicts real-time production data based on a time series perception dynamic knowledge graph and a time series prediction model. The time series perception dynamic knowledge graph is obtained by clustering analysis of historical production data through a time series perception dynamic knowledge graph algorithm.

[0032] Specifically, the multi-state uncertainty full-time series simulation model cleans the real-time production data to form a real-time time series production data set, extracts feature values using the real-time time series production data set, and predicts the health status and device failure probability of the device in combination with the time series perception dynamic knowledge graph. The health status of the device includes: normal state, abnormal state and fault state. If the device is in an abnormal state or a fault state, the relationship between production devices, device failure relationship and predicted device failure recovery time can be obtained according to the time series perception dynamic knowledge graph, the predicted value, probability distribution of production data and predicted completion time of each production link are dynamically calculated using a time series prediction model (such as LSTM), and finally the target washing and separation decision is generated according to the prediction results of the time series perception dynamic knowledge graph, Monte Carlo simulation engine and time series prediction model.

[0033] Step S103: Distribute the target washing and separation strategy to the corresponding production equipment.

[0034] In an embodiment, the system can automatically generate tasks according to the target washing and separation strategy, rely on the built wisdom body and workflow to quickly issue, circulate and execute the tasks, and can use two modes of system automatic distribution and manual confirmation after distribution to ensure the flexibility and reliability of task management.

[0035] Specifically, the intelligent command center (intelligence) task is realized by a workflow to realize the automation of task distribution and circulation, greatly improving the efficiency of task execution, realizing task generation, task content, task allocation, task execution, task monitoring and task feedback, etc. Real-time monitoring of task status ensures transparent and controllable task execution process. At the same time, the system can record the whole life cycle data of the task, which is convenient for subsequent analysis and optimization. In addition, the distribution of system tasks is not limited to physical persons, but also can realize the interaction and circulation of systems, robots and PLC systems.

[0036] The above coal preparation plant scheduling method based on multi-time sequence data analysis provided by the embodiment of the application can cluster and analyze historical production data by a time sequence perception dynamic knowledge graph algorithm to construct a time sequence perception dynamic knowledge graph, can mine potential rules or abnormal patterns between production equipment and process flow, and then realize multi-objective collaborative optimization by fusing and analyzing real-time production data by using a multi-state uncertainty full-time sequence simulation model, generate target washing and selecting decisions, thereby reducing dependence on artificial experience and improving the scientific nature of decisions. At the same time, the above method realizes closed-loop decision optimization of the whole process of coal preparation plant production scheduling by fusing historical data clustering analysis, real-time data monitoring and prediction suggestions, improves equipment utilization, and reduces energy consumption cost.

[0037] In one embodiment, for the foregoing step S102, when generating a target washing and selecting strategy based on real-time production data and a pre-constructed multi-state uncertainty full-time sequence simulation model, the following methods can be used, but are not limited to the following methods: Step 1: Based on the time sequence perception dynamic knowledge graph, the Monte Carlo simulation engine and the time sequence prediction model, the real-time production data is predicted and simulated to obtain the prediction results of the equipment health state and the production data.

[0038] In specific implementation, first, the real-time production data is cleaned to obtain a real-time time sequence production data set; wherein the real-time time sequence production data set includes uncertain coal separation parameters.

[0039] Specifically, the real-time production data includes but is not limited to real-time sensor collected data, laboratory test data, time length fluctuation operation data, production plans and customer demands, etc. The real-time production data is cleaned to obtain a real-time time sequence production data set containing uncertain coal separation parameters after processing.

[0040] Then, the real-time time sequence production data set is feature extracted to obtain feature values of the real-time production data.

[0041] Next, the real-time production data is predicted based on the time sequence perception dynamic knowledge graph and the feature values of the real-time production data to obtain the equipment health state.

[0042] Specifically, the feature values of the real-time production data are extracted by using the real-time time-series production dataset, and the device health state is obtained by predicting the feature values by using the time-series perception dynamic knowledge graph.

[0043] Finally, the prediction value and probability distribution of the production data are obtained by predicting based on the time-series prediction model and the real-time time-series production dataset.

[0044] Specifically, the future possible production scenarios are deduced by using the time-series prediction model for simulation, and the prediction value and probability distribution of the future production data are obtained. Based on the polymorphic uncertainty of the coal preparation scene, there is some uncertainty quantification during simulation. In the embodiment of the present application, the Monte Carlo simulation engine can be used to generate the probability distribution of the uncertain coal preparation parameters. For example, the Monte Carlo simulation is used to generate the probability distribution of the coal quality fluctuation (such as the ash content obeying the normal distribution N(15%, ±2%)).

[0045] Step 2: generating a target washing and separation strategy based on the prediction results of the device health state and the production data.

[0046] In a specific implementation, first, a plurality of washing and separation strategies are generated based on the device health state, the prediction value and probability distribution of the production data; then, a target washing and separation strategy is determined from the plurality of washing and separation strategies based on a preset multi-objective optimization function; wherein the multi-objective optimization function is:

[0047] wherein, MTBF is the average failure-free operating time of the production equipment, α , β , gamma is a weight coefficient.

[0048] Specifically, a plurality of washing and separation strategies are generated in combination with the device health state, the prediction value and probability distribution of the production data, and then the best washing and separation strategy, i.e., the target washing and separation strategy, is selected according to the multi-objective optimization function, with the system reliability optimization as the core target and taking into account the economy, quality standard, etc. In the multi-objective optimization function, MTBF is used to represent production continuity, and the washing and separation qualification rate is used to represent process stability. In the embodiment of the present application, the multi-objective optimization function is constructed in combination with production continuity, process stability and economy, and the washing and separation strategy selected by using the multi-objective optimization function is more in line with the production situation of the coal preparation plant and can optimize the configuration of the coal preparation plant.

[0049] In one embodiment, after the target washing and separation strategy is distributed to the corresponding production equipment, the above further includes: obtaining an actual production result after the distribution of the target washing and separation strategy, and feeding back the actual production result to the historical production database; and optimizing the time-series perception dynamic knowledge graph based on the actual production result and the target washing and separation strategy.

[0050] In specific implementation, after the optimal cleaning and sorting strategy is generated, the final actual production result can also be tracked, and the actual production result is fed back to the historical production database, and a time-aware dynamic knowledge graph relationship is established according to the actual production result and the predicted result (i.e., the target cleaning and sorting strategy), and the time-aware dynamic knowledge graph is continuously optimized.

[0051] For ease of understanding, the embodiment of the present application also provides a flowchart of a multi-state uncertainty full-time sequence simulation model, as shown in Figure 2 The flowchart comprises the following steps: Dynamic perception: production real-time data (collection of sensors, laboratory tests, etc.) enter the dynamic perception layer for data cleaning, and after processing, high-quality time sequence data sets with uncertainty are formed.

[0052] Time sequence perception model database: according to the feature values extracted from the dynamic perception data set, the historical optimal solution is found, and the time sequence perception model is optimized. The time sequence perception model can be a pre-constructed time sequence perception dynamic knowledge graph, which is used to capture the dependence relationship between coal quality parameters and equipment efficiency, and then calculate the corresponding coal quality and equipment relationship at this time.

[0053] Simulation: the time sequence data set drives the simulation deduction layer to construct a time sequence prediction model and perform simulation, deduces possible future production scenarios, and outputs the probability distribution of future production indicators. During simulation, there are some uncertain quantities, which can be quantified by Monte Carlo simulation to generate a coal quality fluctuation probability distribution (such as ash content obeying a normal distribution N(15%, ±2%)).

[0054] Optimal strategy generation: analyze the probability distribution results output by the simulation deduction layer, take the system reliability optimization as the core target (considering economy, quality standard, etc.), search and recommend the best cleaning and sorting operation parameters (such as medium density set value, flow rate, reagent amount) or strategy (such as equipment start-stop combination, coal blending adjustment), and generate the best cleaning and sorting strategy.

[0055] Feedback loop: according to the cleaning and sorting strategy adjusted by the model this time and the actual result, a time-aware dynamic knowledge graph relationship is established, and the model is continuously optimized.

[0056] The key of the historical data clustering analysis is to effectively analyze the historical data, but in the traditional clustering analysis, the time dimension is embedded in the entity or relationship, the four-tuple is reduced to a three-tuple, and then the static knowledge graph is completed according to the static knowledge graph completion theory, the scene elements are lost, and the clustering analysis result is greatly discounted. Based on this, the time sequence perception encoder and the time sequence convolution decoder are analyzed based on the historical production data in the embodiment of the present application, and a time sequence perception dynamic knowledge graph is constructed. As shown in Figure 3 The construction of the time sequence perception dynamic knowledge graph can be realized in the following ways, but is not limited to the following ways: Firstly, historical production data is acquired, and the historical production data is cleaned to obtain historical time series production data.

[0057] Then, the historical time series production data is subjected to feature extraction of entity quadruples based on a time series perception encoder to obtain entity embedding vectors.

[0058] In a specific implementation, the time series perception encoder is based on an improved convolutional neural network for feature extraction of entity quadruples, and the improved convolutional neural network is based on the time sequence of the historical time series production data to perform feature extraction of entity quadruples in the spatial dimension and the time dimension, and the entity quadruples include entities (s, mainly referring to devices, sensors, etc.), relationships (r, mainly referring to physical relationships and logical relationships between entities), events (o, mainly some normal, abnormal, fault, etc. phenomena) and time (t). As shown in Table 1, compared with the traditional convolutional neural network, the time series perception encoder used in the embodiment of the present application slides in the spatial and time dimensions (height, width, time), and the network structure (such as 3D convolution, recursive connection) depends on the time sequence of the input, which is used to understand and utilize the spatio-temporal causal relationship and dynamic evolution, and is mainly used for time series prediction tasks.

[0059] Table 1 Comparison of time series perception encoder and traditional convolutional neural network

[0060] Specifically, the time series perception encoder embeds the time dimension into entities and relationships as vectors of the same scale, and realizes feature extraction of entity quadruples through an improved graph convolutional neural network, and models the time in the entity quadruples and other entities (coal preparation equipment, sensors, etc.) and relationships (washing and separation process, equipment process) as embedding vectors of the same scale. Based on this, in the embodiment of the present application, when the entity quadruples are subjected to feature extraction based on the time series perception encoder to obtain entity embedding vectors, the following methods can be used, but are not limited to: first, the entity quadruples are subjected to feature extraction based on the time series perception encoder to obtain feature vectors of the entity quadruples; and then the feature vectors of the entity quadruples are subjected to feature fusion through an attention mechanism to obtain the entity embedding vectors.

[0061] In the embodiment of the present application, the time information is fused into the entities through the attention mechanism, and each node learns the neighborhood features with emphasis through the improved graph attention mechanism, and all quadruples (s, r, o, t) are embedded and represented as an input matrix of k x 4 in a k-dimensional space.

[0062] Then, the entity embedding vectors are decoded based on a time series convolutional decoder to obtain new entity quadruples, and the similarity between the entity quadruples and the new entity quadruples is calculated.

[0063] In specific implementation, after the dynamic knowledge base (i.e. historical production data) is encoded by the time-aware encoder, the central entity captures the features of multiple dimensions in the quadruple, especially the time dimension features, by aggregating the neighborhood features. The time convolution decoder takes the quadruple with the size of k x 4 after embedding as the input matrix, and the convolution layer uses multiple convolution kernels with the size of 1 x 4 to extract the quadruple features from multiple angles for decoding calculation to obtain a new entity quadruple, and the similarity of the quadruple is evaluated by converting the matrix into a numerical value.

[0064] Finally, if the similarity exceeds the pre-set similarity threshold, a time-aware dynamic knowledge graph is constructed based on the entity quadruple.

[0065] In specific implementation, if the similarity between the entity quadruple and the new entity quadruple exceeds the similarity threshold (such as 80% or 90%, which can be set according to the actual production scene), it indicates that the prediction performance of the time convolution decoder meets the requirements, and then the time-aware encoder and the time convolution decoder can be used for clustering analysis of the historical production data to construct a time-aware dynamic knowledge graph.

[0066] In the embodiment of the present application, based on the establishment of the relationship of the time-aware dynamic knowledge graph, when the equipment fails or the coal quality is abnormal, the multi-factor analysis can be performed in the same time dimension, thereby improving the accuracy of the clustering analysis.

[0067] For ease of understanding, the present application also provides a full-process schematic diagram of production scheduling of a coal preparation plant, as shown in Figure 4 As shown, the present application realizes the full-cycle closed-loop management of the scheduling of the coal preparation plant by constructing a "history-real-time-future" three-stage time series data analysis framework, and truly realizes the predictive production management: Historical data clustering analysis: the time-aware dynamic knowledge graph algorithm is used to perform clustering analysis on the historical production data (such as equipment failure records, coal quality parameters, and energy consumption curves), and identify potential rules and abnormal patterns.

[0068] Real-time monitoring and dynamic adjustment: integrate Internet of Things devices (such as sensors and video monitoring) to collect real-time data, and combine digital twin technology to construct a virtual mapping model to realize the second-level response of the device state and the process flow.

[0069] Prediction and collaborative optimization: based on the time prediction model (such as LSTM and ARIMA), the future production demand, the probability of equipment failure, and the energy consumption trend are predicted to generate a multi-objective optimization scheduling scheme.

[0070] Specifically, real-time sensing data is collected and uploaded to the data warehouse. The pre-constructed time series knowledge graph is used in combination with large model algorithms for prediction reasoning and full-time series simulation to achieve dynamic production scheduling, dynamic strategy recommendation, fault recovery prediction, real-time monitoring and adjustment in coal preparation plants, and task distribution and task flow through the command system.

[0071] 1) Dynamic production scheduling Based on real-time equipment operation data (such as equipment failure conditions, equipment processing capacity, washing and separation efficiency), coal quality parameters (ash content, moisture content), storage capacity, etc., the expected completion time of each production link is dynamically calculated through a time series prediction model (such as LSTM), and the optimal washing and separation production plan is given.

[0072] When the raw coal storage changes or the raw coal washing capacity suddenly increases, the system reminds the dispatcher to extend the shift or activate other production lines through "flexible completion time".

[0073] 2) Dynamic strategy recommendation In combination with real-time load, energy consumption, and coal quality fluctuation data, dynamic production strategies (such as adjusting the air pressure parameters of the jig, switching the separation medium ratio) are generated, and the suggestions are pushed through the visual interface.

[0074] 3) Fault recovery prediction When equipment failure (such as hydraulic system leakage of filter press) occurs, the system predicts the maintenance completion time based on historical maintenance records, spare parts inventory, and personnel location, and dynamically adjusts the production plan (such as switching to backup equipment or reducing processing capacity). The system associates fault types, maintenance resources, and process dependency relationships through a knowledge graph to improve fault recovery efficiency.

[0075] 4) Real-time monitoring and dynamic adjustment Real-time analysis of device current, temperature, and vibration data is performed using edge computing, and the device health status is predicted based on the dynamic knowledge graph of time series perception, and maintenance tasks are given in a timely manner to reduce unplanned downtime.

[0076] 5) Task distribution The system automatically generates tasks based on status and warning information, relies on the built-in wisdom and workflow to quickly issue, transfer, and execute tasks.

[0077] The coal preparation plant scheduling method based on multi-time sequence data analysis provided by the embodiment of the present application balances production efficiency, energy consumption and equipment life by using the multi-state uncertainty full-time sequence simulation model, realizes multi-objective collaborative optimization, realizes the upgrade from "early warning-suggestion" to "prediction-automatic adjustment-feedback calibration" according to time sequence analysis processing, realizes the continuous evolution of the model, reduces the dependence on artificial experience through time sequence data fusion analysis, reduces the misjudgment risk caused by one-sidedness of data, and improves the scientific nature of decision-making; the real-time closed-loop feedback mechanism can quickly respond to sudden working conditions (such as equipment failure and sudden change of coal quality), reduce downtime, and enhance the robustness of the system; the equipment utilization rate is improved through predictive scheduling, and the energy consumption cost is reduced.

[0078] For the coal preparation plant scheduling method based on multi-time sequence data analysis provided by the foregoing embodiment, the embodiment of the present application provides a coal preparation plant scheduling system based on multi-time sequence data analysis, which refers to a structure diagram of a coal preparation plant scheduling system based on multi-time sequence data analysis shown in Figure 5 The system mainly includes the following parts: The data acquisition module 501 is configured to acquire real-time production data of the coal preparation plant based on Internet of Things devices. The prediction module 502 is configured to generate a target washing strategy based on the real-time production data and a pre-constructed multi-state uncertainty full-time sequence simulation model; the multi-state uncertainty full-time sequence simulation model is configured to predict the real-time production data based on a time sequence perception dynamic knowledge graph, a Monte Carlo simulation engine and a time sequence prediction model; the time sequence perception dynamic knowledge graph is obtained by clustering analysis of historical production data through a time sequence perception dynamic knowledge graph algorithm. The distribution module 503 is configured to distribute the target washing strategy to corresponding production equipment.

[0079] The coal preparation plant scheduling system based on multi-time sequence data analysis provided by the embodiment of the present application can mine potential rules or abnormal patterns between production equipment and process flow by clustering analysis of historical production data through a time sequence perception dynamic knowledge graph algorithm to construct a time sequence perception dynamic knowledge graph, and then realize multi-objective collaborative optimization by fusion analysis of real-time production data through a multi-state uncertainty full-time sequence simulation model to generate a target washing decision, thereby reducing the dependence on artificial experience and improving the scientific nature of decision-making; at the same time, the system realizes closed-loop decision optimization of the whole process of coal preparation plant production scheduling by fusing historical data clustering analysis, real-time data monitoring and predictive suggestions, improves the equipment utilization rate, and reduces the energy consumption cost.

[0080] In an implementation, the prediction module 502 is specifically configured to: predict and simulate the real-time production data based on the time-aware dynamic knowledge graph and the time series prediction model to obtain the prediction results of the equipment health status and the production data; and generate the target washing and separation strategy based on the prediction results of the equipment health status and the production data.

[0081] In an implementation, the prediction module 502 is further configured to: perform data cleaning on the real-time production data to obtain a real-time time series production data set; wherein the real-time time series production data set includes uncertain coal separation parameters; perform feature extraction on the real-time time series production data set to obtain feature values of the real-time production data; perform prediction based on the time-aware dynamic knowledge graph and the feature values of the real-time production data to obtain the equipment health status; and perform prediction based on the time series prediction model and the real-time time series production data set to obtain prediction values and probability distributions of the production data, wherein a Monte Carlo simulation engine is used to generate the probability distributions of the uncertain coal separation parameters.

[0082] In an implementation, the prediction module 502 is further configured to: generate a plurality of washing and separation strategies based on the equipment health status, the prediction values and the probability distributions of the production data; determine the target washing and separation strategy from the plurality of washing and separation strategies based on a preset multi-objective optimization function; wherein the multi-objective optimization function is:

[0083] wherein, MTBF is the average failure-free operating time of the production equipment, α , β , gamma is a weight coefficient.

[0084] In an implementation, the system further includes a feedback module configured to: obtain actual production results after distribution of the target washing and separation strategy, and feed back the actual production results to the historical production database; and optimize the time-aware dynamic knowledge graph based on the actual production results and the target washing and separation strategy.

[0085] In an implementation, the system further comprises a clustering analysis module configured to: acquire historical production data, and perform data cleaning on the historical production data to obtain historical time-series production data; perform feature extraction on entity quadruples based on a time-series perception encoder to obtain entity embedding vectors; wherein the time-series perception encoder performs feature extraction on the entity quadruples based on an improved convolutional neural network, the improved convolutional neural network performs feature extraction on the entity quadruples in spatial and time dimensions based on a time sequence of the historical time-series production data, the entity quadruples comprise entities, relationships, events, and times; decode the entity embedding vectors based on a time-series convolutional decoder to obtain new entity quadruples, and calculate a similarity between the entity quadruples and the new entity quadruples; and if the similarity exceeds a pre-set similarity threshold, construct a time-series perception dynamic knowledge graph based on the entity quadruples.

[0086] In an implementation, the clustering analysis module is further configured to: perform feature extraction on the entity quadruples based on the time-series perception encoder to obtain feature vectors of the entity quadruples; and perform feature fusion on the feature vectors of the entity quadruples through an attention mechanism to obtain the entity embedding vectors.

[0087] It should be noted that the system provided by the embodiments of the present application has the same implementation principle and technical effects as the foregoing method embodiments, and for brevity, the parts not mentioned in the system embodiment part can be referred to the corresponding content in the foregoing method embodiments. The specific numerical values provided in the embodiments of the present application are only exemplary and are not limited herein.

[0088] The embodiments of the present application further provide an electronic device, specifically, the electronic device comprises a processor and a storage device; the storage device stores a computer program, and the computer program performs the method according to any one of the above embodiments when executed by the processor.

[0089] Figure 6 A structural schematic diagram of an electronic device provided by the embodiments of the present application is shown in the figure, and the electronic device 100 comprises a processor 60, a memory 61, a bus 62, and a communication interface 63, wherein the processor 60, the communication interface 63, and the memory 61 are connected through the bus 62; the processor 60 is configured to execute executable modules stored in the memory 61, such as a computer program.

[0090] The memory 61 can include a high-speed random access memory (RAM), and can also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 63 (which can be wired or wireless), and the Internet, a wide area network, a local area network, a metropolitan area network, etc. can be used.

[0091] The bus 62 can be an ISA bus, a PCI bus, an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one bidirectional arrow is used in the figure, but it does not mean that there is only one bus or only one type of bus.

[0092] The memory 61 is used to store programs, and the processor 60 executes the programs after receiving execution instructions. The method executed by the device defined by the flow process disclosed in any of the embodiments of the application can be applied to the processor 60 or implemented by the processor 60.

[0093] The processor 60 can be an integrated circuit chip with signal processing capability. In the implementation process, each step of the above method can be completed by the integrated logic circuit of hardware in the processor 60 or the instruction in the form of software. The processor 60 described above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. It can implement or execute the disclosed methods, steps and logic block diagrams in the embodiments of the application. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. The storage medium in the art. The storage medium is located in the memory 61, and the processor 60 reads the information in the memory 61 and combines the hardware to complete the steps of the above method.

[0094] The computer program product of the readable storage medium provided by the embodiment of the present application comprises a computer readable storage medium storing program codes, and the program codes comprise instructions for executing the method described in the foregoing method embodiments. The specific implementation can be referred to the foregoing method embodiments, and will not be described here.

[0095] When the functions are realized in the form of software function units and sold or used as independent products, the functions can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art can be embodied in the form of software products. The computer software product is stored in a storage medium and includes instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0096] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present application, which are used to illustrate the technical solutions of the present application, but not to limit the present application. The protection scope of the present application is not limited to this. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily think of changes to the technical solutions recorded in the foregoing embodiments, or replace some technical features with equivalent replacements within the technical range disclosed by the present application. The modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A coal preparation plant scheduling method based on multi-time series data analysis, characterized in that: include: Collect real-time production data from coal preparation plants based on IoT devices; Generate a target selection strategy based on the real-time production data and a pre-built polymorphic uncertainty full-time series simulation model; wherein the polymorphic uncertainty full-time series simulation model predicts the real-time production data based on a time-series-aware dynamic knowledge graph and a time-series prediction model, and the time-series-aware dynamic knowledge graph is obtained by clustering and analyzing historical production data using a time-series-aware dynamic knowledge graph algorithm; Distribute the target washing strategy to the corresponding production equipment.

2. The method according to claim 1, characterized in that Generate a target selection strategy based on the real-time production data and a pre-built multi-state uncertainty full-time simulation model, including: Predict and simulate the real-time production data based on a time-series-aware dynamic knowledge graph and a time-series prediction model to obtain prediction results of equipment health status and production data; A target washing strategy is generated based on the prediction results of the equipment health status and production data.

3. The method according to claim 2, characterized in that The real-time production data is predicted and simulated based on a time-series-aware dynamic knowledge graph and a time-series prediction model to obtain prediction results for equipment health status and production data, including: Performing data cleaning on the real-time production data to obtain a real-time time series production data set; wherein the real-time time series production data set includes uncertain coal preparation parameters; Performing feature extraction on the real-time time series production data set to obtain feature values ​​of the real-time production data; Predicting the equipment health status based on the time-series-aware dynamic knowledge graph and the feature values ​​of the real-time production data; Prediction is performed based on the time series prediction model and the real-time time series production data set to obtain predicted values ​​and probability distribution of production data; wherein, a Monte Carlo simulation engine is used to generate the probability distribution of the uncertain coal preparation parameters.

4. The method according to claim 3, characterized in that Generate a target cleaning strategy based on the prediction results of the equipment health status and production data, including: generating a plurality of washing strategies based on the health status of the equipment, the predicted value of the production data, and the probability distribution; Determine a target washing strategy from the plurality of washing strategies based on a preset multi-objective optimization function; wherein the multi-objective optimization function is: in, MTBF is the average trouble-free working time of production equipment, α 、 β 、 γ is the weight coefficient.

5. The method according to claim 1, wherein After distributing the target washing strategy to the corresponding production equipment, the method further includes: Obtaining actual production results after the target washing strategy is distributed, and feeding the actual production results back to a historical production database; Based on the actual production results and the target washing strategy, the time-series-aware dynamic knowledge graph is optimized.

6. The method according to claim 1, characterized in that Before collecting real-time production data of the coal preparation plant based on IoT devices, it also includes: Acquire historical production data, and perform data cleaning on the historical production data to obtain historical time-series production data; Performing feature extraction of entity quads on the historical time-series production data based on a time-series-aware encoder to obtain an entity embedding vector; wherein the time-series-aware encoder performs feature extraction of entity quads based on an improved convolutional neural network, and the improved convolutional neural network performs feature extraction of entity quads in spatial and temporal dimensions based on the time sequence of the historical time-series production data, wherein the entity quads include: entity, relationship, event, and time; Decoding the entity embedding vector based on a temporal convolutional decoder to obtain a new entity quadruple, and calculating the similarity between the entity quadruple and the new entity quadruple; If the similarity exceeds a preset similarity threshold, a time-series-aware dynamic knowledge graph is constructed based on the entity quadruple.

7. The method according to claim 6, characterized in that Feature extraction is performed on the entity quadruple based on a temporal-aware encoder to obtain an entity embedding vector, including: Performing feature extraction of entity quadruple on the historical time series production data based on a time series-aware encoder to obtain a feature vector of the entity quadruple; The feature vectors of the entity quadruple are fused through the attention mechanism to obtain the entity embedding vector.

8. A coal preparation plant dispatching system based on multi-time series data analysis, characterized in that: include: Data acquisition module, used to collect real-time production data of coal preparation plants based on IoT devices; A prediction module is configured to generate a target washing strategy based on the real-time production data and a pre-built polymorphic uncertainty full-time series simulation model; wherein the polymorphic uncertainty full-time series simulation model predicts the real-time production data based on a time-series-aware dynamic knowledge graph and a time-series prediction model, and the time-series-aware dynamic knowledge graph is obtained by clustering historical production data using a time-series-aware dynamic knowledge graph algorithm; The distribution module is used to distribute the target washing strategy to the corresponding production equipment.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are performed.

Images