A Smart Coal Blending Method Based on Predictive Models

Abstract

This invention relates to an intelligent coal blending method based on a predictive model. The method includes the following steps: Step 1: Data acquisition and preprocessing; Step 2: Model training, where the model adopts an Encoder-Decoder structure, wherein the Encoder model is a state estimation model and the Decoder model is a system dynamic model; Step 3: Quality control of the finished coal at the outlet. This technical solution significantly improves the quality stability of the finished coal, ensuring that each batch accurately meets the stringent requirements of customers, and greatly reduces reliance on manual operation, thereby significantly reducing labor costs. More importantly, the intelligent coal blending system can precisely control the blending ratio of raw coal, avoiding the previous practice of producing finished coal with a quality level higher than the actual demand to ensure quality standards. This further optimizes production costs while ensuring quality, achieving a dual improvement in economy and quality.

Description

Technical Field

[0001] This invention relates to an intelligent coal blending method, specifically an intelligent coal blending method based on a prediction model, and belongs to the field of intelligent coal blending technology. Background Technology

[0002] Coal blending is a clean and efficient coal utilization technology that mixes raw coal of different qualities in a certain ratio to obtain finished coal that meets requirements. Traditional coal blending involves on-site operators manually estimating and setting the proportions of raw coal of different qualities by manipulating the feeder gate under the product bin. Due to significant fluctuations in the quality of raw coal in the product bin, delayed feedback from online ash content testing, and poor adaptability to time-varying and nonlinear transport in the system, traditional blending methods cannot adjust the blending ratio in real time according to changes in the quality of the finished coal, resulting in unstable quality of the finished coal and failing to meet customer needs. Therefore, a new solution is urgently needed to address this technical problem. Summary of the Invention

[0003] This invention addresses the technical problems existing in the prior art by providing an intelligent coal blending method based on a predictive model. This technical solution improves the stability of commercial coal quality and avoids the waste of high-priced high-quality coal while meeting the needs of product users.

[0004] To achieve the above objectives, the technical solution of the present invention is as follows: an intelligent coal blending method based on a prediction model, the method comprising the following steps:

[0005] Step 1: Data acquisition and preprocessing

[0006] Step 2: Model training. The model adopts an Encoder-Decoder structure, where the Encoder model is the state estimation model and the Decoder model is the system dynamics model.

[0007] Step 3: Quality control of exported finished coal.

[0008] Step 1: Data acquisition and preprocessing, specifically as follows:

[0009] Step 1-1: Collect historical data of the intelligent coal blending process. The dataset consists of several complete coal blending orders, and the quality data of the raw coal bunker is recorded in each order. And production data that varies over time, with each time sample in the sequence data including the dosing ratio. and the quality of exported finished coal ,in Indicates the order number. [1, N] represents the raw coal bunker number. Indicates time, Indicates order The Middle The quality of each raw coal bunker Indicates order middle Time of the first The blending ratio of each raw coal bunker, Indicates order middle The quality of exported finished coal at any given time;

[0010] Steps 1-2: Data preprocessing. The exported finished coal quality data undergoes center-average filtering to eliminate measurement noise. The center-average filtering method is as follows:

[0011]

[0012] in The filter width, As a preset value, This represents the quality of the exported finished coal at time k in order i. This represents the quality of the exported finished coal after filtering at time t in order i.

[0013] Step 2: Model training is as follows:

[0014] Step 2-1: The state estimation model predicts the current state of the system based on the system's most recent input and output history sequences, and its form is as follows:

[0015]

[0016] in Let this be the system state at time t in order i. In order i The blending sequence data of the j-th raw coal bunker from time t-1 to time t-1. This represents the quality of the j-th raw coal bunker in order i. In order i The finished coal quality sequence data from time +1 to time t, where T1 is the first preset time duration. For the parameters of the state estimation model,

[0017] Step 2-2: System dynamic model. The system dynamic model includes two sub-models: the state transition model and the output model.

[0018] The state transition model describes the state transition characteristics of the system, and its form is as follows:

[0019]

[0020] in, Let this be the system state at time t in order i. Let i be the blending ratio of the j-th raw material coal bunker at time t in order i. For the quality of the j-th raw coal bunker in order i, The parameters of the state transition model,

[0021] The output model maps the system state to the controlled output value, and its form is as follows:

[0022]

[0023] in Let this be the system state at time t in order i. This represents the system output at time t in order i, i.e., the predicted quality of the finished coal. These are the parameters for the output model.

[0024] The state estimation model is an encoder network that transforms a sequence into a state vector, implemented using network structures such as transformers or LSTM. The system dynamics model is a decoder network that describes the dynamic characteristics of the system, and can be implemented using an LSTM network. During implementation, it is necessary to ensure that the dimensions of the state vectors in the encoder and decoder networks are equal. The output model consists of a simple two-layer linear network, with a non-linear activation function used between the two linear layers to increase the non-linearity of the model.

[0025] Step 3: Quality control of exported finished coal, specifically as follows:

[0026] The optimal blending ratio of each raw coal bunker is determined based on the trained state estimation model and system dynamic model, thereby achieving precise control of the quality of the finished coal at the outlet. The blending ratio of each raw coal bunker meets certain constraints:

[0027] Constraint (1): ,

[0028] Constraint (2): ,

[0029] in and These represent the minimum and maximum blending ratios of the j-th raw material coal bin in order i, respectively.

[0030] Step 3 also includes the following constraints:

[0031] Constraint (3): When When the quality requirements for exported finished coal are less than those of order i, the minimum blending ratio of the best quality raw coal is the current blending ratio of that type of coal, i.e. ;when When the quality requirements for exported finished coal exceed those of order i, the highest blending ratio of the best quality raw coal is the current blending ratio of that type of coal, i.e. The quality of the exported finished coal is obtained at time t. Then, the following steps are performed to determine the optimal blending ratio, thereby controlling the quality of the exported finished coal to meet customer requirements.

[0032] (1) Construct a historical production sequence. The historical production sequence in production order i includes the blending ratio of each raw coal bunker from t-T1 to the current time t-1. The quality sequence of finished coal from time t-T1+1 to time t And the quality data of each raw coal bunker, Where T1 is the first preset duration,

[0033] (2) Based on the constructed historical production sequence, the system state at the current moment is predicted using the state estimation model. ,

[0034] (3) Generate several blending strategies for raw coal bunkers. First, determine the minimum blending ratio for each coal bunker. and the highest admixture ratio Next, for raw coal bunkers 1 to N-1, within the interval... A uniform sampling process is performed within the N-1 coal bunkers to obtain m values; then, the blending ratios of these N-1 coal bunkers are arranged and combined to obtain... The blending strategy, wherein the blending ratio of the Nth raw material coal bunker can be determined by... To obtain; ultimately, to be dissatisfied. Remove the d policies that have the conditions.

[0035] (4) Based on Seed mixing strategy to generate The predicted output sequence, for each dosing strategy, is based on the current system state obtained in step 2. and the quality of each raw coal bunker A predicted output sequence can be obtained iteratively using the system dynamic model and the output model, that is, from the current time t to the future. At any given time, the current blending strategy is continuously executed to obtain the finished coal quality estimation sequence.

[0036] (5) First, from Select from the finished coal quality estimation sequences that are within the finished coal quality setpoint A sequence that fluctuates within an allowed range, i.e. Let ns be the number of sequences that meet the criteria after screening; then, calculate the relationship between each selected sequence and the set value of finished coal quality. The sum of absolute deviation values Secondly, calculate the corresponding product cost when performing the dosing ratio for each selected sequence. ,in Let the price of the j-th raw coal bunker in order i be denoted; then, calculate the overall target value. ,in The weighting coefficient is used to balance the control effect and product cost; finally, the dosing ratio with the smallest overall target value is selected as the decision result at time t.

[0037] An electronic device includes: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the intelligent coal blending method based on a predictive model.

[0038] A computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the aforementioned intelligent coal blending method based on a predictive model.

[0039] Compared to existing technologies, this invention has the following advantages: 1. The intelligent coal blending method proposed in this invention, based on a predictive model, determines the optimal blending ratio at the current moment based on the quality of each raw coal bunker, the blending ratio sequence data of each raw coal bunker, and the finished coal quality sequence data, effectively overcoming problems such as large fluctuations in raw coal quality, time-varying and nonlinear operation of the system; 2. This invention significantly improves the quality stability of finished coal, ensuring that each batch accurately meets the strict requirements of customers, and greatly reduces reliance on manual operation, thereby significantly reducing labor costs. Labor costs are expected to decrease by 20%-40%. Furthermore, the system can precisely control the blending ratio of raw coal, avoiding the previous practice of producing finished coal higher than actual demand to ensure quality standards, reducing raw material costs by approximately 5%-15% and operating costs by approximately 5%. In terms of production efficiency, intelligent scheduling and real-time monitoring can shorten the production cycle by 10%-30%. These improvements not only ensure quality compliance but also further optimize production costs, achieving a dual improvement in economy and quality, bringing significant economic benefits and market competitive advantages to enterprises. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0041] Figure 2 This is a schematic diagram of the overall control process.

[0042] Figure 3 This is a diagram of the actual control process. Detailed Implementation

[0043] To enhance understanding of the present invention, the embodiments will be described in detail below with reference to the accompanying drawings.

[0044] Example 1: See Figures 1-3 A smart coal blending method based on a predictive model, the method comprising the following steps:

[0045] Step 1: Data acquisition and preprocessing

[0046] Step 2: Model training. The model adopts an Encoder-Decoder structure, where the Encoder model is the state estimation model and the Decoder model is the system dynamics model.

[0047] Step 3: Quality control of exported finished coal.

[0048] Step 1: Data acquisition and preprocessing, specifically as follows:

[0049] Step 1-1: Collect historical data of the intelligent coal blending process. The dataset consists of several complete coal blending orders, and the quality data of the raw coal bunker is recorded in each order. And production data that varies over time, with each time sample in the sequence data including the dosing ratio. and the quality of exported finished coal ,in Indicates the order number. [1, N] represents the raw coal bunker number. Indicates time, Indicates order The Middle The quality of each raw coal bunker Indicates order middle Time of the first The blending ratio of each raw coal bunker, Indicates order middle The quality of exported finished coal at any given time;

[0050] Steps 1-2: Data preprocessing. The exported finished coal quality data undergoes center-average filtering to eliminate measurement noise. The center-average filtering method is as follows:

[0051]

[0052] in The filter width, As a preset value, This represents the quality of the exported finished coal at time k in order i. This represents the quality of the exported finished coal after filtering at time t in order i.

[0053] Step 2: Model training is as follows:

[0054] Step 2-1: The state estimation model predicts the current state of the system based on the system's most recent input and output history sequences, and its form is as follows:

[0055]

[0056] in Let this be the system state at time t in order i. In order i The blending sequence data of the j-th raw coal bunker from time t-1 to time t-1. This represents the quality of the j-th raw coal bunker in order i. In order i The finished coal quality sequence data from time +1 to time t, where T1 is the first preset time duration. For the parameters of the state estimation model,

[0057] Step 2-2: System dynamic model. The system dynamic model includes two sub-models: the state transition model and the output model.

[0058] The state transition model describes the state transition characteristics of the system, and its form is as follows:

[0059]

[0060] in, Let this be the system state at time t in order i. Let i be the blending ratio of the j-th raw material coal bunker at time t in order i. For the quality of the j-th raw coal bunker in order i, The parameters of the state transition model,

[0061] The output model maps the system state to the controlled output value, and its form is as follows:

[0062]

[0063] in Let this be the system state at time t in order i. This represents the system output at time t in order i, i.e., the predicted quality of the finished coal. These are the parameters for the output model.

[0064] The state estimation model is an encoder network that transforms a sequence into a state vector, implemented using network structures such as transformers or LSTM. The system dynamics model is a decoder network that describes the dynamic characteristics of the system, and can be implemented using an LSTM network. During implementation, it is necessary to ensure that the dimensions of the state vectors in the encoder and decoder networks are equal. The output model consists of a simple two-layer linear network, with a non-linear activation function used between the two linear layers to increase the non-linearity of the model.

[0065] Based on the preprocessed data, a training dataset is constructed. Each training sample in the training set contains three parts: the input to the state estimation model, a partial input to the system dynamics model, and the expected output of the output model. The input to the state estimation model includes... ,in Given the unfiltered quality data of the finished coal product at the export site, the output of the state estimation model is: The input to the system dynamic model is ,in Provided by the state estimation model, Provided by the dataset, the output of the system dynamic model is The input to the output model is... The output of the output model is Where T1 and T2 are the first and second preset durations, respectively. Output based on the output model. The expected output of the model in the dataset The loss is calculated to update the parameters of the state estimation model, the system dynamics model, and the output model.

[0066] Step 3: Quality control of exported finished coal, specifically as follows:

[0067] The optimal blending ratio of each raw coal bunker is determined based on the trained state estimation model and system dynamic model, thereby achieving precise control of the quality of the finished coal at the outlet. The blending ratio of each raw coal bunker meets certain constraints:

[0068] Constraint (1): ,

[0069] Constraint (2): ,

[0070] in and These represent the minimum and maximum blending ratios of the j-th raw material coal bin in order i, respectively.

[0071] Step 3 also includes the following constraints:

[0072] Constraint (3): When When the quality requirements for exported finished coal are less than those of order i, the minimum blending ratio of the best quality raw coal is the current blending ratio of that type of coal, i.e. ;when When the quality requirements for exported finished coal exceed those of order i, the highest blending ratio of the best quality raw coal is the current blending ratio of that type of coal, i.e. The quality of the exported finished coal is obtained at time t. Then, the following steps are performed to determine the optimal blending ratio, thereby controlling the quality of the exported finished coal to meet customer requirements.

[0073] (1) Construct a historical production sequence. The historical production sequence in production order i includes the blending ratio of each raw coal bunker from t-T1 to the current time t-1. The quality sequence of finished coal from time t-T1+1 to time t And the quality data of each raw coal bunker, Where T1 is the first preset duration,

[0074] (2) Based on the constructed historical production sequence, the system state at the current moment is predicted using the state estimation model. ,

[0075] (3) Generate several blending strategies for raw coal bunkers. First, determine the minimum blending ratio for each coal bunker. and the highest admixture ratio Next, for raw coal bunkers 1 to N-1, within the interval... A uniform sampling process is performed within the N-1 coal bunkers to obtain m values; then, the blending ratios of these N-1 coal bunkers are arranged and combined to obtain... The blending strategy, wherein the blending ratio of the Nth raw material coal bunker can be determined by... To obtain; ultimately, to be dissatisfied. Remove the d policies that have the conditions.

[0076] (4) Based on Seed mixing strategy to generate The predicted output sequence, for each dosing strategy, is based on the current system state obtained in step 2. and the quality of each raw coal bunker A predicted output sequence can be obtained iteratively using the system dynamic model and the output model, that is, from the current time t to the future. At any given time, the current blending strategy is continuously executed to obtain the finished coal quality estimation sequence.

[0077] (5) First, from Select from the finished coal quality estimation sequences that are within the finished coal quality setpoint A sequence that fluctuates within an allowed range, i.e. Let ns be the number of sequences that meet the criteria after screening; then, calculate the relationship between each selected sequence and the set value of finished coal quality. The sum of absolute deviation values Secondly, calculate the corresponding product cost when performing the dosing ratio for each selected sequence. ,in Let the price of the j-th raw coal bunker in order i be denoted; then, calculate the overall target value. ,in The weighting coefficient is used to balance the control effect and product cost; finally, the dosing ratio with the smallest overall target value is selected as the decision result at time t.

[0078] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention. Equivalent transformations or substitutions made based on the above technical solutions all fall within the scope of protection of the claims of the present invention.

Claims

1. A smart coal blending method based on a predictive model, characterized in that, The method The process includes the following steps: Step 1: Data acquisition and preprocessing. Step 2: Model training. The model adopts an Encoder-Decoder structure, where the Encoder model is the state estimation model and the Decoder model is the system dynamics model. Step 3: Quality control of finished coal at export; specifically as follows: Based on the trained state estimation model and system dynamic model, determine the optimal blending ratio of each raw coal bin at time t, thereby achieving precise control of the quality of finished coal at export. The blending ratio of each raw coal bin meets certain constraints: Constraint (1): , Constraint (2): ， in and These represent the minimum and maximum blending ratios of the j-th raw coal silo in order i, respectively; Step 3 also includes the following constraints: Constraint (3): When When the quality of the finished coal exported is less than the requirements of order i, the minimum blending ratio of the best quality raw coal is the current blending ratio of that type of coal, i.e. ; This represents the quality of the exported finished coal at time t in order i; when When the quality requirements of the exported finished coal exceed those of order i, the highest blending ratio of the best quality raw coal is the current blending ratio of that type of coal, i.e. The quality of the exported finished coal is obtained at time t. Then, the following steps are performed to determine the optimal blending ratio, thereby controlling the quality of the exported finished coal to meet customer requirements. (1) Construct a historical production sequence. The historical production sequence in production order i includes the blending ratio sequence of each raw coal bunker from time t-T1 to time t-1. The quality sequence of finished coal from time t-T1+1 to time t and quality data of each raw coal bunker Where T1 is the first preset duration, (2) Based on the constructed historical production sequence, the system state at the current moment is predicted using a state estimation model. ; (3) Generate blending strategies for several raw coal bunkers. First, determine the minimum blending ratio for each coal bunker. and the highest blending ratio Next, for raw coal bunkers 1 to N-1, within the interval We perform uniform sampling within the range j∈[1,N-1] to obtain m values; then, we arrange and combine the blending ratios of these N-1 coal bunkers to obtain... A blending strategy, wherein the blending ratio of the Nth raw coal bunker can be determined by... To obtain; ultimately, to be dissatisfied. Remove the d policies that have the conditions. (4) Based on A blending strategy to generate The predicted output sequence, for each blending ratio strategy, is based on the current system state obtained in step 2. and the quality of each raw coal bunker A predicted output sequence can be obtained iteratively using the system dynamic model and output model. That is, from the current time t to the future time t+T2, the current blending ratio strategy is continuously executed to obtain the finished coal quality estimation sequence. T1 represents the system state at time t in order i, and T2 represents the second preset duration. (5) First, from Select from the finished coal quality estimation sequences that are within the finished coal quality setpoint A sequence that fluctuates within an allowed range, i.e. Let ns be the number of sequences that meet the criteria after filtering. This represents the quality of the exported finished coal at time k in order i. Next, the quality set value of each selected sequence is calculated. The sum of absolute deviation values Secondly, calculate the corresponding product cost when performing the blending ratio for each selected sequence. ,in Let the price of the j-th raw coal bunker in order i be denoted; then, calculate the overall target value. , where α is the weighting coefficient used to balance the control effect and product cost; finally, the blending ratio corresponding to the minimum overall target value is selected as the decision result at time t.

2. The intelligent coal blending method based on a prediction model according to claim 1, characterized in that, Step 1: Data Acquisition and Preprocessing, specifically as follows: Step 1-1: Collect historical data of the intelligent coal blending process. The dataset consists of several complete coal blending orders, and the quality data of the raw coal bunker is recorded in each order. And production data that changes over time, with each time sample in the production data including the blending ratio. and the quality of exported finished coal ,in Let j represent the order number, j∈[1,N] represent the raw coal bunker number, and t represent the time. This represents the quality of the j-th raw coal bunker in order i. This represents the blending ratio of the j-th raw material coal bunker at time t in order i. This represents the quality of the exported finished coal at time t in order i; Steps 1-2: Data preprocessing. The exported finished coal quality data undergoes center-average filtering to eliminate measurement noise. The center-average filtering method is as follows: Where 2K+1 is the filter width, and K is the preset value. This represents the quality of the exported finished coal at time k in order i. This represents the quality of the exported finished coal after filtering at time t in order i.

3. The intelligent coal blending method based on a prediction model according to claim 2, characterized in that, Step 2: Model training is as follows: Step 2-1: The state estimation model predicts the current state of the system based on the system's most recent input and output history sequences, and its form is as follows: in Let this be the system state at time t in order i. In order i Time's up The blending ratio sequence data of the j-th raw coal bunker at time j. This represents the quality of the j-th raw coal bunker in order i. This represents the sequence data of finished coal quality from time t-T1+1 to time t in order i. Let θ be the first preset duration, and let θ be the parameters of the state estimation model. Step 2-2: System Dynamic Model. The system dynamic model consists of two sub-models: a state transition model and an output model. The state transition model describes the state transition characteristics of the system and takes the following form: , in, Let this be the system state at time t in order i. Let i be the blending ratio of the j-th raw material coal bunker at time t in order i. Let be the mass of the j-th raw coal bunker in order i, and φ be the parameters of the state transition model. The output model maps the system state to the controlled output value, and its form is as follows: ， in Let this be the system state at time t in order i. This represents the system output at time t in order i, i.e., the predicted quality of the finished coal. These are the parameters for the output model.

4. An electronic device, characterized in that, include: A processor, and a memory for storing processor-executable instructions; The processor is configured to execute the instructions to implement the intelligent coal blending method based on a predictive model as described in any one of claims 1 to 3.

5. A computer-readable storage medium storing computer instructions thereon, characterized in that: When executed by a processor, the computer instructions implement the intelligent coal blending method based on a predictive model as described in any one of claims 1 to 3.

Images