Coal dressing decision-making method and system based on multi-source data integration

Abstract

The invention provides a coal dressing decision-making method and system based on multi-source data integration, relates to the technical field of coal washing and processing, and aims to solve the technical problems of low coal dressing efficiency and the like caused by insufficient data utilization and lack of scientificity and self-adaptive ability in the existing coal dressing decision-making method. According to the method, firstly, a multi-source coal dressing data set containing coal quality detection, equipment operation parameters and market dynamic indexes is obtained; performing standardization preprocessing on the coal quality detection data sequence, extracting coal quality characteristic groups such as ash distribution, and extracting equipment operation state characteristic groups such as separation equipment current change from the equipment operation parameter sequence; performing multi-dimensional feature fusion on the feature group and the market dynamic index sequence based on a preset deep neural network model to generate a coal dressing process parameter adjustment strategy combination; and finally, according to the strategy combination, the coal preparation equipment is controlled to execute operations such as separation density optimization, and the deep neural network model is updated according to real-time separation effect feedback.

Description

Technical Field

[0001] This application relates to the technical field of coal washing and processing, and particularly to a coal preparation decision-making method and system based on multi-source data integration. Background Art

[0002] In the field of intelligent coal washing and processing, traditional coal preparation decision-making methods mainly rely on manual experience for decision-making. In terms of data acquisition, the collection of coal preparation production data in the past was relatively one-sided. For example, regarding coal quality detection data, often only a small number of key indicators were obtained, lacking a comprehensive and detailed detection of coal quality. Moreover, the acquisition method was mostly periodic discrete sampling, which could not capture the dynamic information of coal quality changing over time in a timely manner. Key data such as the change trend of ash distribution, the real-time fluctuation of sulfur content, and the dynamic correlation between volatile matter and other components were difficult to effectively obtain, which made the overall evaluation of coal quality have serious defects. For equipment operation parameters, only some basic operation status parameters were concerned, such as whether the equipment was turned on and the operation duration, ignoring many key parameters that had an important impact on coal preparation effects, such as the current change of the separation equipment, vibration frequency, and medium flow rate, resulting in the inability to deeply understand the actual operation status of the equipment and the synergistic relationship between various parameters. At the same time, market dynamic data was almost completely ignored in the traditional coal preparation decision-making process, and important information such as market price fluctuations and diverse demands for coal product quality and specifications in different periods was not included in the decision-making consideration, making the coal preparation production seriously out of touch with the actual market demand.

[0003] In terms of data processing and analysis, traditional methods are extremely rough and simple. For the limited data obtained, mostly only simple statistical analyses are carried out, such as calculating the average value, maximum value, and minimum value, etc., and it is impossible to deeply explore the hidden features and laws behind the data. In coal quality analysis, it is impossible to extract the features that truly reflect coal quality characteristics and have guiding significance for coal preparation decision-making from numerous coal quality indicators. For equipment operation parameters, it is also impossible to effectively extract the key features that can characterize the good or bad operation status of the equipment and the impact on the coal preparation process, let alone making scientific decisions using these features.

[0004] In terms of coal preparation decision-making and process parameter adjustment, the traditional decision-making process mainly relies on the personal experience of operators, lacking scientific theoretical basis and systematic method support. Process parameter adjustment is usually based on pre-set fixed rules, which are often formulated under specific conditions and lack flexibility and adaptability. When the coal quality changes, the equipment performance fluctuates, or the market demand changes, this decision-making and parameter adjustment method based on fixed rules cannot make reasonable decisions and effective parameter adjustments in a timely manner according to the actual situation, resulting in low coal preparation efficiency and unstable product quality. Summary of the Invention

[0005] In view of the above, to at least partially address the deficiencies in the prior art, in a first aspect, an embodiment of the present application provides a coal preparation decision-making method based on multi-source data integration, and the method includes: Obtain a multi-source coal preparation data set during the coal preparation production process, where the multi-source coal preparation data set includes a coal quality detection data sequence, an equipment operation parameter sequence, and a market dynamic index sequence; Perform standardized preprocessing on the coal quality detection data sequence to generate a standardized coal quality data sequence, and extract a coal quality feature group based on the standardized coal quality data sequence. The coal quality feature group includes an ash content distribution feature, a sulfur content fluctuation feature, and a volatile matter correlation feature; Extract time-series features from the equipment operation parameter sequence to generate an equipment operation state feature group. The equipment operation state feature group includes a current change feature of the separation equipment, a vibration frequency correlation feature, and a medium flow rate matching feature; Based on a preset deep neural network model, perform multi-dimensional feature fusion processing on the coal quality feature group, the equipment operation state feature group, and the market dynamic index sequence, and generate a combination of coal preparation process parameter adjustment strategies; According to the combination of coal preparation process parameter adjustment strategies, control the coal preparation equipment to perform sorting density optimization operations, medium flow rate adjustment operations, and product particle size classification operations, and update the deep neural network model based on real-time sorting effect feedback.

[0006] In a second aspect, an embodiment of the present application further provides a coal preparation decision-making system based on multi-source data integration, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions, or codes, and the processor is used to execute the programs, instructions, or codes in the machine-readable storage medium to implement the above-mentioned coal preparation decision-making method based on multi-source data integration.

[0007] In summary, the coal preparation decision-making method and system based on multi-source data integration provided by the embodiments of the present application achieve the intelligence and high efficiency of the coal preparation process, and can improve the coal preparation efficiency and the stability of product quality. Specifically, by integrating the dynamic correlation processing mechanism of multi-source data, different types of data such as coal quality detection data, equipment operation parameters, and market dynamic indicators are deeply fused to form a characterization method that can comprehensively and accurately reflect the characteristics of the coal preparation process, no longer limited to the simple analysis of a single data type by traditional methods, and comprehensively capture the complex relationships of various factors in the coal preparation process. On this basis, the preset deep neural network model uses advanced machine learning algorithms to deeply mine the process feature set, realizing the accurate prediction of key parameters in coal preparation, and changing the uncertainty of relying on empirical judgment in the past. The generated parameter set not only has the ability to quickly respond to changes in the coal preparation process, but also interacts closely with the production control unit through a closed-loop feedback mechanism, enabling the generation and execution of instructions to have an adaptive characteristic, being able to dynamically adjust according to real-time production conditions, making the coal quality optimization process successfully transform from a traditional experience-dependent type to a scientific model prediction type, thereby significantly improving the accuracy and timeliness of parameter adjustment, effectively reducing the quality fluctuations in the coal preparation process, and improving production efficiency.

[0008] Furthermore, through the dynamic update mechanism of the parameter weights of the deep neural network model by real-time sorting effect feedback (which can correspond to quality indicators), a control system with the ability of continuous evolution is constructed, enabling the model to automatically correct the decision-making logic according to changing production conditions, and ensuring stable and efficient performance in a complex and changeable industrial production environment.

[0009] Other features and advantages of the present application will be described in detail in the subsequent specific implementation part. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings based on the above drawings without creative efforts.

[0011] In order to more completely understand the present application and its beneficial effects, the following description will be made in conjunction with the drawings, where the same reference numerals in the following description represent the same parts.

[0012] Figure 1 It is a schematic flow chart of a coal preparation decision-making method based on multi-source data integration provided by the embodiments of the present application.

[0013] Figure 2It is a schematic diagram of the hardware environment of a coal preparation decision-making system based on multi-source data integration provided by an embodiment of the present application. Detailed implementation manners

[0014] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present application.

[0015] The present application will be specifically described below with reference to the accompanying drawings of the specification. Figure 1 It is a schematic flowchart of a coal preparation decision-making method based on multi-source data integration provided by an embodiment of the present application. The coal preparation decision-making method based on multi-source data integration includes the following steps S110-S150, which will be introduced in detail below.

[0016] Step S110: Obtain a multi-source coal preparation data set in the coal preparation production process. The multi-source coal preparation data set includes a coal quality detection data sequence, an equipment operation parameter sequence, and a market dynamic index sequence.

[0017] In this embodiment, taking the production and manufacturing scenario of a coal preparation plant as an example, the production system of the coal preparation plant can continuously collect various types of data. For example, the coal quality detection data sequence can be obtained by comprehensively detecting different batches of coal samples. Suppose different batches of coal samples come from different mining areas of the same coal mine. Such as coal samples A1, A2, A3, etc., corresponding to different detection time points t1, t2, t3 respectively. The coal quality detection data sequence contains the detection values of each coal sample at each time point regarding ash content, sulfur content, volatile matter, etc. The equipment operation parameter sequence includes records of the operating states of various equipment in the coal preparation plant. For example, for sorting equipment B1, B2, B3, parameters such as current values, vibration frequency values, and medium flow velocity values at each moment t1, t2, t3, etc. constitute this sequence. The market dynamic index sequence is a set of information related to the coal preparation product market. For example, the clean coal price trend index includes a data sequence composed of price change rates in a series of different time periods. Another example is that the market demand prediction index may be a set of predicted values of demand in different future time intervals based on past sales data and market research. The inventory turnover rate index is a set of values reflecting the inventory turnover situation calculated from data such as inventory quantity and sales quantity. Through the above methods, data in multiple aspects such as coal quality, equipment operation, and market dynamics involved in the coal preparation production process can be comprehensively collected, providing a data basis for subsequent analysis and decision-making.

[0018] Step S120: Perform standardized preprocessing on the coal quality detection data sequence to generate a standardized coal quality data sequence, and extract a coal quality feature group based on the standardized coal quality data sequence. The coal quality feature group includes an ash distribution feature, a sulfur content fluctuation feature, and a volatile matter correlation feature.

[0019] Among them, the standardized preprocessing of the obtained coal quality detection data sequence in step S120 can be realized through the following steps S121 - S129, which are described as follows.

[0020] Step S121: Perform data format conversion processing on the original ash detection value, sulfur content detection value, and volatile matter detection value in the coal quality detection data sequence to generate an ash standardized sequence, a sulfur content standardized sequence, and a volatile matter standardized sequence with a unified dimension.

[0021] In this embodiment, since the original coal quality detection data may come from different detection devices or follow different detection standards, the dimensions of the ash, sulfur content, and volatile matter detection values may be inconsistent. For the consistency of subsequent processing, data format conversion is required. Assume there is a set of standard dimension systems. Convert all ash detection values to this standard dimension to generate an ash standardized sequence. For example, convert ash detection values expressed in different units through specific conversion coefficients so that all ash values in this sequence have the same dimension. Similarly, perform similar operations on the sulfur content detection value and the volatile matter detection value to generate a sulfur content standardized sequence and a volatile matter standardized sequence. Through the above method, different types of coal quality detection data can be unified in dimension, facilitating subsequent data processing and analysis.

[0022] Step S122: Perform missing value filling processing on the ash standardized sequence, and generate a complete ash continuous sequence based on the linear interpolation algorithm of adjacent coal sample detection values.

[0023] Among them, in the ash normalization sequence, there may be missing values due to detection link failures or other reasons. For example, in the detection sequence of coal samples D1, D2, D3, and D4, the ash detection value corresponding to coal sample D3 is missing. At this time, by traversing the ash normalization sequence, the time stamp interval where the missing value is located is identified, and the time stamp corresponding to coal sample D3 is found. Then, the previous valid ash detection value in the time stamp interval where the missing value is located is obtained, such as the ash value of coal sample D2, and the next valid ash detection value, such as the ash value of coal sample D4. The time stamp difference between the two valid ash detection values of D2 and D4. Assume that the time stamp of coal sample D2 is t6 and the time stamp of coal sample D4 is t8, and the difference is t8 - t6. According to this time stamp difference, linear interpolation weight coefficients are generated. For example, the weight coefficients can be (t8 - the current missing value time stamp) / (t8 - t6) and (the current missing value time stamp - t6) / (t8 - t6). Based on these two linear interpolation weight coefficients, the previous valid ash detection value and the next valid ash detection value are weighted and summed, that is, ((t8 - the current missing value time stamp) / (t8 - t6)) × the ash value of coal sample D2 + ((the current missing value time stamp - t6) / (t8 - t6)) × the ash value of coal sample D4, to generate the filled ash value at the time stamp where the missing value is located. The filled ash value is inserted into the corresponding time stamp position of the ash normalization sequence to generate a continuous ash sequence without missing values. After that, a dimensional consistency check is performed on this continuous ash sequence to ensure that the units of all ash values in this sequence are consistent with the sulfur value units of the sulfur normalization sequence and the volatile value units of the volatile normalization sequence under the preset dimensional conversion rules.

[0024] In another alternative embodiment, step S122 may include the following steps S1221-S1226.

[0025] Step S1221: Traverse each coal sample detection time stamp in the ash normalization sequence to identify the time stamp interval where the missing value is located.

[0026] In this embodiment, the detection time stamps corresponding to each coal sample in the ash normalization sequence can be viewed in chronological order. If it is found that the ash value between time stamps T1 and T2 is missing, that interval is determined as the time stamp interval where the missing value is located. In the above manner, the position of the missing data in the ash normalization sequence can be accurately found, preparing for filling the missing value later.

[0027] Step S1222: Obtain the previous valid ash detection value and the next valid ash detection value in the time stamp interval where the missing value is located.

[0028] Among them, based on the time stamp interval where the missing value is located as found above, assume that the previous valid ash detection value is A1, corresponding to the time stamp T0, and the next valid ash detection value is A2, corresponding to the time stamp T3. Locate these two valid ash detection values by positioning the time stamp. Through the above method, relevant data for linear interpolation calculation can be obtained, providing a necessary condition for filling the missing value.

[0029] Step S1223: Calculate the time stamp difference between the previous valid ash detection value and the next valid ash detection value, and generate a linear interpolation weight coefficient according to the time stamp difference.

[0030] For example, the difference between time stamps T3 and T0 can be calculated, denoted as ΔT, and a linear interpolation weight coefficient is generated according to the proportional relationship between the time stamp difference and the time length of the interval where the missing value is located. For example, if the interval where the missing value is located is from T1 to T2, and the time length of this interval is Δt, then the linear interpolation weight coefficient W1 = (T3 - T1) / ΔT, W2 = (T2 - T0) / ΔT. Through the above method, the weight coefficient for weighted calculation of the missing value can be obtained, ensuring the rationality of the filled value.

[0031] Step S1224: Based on the linear interpolation weight coefficient, perform weighted summation on the previous valid ash detection value and the next valid ash detection value to generate the filled ash value at the time stamp where the missing value is located.

[0032] Among them, for example, according to the linear interpolation weight coefficient generated previously, calculate the filled ash value F = W1 * A1 + W2 * A2, so as to obtain the filled ash value at the time stamp where the missing value is located. In this way, the missing ash value can be reasonably filled using the existing data, making the ash sequence more complete.

[0033] Step S1225: Insert the filled ash value into the corresponding time stamp position of the ash normalization sequence to generate a continuous ash sequence without missing values.

[0034] Insert the calculated filled ash value into the time stamp position corresponding to the missing value in the ash normalization sequence. For example, insert the filled ash value between time stamps T1 and T2, so as to form a continuous ash sequence without missing values. In this way, the ash normalization sequence can be made complete, meeting the requirements for data integrity in subsequent analysis.

[0035] Step S1226: Perform a dimensional consistency check on the continuous ash sequence to ensure that the units of all ash values in the continuous ash sequence are consistent with the units of the sulfur values in the sulfur normalization sequence and the units of the volatile matter values in the volatile matter normalization sequence under the preset dimensional conversion rules.

[0036] In this embodiment, according to the preset dimension conversion rule, it is checked whether the units of all ash values in the continuous ash sequence are consistent with the sulfur value unit of the sulfur standardized sequence and the volatile matter value unit of the volatile matter standardized sequence. In this way, the dimensional consistency between different coal quality data sequences can be ensured, and subsequent analysis errors caused by dimensional problems can be avoided.

[0037] Step S123: Perform outlier correction processing on the sulfur standardized sequence, truncate and replace the sulfur detection values that exceed the sulfur threshold interval based on the preset sulfur threshold interval, and generate a corrected sulfur sequence.

[0038] In this embodiment, a reasonable sulfur threshold interval can be preset. For example, the lower limit is L1 and the upper limit is U1. For each sulfur detection value in the sulfur standardized sequence, such as the sulfur detection values corresponding to coal samples E1, E2, E3, etc., assuming that the sulfur detection value of coal sample E2 is S2, if S2 is less than L1 or greater than U1, it is determined as an outlier. For the case where it is less than L1, it is truncated and replaced with L1; for the case where it is greater than U1, it is truncated and replaced with U1, thereby generating a corrected sulfur sequence. In this way, the outliers in the sulfur standardized sequence can be removed, making the data more reliable and facilitating the subsequent accurate analysis of sulfur-related characteristics.

[0039] Step S124: Perform noise filtering processing on the volatile matter standardized sequence, and smooth the volatile matter standardized sequence using the sliding window mean algorithm to generate a denoised volatile matter sequence.

[0040] For example, a sliding window size of n can be set, assuming n is 5. For a group of volatile matter detection values starting from coal sample F1 in the volatile matter standardized sequence, first take the first 5 values, that is, the volatile matter detection values of coal samples F1, F2, F3, F4, and F5, calculate their average value, and use this average value as the denoised volatile matter value of coal sample F3. Then the sliding window moves one position backward, take the volatile matter detection values of coal samples F2, F3, F4, F5, and F6, calculate the average value as the denoised volatile matter value of coal sample F4, and so on, process the entire volatile matter standardized sequence to generate a denoised volatile matter sequence. In this way, the volatile matter standardized sequence can be smoothed, the noise interference in it can be removed, and a data sequence that can better reflect the true change trend of the volatile matter can be obtained.

[0041] Step S125: Merge the continuous ash sequence, the corrected sulfur sequence, and the denoised volatile matter sequence into a standardized coal quality data sequence, and store the standardized coal quality data sequence in a distributed database.

[0042] In this embodiment, the processed ash continuous sequence, corrected sulfur content sequence, and denoised volatile matter sequence can be merged according to the time stamp of the coal sample or other preset corresponding relationships. For example, the ash value, sulfur value, and volatile matter value of coal sample G1 at time point t9 are respectively from the values at the corresponding time points in the ash continuous sequence, corrected sulfur content sequence, and denoised volatile matter sequence and combined together to form a record, thus constituting a standardized coal quality data sequence. Then, this standardized coal quality data sequence is stored in a distributed database for subsequent ready access. Among them, the distributed database can be a cloud storage space built on a cloud server for storing coal mine big data to facilitate intelligent analysis of coal mine production. In this way, the integrated and processed coal quality data is stored safely and reliably, providing a unified data source for further extraction of coal quality characteristics and coal preparation decision-making.

[0043] Step S126: Perform frequency domain transformation processing on the ash continuous sequence to generate an ash frequency spectrum distribution feature, and extract the frequency band whose main frequency amplitude in the ash frequency spectrum distribution feature exceeds a preset ash amplitude threshold as the ash distribution feature.

[0044] As an example, the method for performing frequency domain transformation on the ash continuous sequence can be the Fourier transform method, which converts the ash continuous sequence in the time domain to the frequency domain to obtain the ash frequency spectrum distribution feature, which contains multi-dimensional results of different frequency components and their corresponding amplitudes. For example, in some scenarios, such as during the coal preparation process, the ash content shows periodic fluctuations due to reasons such as equipment vibration and raw material batch differences. The Fourier transform method can effectively capture the dominant frequency components. In a possible implementation manner, assuming that the preset ash amplitude threshold is T1, in the obtained ash frequency spectrum distribution feature, search for the frequency bands whose main frequency amplitudes exceed T1, and mark these frequency bands as the ash distribution feature. For example, after frequency domain transformation, frequencies f1 - f10 and their corresponding amplitudes A1 - A10 are obtained. If A3, A5, and A7 exceed T1, the corresponding frequency ranges constitute the ash distribution feature.

[0045] Step S127: Perform fluctuation period analysis on the corrected sulfur content sequence, calculate the absolute value sequence of the differences between adjacent sulfur content detection values, and perform moving average processing on the absolute value sequence of the differences to generate a sulfur content fluctuation intensity sequence, and mark the intervals in the sulfur content fluctuation intensity sequence that exceed the preset sulfur content fluctuation threshold as sulfur content fluctuation features.

[0046] In this embodiment, for the corrected sulfur content sequence, the absolute value of the difference between adjacent sulfur content detection values can be calculated in sequence. For example, for the sulfur content detection values S_H1, S_H2, and S_H3 of coal samples H1, H2, and H3, calculate |S_H2 - S_H1|, |S_H3 - S_H2|, etc. to form a sequence of absolute difference values. Then, perform a moving average process on this sequence of absolute difference values. Assume that the moving average window size is m. If m is 3, first take the first 3 absolute difference values, calculate the average value as the first sulfur content fluctuation intensity value after moving average. Move the window one position backward, then take 3 values and calculate the average value, and so on, to generate a sulfur content fluctuation intensity sequence. In addition, a sulfur content fluctuation threshold T2 can be preset in advance, and the intervals in the sulfur content fluctuation intensity sequence that exceed T2 are marked as sulfur content fluctuation characteristics, so as to fully reflect the stability of sulfur content in coal quality.

[0047] Step S128: Perform a correlation analysis on the denoised volatile matter sequence and the ash content continuous sequence, calculate the Pearson correlation coefficient between the volatile matter detection value and the corresponding ash content detection value, and mark the interval where the Pearson correlation coefficient is lower than the preset correlation threshold as the volatile matter association characteristic.

[0048] In this embodiment, for the denoised volatile matter sequence and the ash content continuous sequence, according to the same time stamp or the corresponding relationship of coal samples, calculate the Pearson correlation coefficient for each pair of volatile matter detection values and ash content detection values. For example, the volatile matter detection value of coal sample I1 at time point t10 is V1, and the ash content detection value is A1. The volatile matter detection value of coal sample I2 at time point t11 is V2, and the ash content detection value is A2, and so on, to calculate the Pearson correlation coefficients of multiple groups of data. Preset a correlation threshold T3, and mark the interval where the Pearson correlation coefficient is lower than T3 as the volatile matter association characteristic. In this way, the correlation relationship between the volatile matter and the ash content can be analyzed, and the volatile matter association characteristic can be extracted to provide a reference for the coal preparation process.

[0049] Step S129: Associate the ash content distribution characteristic, the sulfur content fluctuation characteristic, and the volatile matter association characteristic with the corresponding coal sample batch number to generate the coal quality characteristic group.

[0050] In this embodiment, each coal sample can have a corresponding batch number. For example, coal samples J1, J2, and J3 belong to batch K1. Associate the previously obtained ash content distribution characteristic, sulfur content fluctuation characteristic, and volatile matter association characteristic with the corresponding coal sample batch number. For example, the ash content distribution characteristic, sulfur content fluctuation characteristic, and volatile matter association characteristic corresponding to the coal samples in batch K1 are combined together to form a coal quality characteristic group, and then different coal quality characteristics are corresponded to specific coal sample batches, which is convenient for subsequent coal preparation decision-making analysis based on the coal sample batches.

[0051] Step S130: Extract the temporal characteristics from the sequence of the device operation parameters to generate a set of device operation state characteristics, where the set of device operation state characteristics includes the current change characteristics of the sorting device, the vibration frequency correlation characteristics, and the medium flow rate matching characteristics.

[0052] Among them, the process of extracting the temporal characteristics and other processing from the sequence of the processing device operation parameters in step S130 may include the following steps S131 - S134, which will be introduced in detail below.

[0053] Step S131: Perform trend decomposition processing on the current time - series data of the sorting device in the sequence of the device operation parameters, separate the long - term trend component of the current, the periodic fluctuation component, and the residual component, and mark the interval where the slope of the long - term trend component of the current exceeds the preset slope threshold as the current change characteristics.

[0054] In this embodiment, for the current time - series data of the sorting device, assume that there is a series of current values I1 - I9 at time points t12 - t20. An appropriate trend decomposition method is adopted, such as decomposing it into a long - term trend component of the current, which reflects the overall change trend of the current in a relatively long time period, a periodic fluctuation component reflecting the periodic current change, and a residual component which is the remaining part after removing the long - term trend and periodic fluctuation. For example, a slope threshold T4 can be preset, and the slope of the long - term trend component of the current is calculated. For two adjacent points on the long - term trend component of the current, such as points (t13, I13_long) and (t14, I14_long), the slope is (I14_long - I13_long) / (t14 - t13), and the interval where the slope exceeds T4 is marked as the current change characteristics. Through the above method, the characteristics reflecting the current change trend can be extracted from the current time - series data, which helps to understand the change of the operation state of the sorting device.

[0055] Step S132: Perform joint analysis on the vibration frequency time - series data and the medium flow rate time - series data, calculate the covariance matrix of the mean value of the vibration frequency and the mean value of the medium flow rate within the same time window, and mark the window where the elements on the main diagonal of the covariance matrix exceed the preset covariance threshold as the vibration frequency correlation characteristics.

[0056] For example, set a time window size of p, and assume p is 4. For the vibration frequency time series data and the medium flow velocity time series data, within each time window, starting from time point t21, take the vibration frequency values f1 - f4 and the medium flow velocity values v1 - v4 from t21 - t24, and calculate the average vibration frequency (f1 + f2 + f3 + f4) / 4 and the average medium flow velocity (v1 + v2 + v3 + v4) / 4 respectively. Then calculate the covariance matrix of these two averages. Preset a covariance threshold T5, and in the covariance matrix, check the elements on the main diagonal, and mark the windows exceeding T5 as vibration frequency correlation features.

[0057] Step S133: Detect the mutation points of the medium flow velocity time series data, calculate the cumulative deviation amount of the medium flow velocity sequence based on the cumulative sum algorithm, and obtain the medium flow velocity matching feature according to the moment when the cumulative deviation amount exceeds the preset deviation threshold.

[0058] Among them, first extract the corresponding sequence of timestamps and medium flow velocity values in the medium flow velocity time series data. Assume that starting from time point t25, there are medium flow velocity values v5 - v10, and generate a medium flow velocity change curve in timestamp order. Calculate the global average of this medium flow velocity change curve, for example, (v5 + v6 + v7 + v8 + v9 + v10) / 6, and generate the medium flow velocity reference value corresponding to each timestamp based on this global average. Accumulate the difference between the actual medium flow velocity value and the medium flow velocity reference value at each timestamp. For example, for time point t25, the difference is v5 - reference value, and start accumulating from t25 to generate a cumulative deviation amount sequence. Monitor the change trend of the cumulative deviation amount sequence in real time, preset a deviation threshold T6, and when the cumulative deviation amount exceeds T6, record the current timestamp as the candidate moment of the mutation point. Compare the average values within the front and rear time windows of the actual medium flow velocity value at the candidate moment. Assume that the sizes of the front and rear time windows are both q, and q is 2. Calculate the average value within the previous time window, such as the average medium flow velocity at the two time points before the candidate moment, and the average value within the subsequent time window. If the differences between the actual medium flow velocity value and the average value within the previous time window and the average value within the subsequent time window both exceed the flow velocity fluctuation tolerance T7, then confirm that the candidate moment of the mutation point is an effective medium flow velocity mutation point. Associate the timestamp of the effective medium flow velocity mutation point and the corresponding medium flow velocity change amplitude with the sorting device number in the device operation parameter sequence to generate the medium flow velocity matching feature.

[0059] Specifically, the above step S133 may include the following steps S1331 - S1336. Step S1331: Extract the corresponding sequence of timestamps and medium flow velocity values from the medium flow velocity time series data, and generate a medium flow velocity change curve in the order of timestamps.

[0060] In this embodiment, each timestamp and its corresponding medium flow velocity value can be taken out from the medium flow velocity time series data, arranged in ascending order of timestamps to form a sequence, and a medium flow velocity change curve can be drawn accordingly. In the above manner, the change of the medium flow velocity over time can be visually presented, preparing for subsequent mutation point detection.

[0061] Step S1332: Calculate the global mean of the medium flow velocity change curve, and generate a medium flow velocity reference value corresponding to each timestamp based on the global mean.

[0062] In this embodiment, the global mean can be obtained by calculating the average value of all medium flow velocity values on the medium flow velocity change curve. For each timestamp, this global mean is used as the medium flow velocity reference value corresponding to that timestamp. In the above manner, a reference value can be provided for determining whether the medium flow velocity has mutated.

[0063] Step S1333: Perform an accumulative calculation on the difference between the actual medium flow velocity value of each timestamp and the medium flow velocity reference value to generate an accumulative deviation amount sequence.

[0064] In this embodiment, for each timestamp, the difference between its actual medium flow velocity value and the medium flow velocity reference value can be calculated, and then these differences are accumulated in sequence to form an accumulative deviation amount sequence. In this way, the accumulative change of the medium flow velocity relative to the reference value can be quantified, facilitating the discovery of sudden changes in the flow velocity.

[0065] Step S1334: Real-time monitor the change trend of the accumulative deviation amount sequence. When the accumulative deviation amount exceeds the preset deviation threshold, record the current timestamp as the candidate moment of the mutation point.

[0066] In this embodiment, the accumulative deviation amount sequence can be continuously observed, and a preset deviation threshold is set, such as the threshold is D1. When the accumulative deviation amount is greater than D1, record the timestamp at this time as the candidate moment of the mutation point. In this way, possible mutation points of the medium flow velocity can be preliminarily screened, preparing for further confirmation later.

[0067] Step S1335: Compare the mean values within the front and rear time windows of the actual medium flow velocity value at the candidate moment. If the differences between the actual medium flow velocity value and the mean value of the previous time window and the mean value of the subsequent time window both exceed the flow velocity fluctuation tolerance, confirm that the candidate moment of the mutation point is an effective medium flow velocity mutation point.

[0068] In this embodiment, for the candidate moments of mutation points, front and rear time windows can be set, for example, both the front and rear time windows are 5 minutes. Calculate the average values of the medium flow rates within the front and rear time windows respectively. For example, they can be set as M1 and M2 respectively, and then calculate the differences between the actual value of the medium flow rate at the candidate moment and M1, M2. Then, a flow rate fluctuation tolerance can be set, such as the tolerance is E1. If both of these differences are greater than E1, then confirm that this candidate moment is an effective mutation point of the medium flow rate. Through the above method, the mutation points of the medium flow rate can be accurately judged, and the accuracy of mutation point detection can be improved.

[0069] Step S1336: Associate the time stamp of the effective mutation point of the medium flow rate and the corresponding change amplitude of the medium flow rate with the sorting equipment number in the equipment operation parameter sequence to generate a medium flow rate matching feature.

[0070] In this embodiment, associate the time stamp of the confirmed effective mutation point of the medium flow rate and the change amplitude of the medium flow rate at this point with the corresponding sorting equipment number in the equipment operation parameter sequence to form a medium flow rate matching feature. Through the above method, the equipment to which the mutation point belongs can be determined, providing specific information for the analysis of the operation state of the coal preparation equipment.

[0071] Step S134: Align the current change feature, the vibration frequency correlation feature, and the medium flow rate matching feature according to the time stamp to generate a set of equipment operation state features.

[0072] In this embodiment, based on the time stamp, perform alignment operations on the current change feature, the vibration frequency correlation feature, and the medium flow rate matching feature extracted previously. For example, for a specific time stamp T, combine the three corresponding features at this time stamp to form a set of equipment operation state features. Through the above method, different features during the equipment operation process can be integrated, comprehensively reflecting the equipment operation state, and providing a basis for coal preparation decision-making in terms of equipment operation.

[0073] Step S140: Based on a preset deep neural network model, perform multi-dimensional feature fusion processing on the coal quality feature group, the equipment operation state feature group, and the market dynamic index sequence to generate a combination of coal preparation process parameter adjustment strategies.

[0074] Among them, step S140 may include the following steps S141 - S145.

[0075] Step S141: Map the ash distribution feature, the sulfur content fluctuation feature, and the volatile matter correlation feature in the coal quality feature group into a first feature vector.

[0076] In this embodiment, according to a specific mapping rule, the ash distribution feature, the sulfur fluctuation feature and the volatile matter correlation feature are respectively converted into vector form, and then spliced ​​together in a certain order to form a first feature vector. For example, the ash distribution feature contains frequency band information, which is converted into vector A, the sulfur fluctuation feature is converted into vector B, and the volatile matter correlation feature is converted into vector C. The first feature vector is formed by splicing A, B, and C in sequence. In the above manner, the coal quality feature group can be converted into a vector form suitable for processing by a deep neural network model, which is convenient for subsequent feature fusion.

[0077] Step S142: mapping the current change feature, the vibration frequency correlation feature and the medium flow rate matching feature in the equipment operation status feature group into a second feature vector.

[0078] In this embodiment, similar to the generation method of the first feature vector, according to a specific mapping rule, the current change feature, the vibration frequency correlation feature and the medium flow rate matching feature are respectively converted into vector forms, and then sequentially spliced ​​into a second feature vector. For example, the current change feature is converted into vector D, the vibration frequency correlation feature is converted into vector E, and the medium flow rate matching feature is converted into vector F. The second feature vector is composed of D, E, and F spliced ​​in sequence. In the above manner, the device operation status feature group is converted into a vector form suitable for deep neural network model processing, laying the foundation for multi-dimensional feature fusion.

[0079] Step S143: Mapping the clean coal price trend index, the market demand forecast index and the inventory turnover rate index in the market dynamic index sequence into a third eigenvector.

[0080] In this embodiment, the clean coal price trend index, market demand forecast index and inventory turnover rate index in the market dynamic index sequence are also processed according to specific mapping rules. Assuming that the clean coal price trend index contains multi-dimensional information such as price increase or decrease trend information and change amplitude, it is converted into vector G according to the rules; the market demand forecast index may involve information such as predicted demand in different time periods, which is converted into vector H; the inventory turnover rate index contains relevant information such as inventory turnover times, which is converted into vector I. Then, vectors G, H, and I are concatenated in sequence to form a third eigenvector. In this way, the market dynamic indicator sequence is converted into a form similar to other related eigenvectors, which is convenient for fusion processing in the deep neural network model.

[0081] Step S144: concatenate the first feature vector, the second feature vector and the third feature vector through the input layer of the deep neural network model to generate a fused feature vector.

[0082] In this embodiment, the input layer of the deep neural network model receives the first feature vector, the second feature vector, and the third feature vector. According to the preset splicing method of the model, these three vectors are spliced end to end. For example, the dimensions of the first feature vector are arranged in sequence first, then the dimensions of the second feature vector are connected, and finally the dimensions of the third feature vector are connected, so as to generate a fused feature vector containing coal quality features, equipment operation status features, and market dynamic index features. This fused feature vector integrates information from multiple aspects and provides a comprehensive data basis for further processing inside the deep neural network model subsequently. Through the above method, feature vectors from different sources can be effectively combined together, enabling the model to comprehensively consider various factors for analysis.

[0083] Step S145: Perform a non-linear transformation on the fused feature vector through the hidden layer of the deep neural network model, output the separation density adjustment amount, the medium flow rate adjustment coefficient, and the particle size classification threshold, and generate a combined coal preparation process parameter adjustment strategy.

[0084] Among them, step S145 may include the following steps S1451 - S1458, which are specifically introduced as follows.

[0085] Step S1451: Input the fused feature vector into the first hidden layer of the deep neural network model, and perform a non-linear mapping on the fused feature vector through the activation function to generate a first intermediate feature vector.

[0086] In this embodiment, when the fused feature vector enters the first hidden layer, the neurons in the hidden layer process each dimension of the fused feature vector according to the activation function. The activation function will output a new value according to a specific non-linear rule based on the magnitude of the input value. For example, for each dimension value X in the fused feature vector, the activation function may, according to a certain rule, such as when X is greater than a certain set value, output a relatively large positive value; when X is less than another set value, output a relatively small negative value or zero, etc. After the activation function processes all dimensions of the fused feature vector, a first intermediate feature vector is generated. This first intermediate feature vector is no longer a simple combination of original features, but after non-linear transformation, it may extract more representative feature information. Through the above method, using the non-linear characteristics of the activation function to perform a preliminary transformation on the fused feature vector can prepare for further extraction of key features subsequently.

[0087] Step S1452: Input the first intermediate feature vector into the second hidden layer of the deep neural network model, perform a feature similarity match with the preset historical adjustment records of process parameters, and screen out the feature dimensions corresponding to the historical adjustment records with a similarity higher than the preset matching threshold.

[0088] In this embodiment, the second hidden layer compares the first intermediate feature vector with the preset historical adjustment records of process parameters. For each dimension of the first intermediate feature vector and the corresponding feature dimensions in the historical adjustment records of process parameters, a specific similarity calculation method is used to measure the similarity between them. For example, a feature distance metric method may be adopted to calculate the distance between two feature dimensions, and the smaller the distance, the higher the similarity. A preset matching threshold is set. For each dimension, if its similarity with the corresponding dimension in the historical adjustment record is higher than this threshold, the feature dimension corresponding to this historical adjustment record is screened out. These screened feature dimensions may contain the key feature information when the process parameters were successfully adjusted in the past. In the above manner, the part with a higher similarity to the current feature vector is screened out from historical experience, providing a reference basis for generating an appropriate process parameter adjustment strategy.

[0089] Step S1453: Reassign weights to the first intermediate feature vector based on the screened feature dimensions to generate a second intermediate feature vector.

[0090] In this embodiment, for the screened feature dimensions, weights are reassigned to each dimension of the first intermediate feature vector according to their relevance or importance to the current situation. For example, for a certain screened feature dimension, if it has been proven to be very critical in past cases of successfully adjusting process parameters, a larger weight is assigned to the corresponding dimension in the first intermediate feature vector; conversely, if a dimension is relatively less important, a smaller weight is assigned. By this way of weight reassignment, the first intermediate feature vector is adjusted to generate a second intermediate feature vector. This second intermediate feature vector highlights more prominently the important features related to the current coal preparation process parameter adjustment, helping the model to generate adjustment strategies more accurately. At the same time, the weights of the feature vector are reasonably adjusted, enabling the model to better focus on the key features and improve the accuracy of generating adjustment strategies.

[0091] Step S1454: Input the second intermediate feature vector into the third hidden layer of the deep neural network model, and perform linear transformations through the separation density adjustment amount output channel, medium flow rate adjustment coefficient output channel, and particle size classification threshold output channel respectively to generate an initial separation density adjustment amount, an initial medium flow rate adjustment coefficient, and an initial particle size classification threshold.

[0092] In this embodiment, when the second intermediate feature vector enters the third hidden layer, it enters three different output channels respectively. In the sorting density adjustment amount output channel, according to the preset linear transformation rule of this channel, the second intermediate feature vector is processed. Suppose the linear transformation rule is to multiply each dimension of the second intermediate feature vector by a set of specific coefficients and then add them up to obtain a value, which is the initial sorting density adjustment amount. Similarly, in the medium flow rate adjustment coefficient output channel and the particle size classification threshold output channel, the second intermediate feature vector is processed in a similar manner according to their respective preset linear transformation rules, and the initial medium flow rate adjustment coefficient and the initial particle size classification threshold are obtained respectively. These initial values are the adjustment amounts of the coal preparation process parameters initially generated by the model, but may need further processing before being applied to actual production. Through the above method, by using the linear transformation of different output channels, initial adjustment values related to the coal preparation process parameters can be generated based on the second intermediate feature vector.

[0093] Step S1455: Perform unit dimension conversion on the initial sorting density adjustment amount to make it consistent with the density measurement unit of the current sorting equipment, and obtain the target sorting density adjustment amount.

[0094] In this embodiment, since the unit dimension of the initial sorting density adjustment amount may not be consistent with the density measurement unit of the current sorting equipment, conversion is required. Suppose the unit of the initial sorting density adjustment amount is a general density unit, while the current sorting equipment uses another specific density unit. By looking up the pre-set unit dimension conversion table or according to a specific conversion formula (the specific formula is not involved here, only for conceptual explanation), the value of the initial sorting density adjustment amount is calculated according to the conversion rule to obtain the target sorting density adjustment amount that is consistent with the density measurement unit of the current sorting equipment. For example, if there is a conversion ratio relationship between the initial unit and the target unit, multiply or divide the initial value by this ratio value to obtain the target value. Through the above method, it is ensured that the unit of the sorting density adjustment amount matches the actual equipment, so that it can be directly applied to the adjustment operation of the sorting equipment.

[0095] Step S1456: Normalize the initial medium flow rate adjustment coefficient to limit its value range within the controllable range of the medium flow rate of the sorting equipment, and obtain the target medium flow rate adjustment coefficient.

[0096] In this embodiment, the value range of the initial medium flow rate adjustment coefficient may be relatively wide and may not necessarily be within the controllable range of the medium flow rate of the sorting device. In order to enable this coefficient to effectively control the medium flow rate, normalization processing is required. Assume that the controllable range of the medium flow rate of the sorting device is from the lower limit value A to the upper limit value B. Through a specific normalization method, such as subtracting the minimum value of the initial medium flow rate adjustment coefficient from it and then dividing by the difference between the maximum value and the minimum value, a value between 0 and 1 is obtained. Then, according to the controllable range of the medium flow rate of the sorting device, a linear transformation is performed on this value between 0 and 1 to make its value range fall between A and B, obtaining the target medium flow rate adjustment coefficient. Through the above method, the initial medium flow rate adjustment coefficient is adjusted to a suitable value range to ensure that it can effectively adjust the medium flow rate within the controllable range of the sorting device.

[0097] Step S1457: Perform integerization processing on the initial particle size classification threshold so that it matches the calibration value of the screen aperture of the classification screen, obtaining the target particle size classification threshold.

[0098] In this embodiment, the initial particle size classification threshold may be a value with a decimal, while the calibration value of the screen aperture of the classification screen is usually an integer. In order to enable the initial particle size classification threshold to match the actual parameters of the classification screen, integerization processing is required. For example, methods such as rounding, rounding up, or rounding down are used to convert the initial particle size classification threshold into an integer. Assume that the initial particle size classification threshold is a decimal X. If the rounding method is used, when the decimal part of X is greater than or equal to 0.5, its integer part is incremented by 1; when the decimal part is less than 0.5, its integer part is directly taken, obtaining the target particle size classification threshold that matches the calibration value of the screen aperture of the classification screen. Through the above method, it is ensured that the particle size classification threshold conforms to the actual parameters of the classification screen, so as to accurately control the product particle size classification operation.

[0099] Step S1458: Associate the target separation density adjustment amount, the target medium flow rate adjustment coefficient, and the target particle size classification threshold according to the equipment number to generate the combined coal preparation process parameter adjustment strategy.

[0100] In this embodiment, a coal preparation plant may have multiple sorting devices, medium pumps, sizing screens and other equipment, and each piece of equipment has its corresponding equipment number. The obtained target sorting density adjustment amount, target medium flow adjustment coefficient and target particle size classification threshold are associated according to the corresponding equipment number. For example, for the sorting device with equipment number 1, the target sorting density adjustment amount, target medium flow adjustment coefficient and target particle size classification threshold applicable to it are combined; for the equipment with equipment number 2, a similar combination is made in the same way. Finally, the parameter adjustment amounts corresponding to all equipment are combined to form a combination of coal preparation process parameter adjustment strategies. In the above way, different process parameter adjustment amounts are corresponded to specific equipment, providing a complete strategy guidance for the precise control of coal preparation equipment.

[0101] Step S150: According to the combination of coal preparation process parameter adjustment strategies, control the coal preparation equipment to perform sorting density optimization operation, medium flow regulation operation and product particle size classification operation, and update the deep neural network model based on the real-time sorting effect feedback.

[0102] In this embodiment, in step S150, according to the combination of coal preparation process parameter adjustment strategies, controlling the coal preparation equipment to perform sorting density optimization operation, medium flow regulation operation and product particle size classification operation specifically includes steps S151 - S15 described below.

[0103] Step S151: Generate an opening adjustment command for the density control valve according to the sorting density adjustment amount in the combination of coal preparation process parameter adjustment strategies; control the density control valve to increase or decrease the medium density based on the opening adjustment command, and monitor the ash content index of the cleaned coal after sorting in real time; if the difference between the ash content index of the cleaned coal and the target ash value exceeds the preset ash content tolerance threshold, recalculate the sorting density adjustment amount and iteratively execute the density optimization operation until the difference is lower than the ash content tolerance threshold.

[0104] In this embodiment, the separation density adjustment amount for a certain separation device is obtained from the combination of coal preparation process parameter adjustment strategies. According to the pre-set relationship, this separation density adjustment amount is converted into an opening adjustment command for the density control valve. For example, if the separation density adjustment amount is positive, it means that the medium density needs to be increased, and according to the device characteristics and the pre-set corresponding relationship, it is calculated how much the opening of the density control valve needs to be increased; conversely, if the separation density adjustment amount is negative, the amount by which the valve opening needs to be decreased is calculated. Then, based on this opening adjustment command, the density control valve is controlled to act, thereby increasing or decreasing the medium density. While the density is being adjusted, the ash content index of the cleaned coal after separation is monitored in real time. Assume that the target ash value is a set value Y, and the preset ash tolerance threshold is Z. The ash content index of the cleaned coal monitored in real time is compared with the target ash value Y, and their difference is calculated. If this difference is greater than the preset ash tolerance threshold Z, it indicates that the current separation density is not appropriate, and the separation density adjustment amount needs to be recalculated. The recalculation may comprehensively consider the current ash content index of the cleaned coal, the previous adjustment history, and other relevant factors (such as coal quality characteristics, etc.), and a new separation density adjustment amount is obtained again through a deep neural network model or other preset calculation methods, and then the above process is repeated until the difference between the ash content index of the cleaned coal and the target ash value is lower than the ash tolerance threshold Z. Through the above method, the separation density can be dynamically adjusted according to the actual separation effect, ensuring that the ash content of the cleaned coal meets the requirements and improving the quality of coal preparation products.

[0105] Step S152: Calculate the target value of the rotational speed of the medium pump according to the medium flow adjustment coefficient in the combination of coal preparation process parameter adjustment strategies; adjust the rotational speed of the medium pump to the target value of the rotational speed through a frequency converter, and collect real-time medium flow velocity sensor data; if the deviation between the real-time medium flow velocity and the target flow velocity exceeds the preset flow velocity deviation threshold, then dynamically correct the target value of the rotational speed based on the proportional-integral-derivative algorithm until the deviation is lower than the flow velocity deviation threshold.

[0106] In this embodiment, a medium flow rate adjustment coefficient is obtained from the combination of coal preparation process parameter adjustment strategies. According to the characteristics of the medium pump and the pre-established relationship model, the target value of the medium pump speed is calculated using this medium flow rate adjustment coefficient. For example, there is a certain functional relationship between the flow rate and the speed of the medium pump. Through this functional relationship and the medium flow rate adjustment coefficient, the corresponding target value of the speed is calculated. Then, the speed of the medium pump is adjusted to this target value through a frequency converter. During the adjustment process, the medium flow rate data is collected in real time by a medium flow rate sensor installed on the pipeline. Assume that the target flow rate is the set value V, and the preset flow rate deviation threshold is W. The medium flow rate collected in real time is compared with the target flow rate V, and the deviation is calculated. If the difference exceeds the preset flow rate deviation threshold W, it indicates that the speed of the current medium pump needs to be further adjusted. At this time, the target value of the speed is dynamically corrected based on the proportional integral derivative algorithm (PID algorithm). The PID algorithm will generate a correction value according to the current deviation value, the change rate of the deviation, the cumulative deviation and other factors, and add this correction value to the original target value of the speed to obtain a new target value of the speed. Then, the speed of the medium pump is adjusted again through the frequency converter, and the deviation between the medium flow rate and the target flow rate is continuously monitored, and the above correction process is repeated until the deviation is lower than the preset flow rate deviation threshold W. In this way, the speed of the medium pump can be accurately controlled, so as to adjust the medium flow rate to meet the requirements of the coal preparation process.

[0107] Step S153: Generate a screen vibration amplitude adjustment parameter according to the particle size classification threshold in the combination of coal preparation process parameter adjustment strategies; adjust the current of the vibration motor of the sizing screen to match the vibration amplitude adjustment parameter, and collect the particle size distribution data of the oversize and the undersize; if the proportion of the target particle size range in the oversize is lower than the preset proportion threshold, increase the particle size classification threshold and re-perform the particle size classification operation until the proportion reaches the proportion threshold.

[0108] In this embodiment, first, a particle size classification threshold is obtained from the combination of coal preparation process parameter adjustment strategies. According to the characteristics of the sizing screen and the relationship between particle size classification and the vibration amplitude of the screen, a screen vibration amplitude adjustment parameter is generated using this particle size classification threshold. For example, through the corresponding relationship obtained from the pre-established mathematical model or experimental data, the particle size classification threshold is substituted into it to calculate a suitable screen vibration amplitude adjustment parameter. This parameter may involve adjustment information such as the size of the vibration amplitude and the vibration frequency to ensure that the sizing screen can effectively classify the products according to the particle size classification threshold.

[0109] Then, according to the generated screen vibration amplitude adjustment parameters, the current of the vibration motor of the sizing screen is adjusted. Since the magnitude of the vibration motor current affects the vibration amplitude of the screen, the vibration amplitude of the screen can be made to meet the requirements of the vibration amplitude adjustment parameters by adjusting the current. After adjusting the vibration motor current, the particle size distribution data of the oversize and undersize materials are collected by the particle size detection device installed on the sizing screen. These data can reflect the effect of the current sizing screen grading according to the set particle size grading threshold, including multi-dimensional information such as the proportion of materials in different particle size ranges.

[0110] Finally, the particle size distribution data of the oversize materials are analyzed to determine the proportion of the target particle size range in the oversize materials. Assume that the preset proportion threshold is P, and the actual proportion is compared with the preset proportion threshold P. If the actual proportion is lower than the preset proportion threshold P, it indicates that the current particle size grading operation has not achieved the expected effect, and it may be that the particle size grading threshold is set unreasonably. At this time, increase the particle size grading threshold, regenerate the screen vibration amplitude adjustment parameters according to the new particle size grading threshold, adjust the vibration motor current, perform the particle size grading operation again, and collect the particle size distribution data of the oversize and undersize materials again. Repeat the above comparison and adjustment process until the proportion of the target particle size range in the oversize materials reaches the preset proportion threshold P. Through the above method, the particle size grading threshold can be dynamically adjusted according to the actual particle size grading effect to ensure that the product particle size grading meets the requirements.

[0111] Furthermore, in step 150, updating the deep neural network model based on the real-time sorting effect feedback includes the following steps S154 - S157.

[0112] Step S154: Collect the set of clean coal quality indicators after performing the sorting density optimization operation, the medium flow rate adjustment operation, and the particle size grading operation. The set of clean coal quality indicators includes the actual ash content value, the actual sulfur content value, and the particle size qualification rate.

[0113] In this embodiment, after completing the sorting density optimization operation, the medium flow rate adjustment operation, and the particle size grading operation, the quality indicators of the clean coal are collected by corresponding detection devices. The actual ash content value of the clean coal is obtained using an ash content detection device, the actual sulfur content value is obtained using a sulfur content detection device, and the particle size qualification rate is calculated through the analysis of the particle size distribution data of the oversize and undersize materials. The actual ash content value, the actual sulfur content value, and the particle size qualification rate together constitute the set of clean coal quality indicators. Through the above method, the quality information of the clean coal after the coal preparation operation is comprehensively collected, providing data support for evaluating the coal preparation effect and updating the deep neural network model.

[0114] Step S155: Perform feature standardization and fusion processing on the difference between the actual ash content value and the target ash content value, the difference between the actual sulfur content value and the target sulfur content value, and the difference between the particle size qualification rate and the target qualification rate to obtain an effect feedback vector.

[0115] In this embodiment, first, calculate the difference between the actual ash content value and the target ash content value, the difference between the actual sulfur content value and the target sulfur content value, and the difference between the particle size qualification rate and the target qualification rate respectively. Assume that the target ash content value is G1, the actual ash content value is G2, then the ash content difference is G2 - G1; the target sulfur content value is S1, the actual sulfur content value is S2, the sulfur content difference is S2 - S1; the target qualification rate is R1, the actual particle size qualification rate is R2, and the particle size qualification rate difference is R2 - R1. Then arrange the three differences in a certain order, and then perform feature fusion processing to obtain an effect feedback vector. For example, the effect feedback vector can be expressed as [G2 - G1, S2 - S1, R2 - R1]. In the above way, the differences between the actual quality indexes and the target indexes after the coal preparation operation are integrated into a vector, which is convenient for subsequent input into the deep neural network model for analysis.

[0116] Step S156: Input the effect feedback vector and the coal preparation process parameter adjustment strategy combination into the deep neural network model, and calculate the strategy effect loss.

[0117] In this embodiment, the deep neural network model receives the effect feedback vector and the coal preparation process parameter adjustment strategy combination. The model internally calculates the strategy effect loss according to its predefined calculation method. Assume that the effect feedback vector contains the differences between the actual indexes and the target indexes. For example, the difference between the actual ash content and the target ash content is denoted as ΔA, the difference between the actual sulfur content and the target sulfur content is denoted as ΔS, the difference in the particle size qualification rate is denoted as ΔR, and the coal preparation process parameter adjustment strategy combination includes the separation density adjustment amount denoted as D, the medium flow rate adjustment coefficient denoted as F, and the particle size classification threshold denoted as T. Comprehensively consider these input elements to evaluate the deviation degree between the actual effect and the expected effect under these parameter combinations. For example, different weights may be assigned to ΔA, ΔS, ΔR, denoted as w1, w2, w3 respectively, and different weights may also be assigned to D, F, T, denoted as v1, v2, v3 respectively, and then calculate the strategy effect loss L based on the set loss calculation function. In this way, the gap between the actual separation effect and the expected effect under the current coal preparation process parameter adjustment strategy combination can be quantified, providing a basis for the subsequent update of the model.

[0118] Step S157: Fine-tune the weight parameters of the deep neural network model based on the strategy effect loss to generate an updated deep neural network model.

[0119] In this embodiment, after obtaining the policy effect loss L, according to the optimization algorithm adopted by the deep neural network model, this loss value is used to fine-tune the weight parameters of the model. Suppose there are numerous weight parameters in the model, such as the weights connecting different neurons, denoted as w11, w12, w21, etc. respectively. The optimization algorithm will adjust these weight parameters according to the policy effect loss L according to specific rules. For example, the common gradient descent algorithm will calculate the gradient of the loss function with respect to each weight parameter. The gradient represents the rate of change of the loss function in the direction of this weight parameter. If the gradient is positive, it means that increasing the value of this weight parameter will increase the loss function; if the gradient is negative, it means that increasing the value of this weight parameter will decrease the loss function. Then, according to the direction and magnitude of the gradient, and the preset learning rate (denoted as η), the weight parameters are adjusted. For the weight parameter w11, its new value after adjustment w11' = w11 - η × (the gradient of the loss function with respect to w11). All the weight parameters in the model are adjusted in a similar manner, so as to generate an updated deep neural network model. In this way, the deep neural network model can optimize its own weight parameters according to the difference between the actual separation effect and the expected effect, and improve the ability of the model to generate a more appropriate process parameter adjustment strategy combination in subsequent coal preparation decision-making.

[0120] Furthermore, the method in this embodiment further includes the process of training the deep neural network model. This process includes the following steps S160 - S190.

[0121] Step S160: Obtain a sample coal preparation data set, where the sample coal preparation data set includes a sample coal quality feature group, a sample equipment operation feature group, a sample market index sequence, and corresponding sample process parameter adjustment records.

[0122] In this embodiment, the sample coal preparation data set for training the deep neural network model can be screened based on the historical production data stored in the data coal preparation decision-making system of the coal preparation plant or the data collected in advance for model training. For the sample coal quality feature group, its source is the detection and feature extraction of different batches of coal samples. For example, within a certain period of time, multiple coal samples such as coal samples P1, P2, P3... are selected, and indicators such as ash content, sulfur content, and volatile matter of each coal sample are detected, and further features such as ash content distribution characteristics, sulfur content fluctuation characteristics, and volatile matter correlation characteristics are extracted, so as to form the sample coal quality feature group. The sample equipment operation feature group and the sample market index sequence can be obtained in a similar way, which will not be elaborated here one by one.

[0123] Step S170: Input the sample coal quality feature group, the sample equipment operation feature group, and the sample market index sequence into the initial deep neural network model to obtain the predicted process parameter adjustment amount.

[0124] In this embodiment, the prepared sample coal quality feature group, sample equipment operation feature group, and sample market index sequence are input into the initial deep neural network model. The ash distribution feature, sulfur content fluctuation feature, volatile matter correlation feature, etc. in the sample coal quality feature group can be input in vector form, and each feature dimension represents a certain aspect of the coal quality characteristics. For example, the ash distribution feature may be a vector containing information on multiple frequency bands, and each frequency band corresponds to a different ash distribution situation. The current change feature, vibration frequency correlation feature, medium flow velocity matching feature, etc. in the sample equipment operation feature group are also input in vector form, and these vectors reflect different characteristics of the equipment operation state. The clean coal price trend index, market demand prediction index, inventory turnover rate index, etc. in the sample market index sequence are also converted into appropriate vector forms and input into the initial deep neural network model. After processing the input sample data, the initial deep neural network model obtains corresponding feature representations and outputs the predicted process parameter adjustment amount, including the predicted separation density adjustment amount, the predicted medium flow rate adjustment coefficient, and the predicted particle size classification threshold, etc. Through the above method, the initial deep neural network model is used to analyze and process the sample data to obtain the predicted value of the coal preparation process parameter adjustment amount based on the current model state, providing a basis for subsequent evaluation of the model performance and optimization of the model.

[0125] Step S180: Calculate the mean square error loss between the predicted process parameter adjustment amount and the sample process parameter adjustment record, and update the weight parameters of the initial deep neural network model based on the gradient descent algorithm.

[0126] In this embodiment, the mean square error loss between the predicted process parameter adjustment amount and the sample process parameter adjustment record can be calculated first. Assume that the predicted separation density adjustment amount in the predicted process parameter adjustment amount is D_pred, and the separation density adjustment amount in the sample process parameter adjustment record is D_real; the predicted medium flow rate adjustment coefficient is F_pred, and the medium flow rate adjustment coefficient in the sample is F_real; the predicted particle size classification threshold is T_pred, and the particle size classification threshold in the sample is T_real. The calculation process of the mean square error loss is as follows: for the separation density adjustment amount, calculate the square of the difference between its predicted value and the actual value, that is, (D_pred - D_real)²; for the medium flow rate adjustment coefficient, calculate (F_pred - F_real)²; for the particle size classification threshold, calculate (T_pred - T_real)². Then add these three squared differences and take the average to obtain the mean square error loss L = [(D_pred - D_real)² + (F_pred - F_real)² + (T_pred - T_real)²] / 3 (here three parameters are taken as an example, and actually more parameters may be involved, and the calculation method is similar). After obtaining the mean square error loss, update the weight parameters of the initial deep neural network model based on the gradient descent algorithm. The core idea of the gradient descent algorithm is to determine the update direction of the weight parameters by calculating the gradient of the loss function (here the mean square error loss) with respect to the model weight parameters. Assume that there is a weight parameter w in the model, and its corresponding gradient is ∂L / ∂w. The gradient represents the rate of change of the loss function in the direction of this weight parameter. If the gradient is positive, it means that increasing the value of this weight parameter will increase the loss function; if the gradient is negative, it means that increasing the value of this weight parameter will decrease the loss function. According to the direction of the gradient and the pre-set learning rate η, update the weight parameters, and the update formula is w' = w - η × (∂L / ∂w). Update all the weight parameters in the model in this way, so that the model adjusts the weight parameters in the direction of reducing the mean square error loss. Through the above method, the error between the predicted value and the actual value is quantified, and the gradient descent algorithm is used to optimize the weight parameters of the model according to the error, so that the model can more accurately output the coal preparation process parameter adjustment amount that meets the actual requirements in subsequent predictions.

[0127] Step S190: When the mean square error loss is lower than the preset convergence threshold, stop training and obtain and save the deep neural network model.

[0128] In this embodiment, after each update of the weight parameters of the initial deep neural network model based on the gradient descent algorithm, the current mean squared error loss can be checked to see if it is lower than a preset convergence threshold. The preset convergence threshold can be a value set in advance according to experience and expectations for model performance, representing the acceptable range of the error between the model prediction value and the actual value of the sample. Suppose the preset convergence threshold is ε. When the calculated mean squared error loss L is less than ε, it indicates that after the model is trained, the error between the predicted process parameter adjustment amount and the sample process parameter adjustment record is within the acceptable range, and the model has achieved a certain degree of accuracy and stability. At this time, the training process of the model is stopped to avoid overfitting caused by overtraining, that is, the model performs well on the sample data but its generalization ability in actual applications decreases. After stopping the training, the deep neural network model at this time is saved, and the saved content includes information such as the structure of the model and the weight parameters optimized through training. These saved models can accurately predict the adjustment amount of the coal preparation process parameters based on the newly input coal quality data, equipment operation data, and market dynamic data during the subsequent coal preparation decision-making process, providing effective decision-making support for coal preparation production. Through the above method, it is ensured that the model stops training and is saved after reaching a certain accuracy, which not only guarantees the effectiveness and practicality of the model but also avoids unnecessary waste of training resources.

[0129] Figure 2 FIG. shows a schematic diagram of exemplary hardware and software components of a coal preparation decision-making system 100 based on multi-source data integration that can implement the inventive concept provided by some embodiments of the present application. For example, the processor 120 can be used on the coal preparation decision-making system 100 based on multi-source data integration and is used to execute the functions in the present invention.

[0130] The coal preparation decision-making system 100 based on multi-source data integration can be a general-purpose server or a special-purpose server, both of which can be used to implement the method of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0131] For example, the coal preparation decision-making system 100 based on multi-source data integration may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as disks, ROM, or RAM, or any combination thereof. Exemplarily, the coal preparation decision-making system 100 based on multi-source data integration may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to the above program instructions. The coal preparation decision-making system 100 based on multi-source data integration further includes an input / output (I / O) interface 150 between the computer and other input / output devices.

[0132] For ease of explanation, only one processor is described in the coal preparation decision-making system 100 based on multi-source data integration. However, it should be noted that the coal preparation decision-making system 100 based on multi-source data integration in the present application may also include multiple processors. Therefore, the steps executed by one processor described in the present application can also be jointly executed or separately executed by multiple processors. For example, if the processor of the coal preparation decision-making system 100 based on multi-source data integration executes steps A and B, it should be understood that steps A and B can also be jointly executed by two different processors or separately executed in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0133] In summary, the coal preparation decision-making method and system based on multi-source data integration provided in the embodiments of the present application achieve the intelligence and high efficiency of the coal preparation process, and can improve the coal preparation efficiency and the stability of product quality. Specifically, by integrating the dynamic association processing mechanism of multi-source data, different types of data such as coal quality detection data, equipment operation parameters, and market dynamic indicators are deeply fused to form a characterization method that can comprehensively and accurately reflect the characteristics of the coal preparation process, no longer limited to the simple analysis of a single data type by traditional methods, and comprehensively capture the complex relationships of various factors in the coal preparation process. On this basis, the preset deep neural network model uses advanced machine learning algorithms to deeply mine the process feature set, realizing the accurate prediction of key coal preparation parameters and changing the uncertainty of relying on experience judgment in the past. The generated parameter set not only has the ability to quickly respond to changes in the coal preparation process, but also closely interacts with the production control unit through a closed-loop feedback mechanism, making the generation and execution of instructions have an adaptive characteristic, capable of dynamically adjusting according to the real-time production situation, so that the coal preparation quality optimization process successfully changes from the traditional experience-dependent type to the scientific model prediction type, thereby significantly improving the accuracy and timeliness of parameter adjustment, effectively reducing the quality fluctuation in the coal preparation process, and improving the production efficiency.

[0134] Furthermore, through a dynamic update mechanism for the parameter weights of the deep neural network model based on real-time sorting effect feedback (which can correspond to quality indicators), a control system with the ability to continuously evolve is constructed, enabling the model to automatically correct its decision logic according to changing production conditions and ensuring stable and efficient performance in complex and changing industrial environments.

[0135] It should be noted that, in order to simplify the description disclosed in this application and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of this application, sometimes multiple features are merged into one embodiment, drawing, or description thereof.

[0136] In the above embodiments, the descriptions of the various embodiments have their own focuses. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. The embodiments, implementation manners, and related technical features of this application can be combined and replaced with each other without conflict. The above are only the preferred embodiments of this application, and do not impose any formal limitations on this application. However, any simple modifications, equivalent changes, and decorations made to the above embodiments based on the technical essence of this application without departing from the content of the technical solution of this application still fall within the scope of the technical solution of this application.

Claims

1. A coal preparation decision-making method based on multi-source data integration, characterized in that, The method includes: Obtaining a multi-source coal preparation data set in the coal preparation production process, where the multi-source coal preparation data set includes a coal quality detection data sequence, an equipment operation parameter sequence, and a market dynamic index sequence; Performing standardized preprocessing on the coal quality detection data sequence to generate a standardized coal quality data sequence, and extracting a coal quality feature group based on the standardized coal quality data sequence, where the coal quality feature group includes an ash distribution feature, a sulfur content fluctuation feature, and a volatile matter correlation feature; Performing time series feature extraction on the equipment operation parameter sequence to generate an equipment operation state feature group, where the equipment operation state feature group includes a current change feature of the separation equipment, a vibration frequency correlation feature, and a medium flow rate matching feature; Based on a preset deep neural network model, performing multi-dimensional feature fusion processing on the coal quality feature group, the equipment operation state feature group, and the market dynamic index sequence, and generating a combination of coal preparation process parameter adjustment strategies; According to the combination of coal preparation process parameter adjustment strategies, controlling the coal preparation equipment to perform sorting density optimization operations, medium flow rate adjustment operations, and product particle size classification operations, and updating the deep neural network model based on real-time sorting effect feedback.

2. The coal preparation decision-making method based on multi-source data integration according to claim 1, characterized in that, Performing standardized preprocessing on the coal quality detection data sequence to generate a standardized coal quality data sequence, including: Performing data format conversion processing on the original ash detection value, sulfur content detection value, and volatile matter detection value in the coal quality detection data sequence to generate an ash standardized sequence, a sulfur content standardized sequence, and a volatile matter standardized sequence with a unified dimension; Performing missing value filling processing on the ash standardized sequence, and generating a complete ash continuous sequence based on the linear interpolation algorithm of adjacent coal sample detection values; Performing outlier correction processing on the sulfur content standardized sequence, and truncating and replacing the sulfur content detection value exceeding the sulfur content threshold interval based on a preset sulfur content threshold interval to generate a corrected sulfur content sequence; Performing noise filtering processing on the volatile matter standardized sequence, and smoothing the volatile matter standardized sequence using a moving window mean algorithm to generate a denoised volatile matter sequence; Combining the ash continuous sequence, the corrected sulfur content sequence, and the denoised volatile matter sequence into a standardized coal quality data sequence, and storing the standardized coal quality data sequence in a distributed database.

3. The coal preparation decision-making method based on multi-source data integration according to claim 2, characterized in that, Extracting a coal quality feature group based on the standardized coal quality data sequence, including: Performing frequency domain transformation processing on the ash continuous sequence to generate an ash frequency spectrum distribution feature, and extracting the frequency band with the main frequency amplitude exceeding a preset ash amplitude threshold in the ash frequency spectrum distribution feature as the ash distribution feature; Performing fluctuation period analysis on the corrected sulfur content sequence, calculating the absolute value sequence of the difference between adjacent sulfur content detection values, and performing moving average processing on the absolute value sequence of the difference to generate a sulfur content fluctuation intensity sequence, and marking the interval exceeding a preset sulfur content fluctuation threshold in the sulfur content fluctuation intensity sequence as the sulfur content fluctuation feature; Performing correlation analysis on the denoised volatile matter sequence and the ash continuous sequence, calculating the Pearson correlation coefficient between the volatile matter detection value and the corresponding ash detection value, and marking the interval with the Pearson correlation coefficient lower than a preset correlation threshold as the volatile matter correlation feature; Associate the ash distribution characteristics, the sulfur content fluctuation characteristics, and the volatile matter correlation characteristics with the corresponding coal sample batch numbers to generate a coal quality characteristic group.

4. The coal preparation decision-making method based on multi-source data integration according to claim 1, characterized in that, Extract time series characteristics from the equipment operation parameter sequence to generate an equipment operation state characteristic group, including: Perform trend decomposition processing on the current time series data of the sorting equipment in the equipment operation parameter sequence, separate the long-term trend component, periodic fluctuation component, and residual component of the current, and mark the interval where the slope of the long-term trend component of the current exceeds the preset slope threshold as the current change characteristic; Perform joint analysis on the vibration frequency time series data and the medium flow velocity time series data, calculate the covariance matrix of the vibration frequency mean value and the medium flow velocity mean value within the same time window, and mark the window where the main diagonal element in the covariance matrix exceeds the preset covariance threshold as the vibration frequency correlation characteristic; Perform mutation point detection on the medium flow velocity time series data, calculate the cumulative deviation amount of the medium flow velocity sequence based on the cumulative sum algorithm, and obtain the medium flow velocity matching characteristic according to the moment when the cumulative deviation amount exceeds the preset deviation threshold; Align the current change characteristic, the vibration frequency correlation characteristic, and the medium flow velocity matching characteristic according to the timestamp to generate an equipment operation state characteristic group.

5. The coal preparation decision-making method based on multi-source data integration according to claim 4, characterized in that, Perform mutation point detection on the medium flow velocity time series data, calculate the cumulative deviation amount of the medium flow velocity sequence based on the cumulative sum algorithm, and obtain the medium flow velocity matching characteristic according to the moment when the cumulative deviation amount exceeds the preset deviation threshold, including: Extract the corresponding sequence of timestamps and medium flow velocity values in the medium flow velocity time series data, and generate a medium flow velocity change curve in the order of timestamps; Calculate the global mean of the medium flow velocity change curve, and generate a medium flow velocity reference value corresponding to each timestamp based on the global mean; Accumulatively calculate the difference between the actual medium flow velocity value and the medium flow velocity reference value at each timestamp to generate a cumulative deviation amount sequence; Real-time monitor the change trend of the cumulative deviation amount sequence. When the cumulative deviation amount exceeds the preset deviation threshold, record the current timestamp as the mutation point candidate moment; Compare the mean values within the front and rear time windows of the actual medium flow velocity value at the candidate moment. If the differences between the actual medium flow velocity value and the mean value of the previous time window and the mean value of the next time window both exceed the flow velocity fluctuation tolerance, confirm that the mutation point candidate moment is an effective medium flow velocity mutation point; Associate the timestamp of the effective medium flow velocity mutation point and the corresponding medium flow velocity change amplitude with the sorting equipment number of the equipment operation parameter sequence to generate a medium flow velocity matching characteristic.

6. The coal preparation decision-making method based on multi-source data integration according to claim 1, characterized in that, Based on a preset deep neural network model, perform multi-dimensional feature fusion processing on the coal quality characteristic group, the equipment operation state characteristic group, and the market dynamic index sequence, and generate a combined coal preparation process parameter adjustment strategy, including: Map the ash distribution characteristics, sulfur content fluctuation characteristics, and volatile matter correlation characteristics in the coal quality characteristic group to a first feature vector; Map the current change characteristic, vibration frequency correlation characteristic, and medium flow velocity matching characteristic in the equipment operation state characteristic group to a second feature vector; Map the clean coal price trend index, market demand forecast index, and inventory turnover rate index in the market dynamic index sequence to a third feature vector; Concatenate the first feature vector, the second feature vector, and the third feature vector through the input layer of the deep neural network model to generate a fused feature vector; Perform a non-linear transformation on the fused feature vector through the hidden layer of the deep neural network model, output the sorting density adjustment amount, medium flow rate adjustment coefficient, and particle size classification threshold, and generate a combination of coal preparation process parameter adjustment strategies.

7. The coal preparation decision-making method based on multi-source data integration according to claim 6, characterized in that, Perform a non-linear transformation on the fused feature vector through the hidden layer of the deep neural network model, output the sorting density adjustment amount, medium flow rate adjustment coefficient, and particle size classification threshold, and generate a combination of coal preparation process parameter adjustment strategies, including: Input the fused feature vector into the first hidden layer of the deep neural network model, and perform a non-linear mapping on the fused feature vector through an activation function to generate a first intermediate feature vector; Input the first intermediate feature vector into the second hidden layer of the deep neural network model, perform a feature similarity match with the preset historical adjustment record of process parameters, and filter out the feature dimensions corresponding to the historical adjustment records with a similarity higher than the preset matching threshold; Reassign weights to the first intermediate feature vector based on the filtered feature dimensions to generate a second intermediate feature vector; Input the second intermediate feature vector into the third hidden layer of the deep neural network model, and perform a linear transformation through the sorting density adjustment amount output channel, medium flow rate adjustment coefficient output channel, and particle size classification threshold output channel respectively to generate an initial sorting density adjustment amount, an initial medium flow rate adjustment coefficient, and an initial particle size classification threshold; Perform a unit dimension conversion on the initial sorting density adjustment amount to make it consistent with the density measurement unit of the current sorting equipment to obtain the target sorting density adjustment amount; Perform a normalization process on the initial medium flow rate adjustment coefficient to limit its value range within the controllable range of the medium flow rate of the sorting equipment to obtain the target medium flow rate adjustment coefficient; Perform an integerization process on the initial particle size classification threshold to make it match the calibration value of the screen aperture of the classification screen to obtain the target particle size classification threshold; Associate the target sorting density adjustment amount, target medium flow rate adjustment coefficient, and target particle size classification threshold according to the equipment number to generate the combination of coal preparation process parameter adjustment strategies.

8. The coal preparation decision-making method based on multi-source data integration according to claim 1, characterized in that, According to the combination of coal preparation process parameter adjustment strategies, control the coal preparation equipment to perform sorting density optimization operations, medium flow rate adjustment operations, and product particle size classification operations, including: Generate an opening adjustment command for the density control valve according to the sorting density adjustment amount in the combination of coal preparation process parameter adjustment strategies; control the density control valve to increase or decrease the medium density based on the opening adjustment command, and monitor the ash content index of the sorted clean coal in real time; if the difference between the ash content index of the sorted clean coal and the target ash value exceeds the preset ash tolerance threshold, recalculate the sorting density adjustment amount and iteratively execute the density optimization operation until the difference is lower than the ash tolerance threshold; Calculate the target speed value of the medium pump according to the medium flow rate adjustment coefficient in the coal preparation process parameter adjustment strategy combination; adjust the speed of the medium pump to the target speed value through the frequency converter, and collect the real-time medium flow rate sensor data; if the deviation between the real-time medium flow rate and the target flow rate exceeds the preset flow rate deviation threshold, dynamically correct the target speed value based on the proportional integral derivative algorithm until the deviation is lower than the flow rate deviation threshold; and Generate the screen vibration amplitude adjustment parameter according to the particle size classification threshold in the coal preparation process parameter adjustment strategy combination; adjust the current of the vibration motor of the classification screen to match the vibration amplitude adjustment parameter, and collect the particle size distribution data of the oversize and the undersize; if the proportion of the target particle size range in the oversize is lower than the preset proportion threshold, increase the particle size classification threshold and re-perform the particle size classification operation until the proportion reaches the proportion threshold.

9. The coal preparation decision-making method based on multi-source data integration according to claim 1, characterized in that, Updating the deep neural network model based on real-time sorting effect feedback, including: Collect the set of clean coal quality indicators after performing the sorting density optimization operation, the medium flow rate adjustment operation, and the particle size classification operation, and the set of clean coal quality indicators includes the actual ash content, the actual sulfur content, and the particle size qualification rate; Perform feature standardization and fusion processing on the difference between the actual ash content and the target ash content, the difference between the actual sulfur content and the target sulfur content, and the difference between the particle size qualification rate and the target qualification rate to obtain an effect feedback vector; Input the effect feedback vector and the coal preparation process parameter adjustment strategy combination into the deep neural network model to calculate the strategy effect loss; Fine-tune the weight parameters of the deep neural network model based on the strategy effect loss to generate an updated deep neural network model.

10. A coal preparation decision-making system based on multi-source data integration, characterized in that, Comprising a processor and a machine-readable storage medium, the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the coal preparation decision-making method based on multi-source data integration according to any one of claims 1-9.

Images