Coal slime water treatment dosing quantity multi-model collaborative prediction method, device and equipment

Abstract

This disclosure provides a multi-model collaborative prediction method, apparatus, equipment, and storage medium for coal slurry water treatment dosage. In some embodiments of this disclosure, the original production process monitoring parameters of the coal slurry water settling stage are obtained; data cleaning is performed on the original production process monitoring parameters to obtain production process monitoring input parameters; these input parameters are input into a spatial domain dosage prediction model to obtain a spatial domain predicted dosage; these input parameters are also input into a frequency domain dosage prediction model to obtain a frequency domain predicted dosage; and a target dosage prediction value is determined based on the spatial domain predicted dosage, a first weight, the frequency domain predicted dosage, and a second weight. This disclosure uses the production process monitoring parameters of the coal slurry water settling stage, along with spatial and frequency domain dosage prediction models, to automatically and accurately predict the dosage of coal slurry water, thereby improving the accuracy of coal slurry water dosage.

Description

Technical Field

[0001] This disclosure relates to the field of coal processing technology, and in particular to a method, apparatus, equipment and storage medium for multi-model collaborative prediction of chemical dosage in coal slurry water treatment. Background Technology

[0002] Coal slurry water treatment is a core step in coal washing and processing, mainly including the classification, concentration, and clarification of coal slurry water. Its core function is to add flocculants (such as polyacrylamide) and coagulant aids (such as polyaluminum chloride) to the coal slurry water to promote the aggregation and bridging of fine coal slurry particles, forming larger flocs. These flocs are then settled in a clarification tank and dewatered by a filter press. Under the premise of strictly controlling the turbidity of the effluent and the moisture content of the filter cake, the wastewater resource is recycled in a closed loop and the coal slurry resource is effectively recovered.

[0003] Existing chemical dosing control technologies for coal slurry water mainly rely on manual experience for adjustment. This manual adjustment depends primarily on operators visually observing floc morphology and effluent turbidity, or on adjusting the dosage based on delayed laboratory concentration test results.

[0004] Currently, the dosage of chemicals in coal slurry water is adjusted manually based on experience, resulting in low precision. Summary of the Invention

[0005] This disclosure provides a multi-model collaborative prediction method, apparatus, and equipment for coal slurry water treatment, which at least solves the problem of low accuracy in existing coal slurry water chemical dosage methods.

[0006] The technical solution disclosed herein is as follows: This disclosure provides a multi-model collaborative prediction method for chemical dosage in coal slurry water treatment, including: Obtain the original production process monitoring parameters for the coal slurry water settling stage; Data cleaning is performed on the original production process monitoring parameters to obtain the production process monitoring input parameters; The production process monitoring input parameters are input into the spatial domain dosing prediction model to obtain the spatial domain predicted dosing amount; The production process monitoring input parameters are input into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage. The target dosage prediction value is determined based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight, wherein the first weight is the weight corresponding to the spatial domain predicted dosage, and the second weight is the weight corresponding to the frequency domain predicted dosage.

[0007] Optionally, the original production process monitoring parameters include: feed flow rate, feed concentration, pH value, underflow flow rate, underflow concentration, overflow turbidity, and settling interface depth; the original production process monitoring parameters for obtaining the coal slurry water settling stage include: The feed flow rate is collected by a flow monitoring device installed on the feed pipeline of the thickener. The feed concentration is collected by a concentration detection device installed on the feed pipe of the thickener. The pH value was collected by installing a pH detector in the middle of the thickening tank; The underflow flow rate is collected by installing a flow monitoring device in the underflow pipe at the bottom of the thickener. The concentration of the underflow is collected by installing a concentration detection device in the underflow pipe at the bottom of the thickener. The turbidity of the overflow water was collected by installing a turbidity meter in the overflow trough of the thickening tank. The settling interface depth was collected by installing a sludge interface meter in the middle of the thickening tank.

[0008] Optionally, the step of inputting the production process monitoring input parameters into the spatial domain dosing prediction model to obtain the spatial domain predicted dosing includes: The production process monitoring input parameters are input into the spatial domain dosing prediction model to obtain trend prediction results, seasonal prediction results, and residual prediction results. Based on the trend prediction results, the seasonal prediction results, and the residual prediction results, the spatial domain predicted dosage is calculated.

[0009] Optionally, the step of inputting the production process monitoring input parameters into the spatial domain dosage prediction model to obtain trend prediction results, seasonal prediction results, and residual prediction results includes: Based on the cycle length, seasonal smoothing window width, trend smoothing window width, and low-pass filter window width, the production process monitoring input parameters are centrally preprocessed to obtain preprocessed data. Enter the outer loop iteration, initialize the data weights, and execute the inner loop; the inner loop includes: smoothing the detrended sequence by periodic grouping to obtain the preliminary seasonal term, smoothing the deseasonal sequence to update the trend term, and performing low-pass filtering correction on the trend term until the trend term and seasonal term converge. In the outer loop, the absolute median difference of the residuals is calculated, and the data weights are updated based on the weight function until the preset number of iterations is reached or the weights converge. The final decomposition results are output as follows: the trend prediction result, the seasonal prediction result, and the residual prediction result.

[0010] Optionally, the step of inputting the production process monitoring input parameters into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage includes: Variational mode decomposition is used to perform frequency domain decomposition on the input parameters of the production process monitoring to obtain multiple stationary mode components with finite bandwidth. Based on a temporal convolutional network, a multi-input prediction sub-model is established for each of the multiple stationary mode components to output the prediction results of multiple sub-models. The prediction results of multiple sub-models are reconstructed to obtain the frequency domain predicted dosage.

[0011] Optionally, determining the target dosage prediction value based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight includes: Historical measured dosage data are divided into rate-period data and forecast-period data; Based on the measured values ​​of the periodic data, the spatial domain predicted dosage, and the frequency domain predicted dosage, calculate the first marginal likelihood corresponding to the spatial domain dosage prediction model and the second marginal likelihood corresponding to the frequency domain dosage prediction model. Based on the first marginal likelihood and the first prior probability corresponding to the spatial domain dosage prediction model, the posterior probability of the spatial domain dosage prediction model is calculated and used as the first weight. Based on the second marginal likelihood and the second prior probability corresponding to the frequency domain dosage prediction model, the posterior probability of the frequency domain dosage prediction model is calculated and used as the second weight. The target dosage prediction value is obtained by weighting the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight.

[0012] This disclosure also provides a multi-model collaborative prediction device for chemical dosage in coal slurry water treatment, comprising: The acquisition module is used to acquire the original production process monitoring parameters of the coal slurry water settling stage; The cleaning module is used to perform data cleaning operations on the original production process monitoring parameters to obtain the production process monitoring input parameters; The first prediction module is used to input the production process monitoring input parameters into the spatial domain dosing prediction model to obtain the spatial domain predicted dosing amount. The second prediction module is used to input the production process monitoring input parameters into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage. The determination module is used to determine the target dosage prediction value based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight, wherein the first weight is the weight corresponding to the spatial domain predicted dosage, and the second weight is the weight corresponding to the frequency domain predicted dosage.

[0013] This disclosure also provides an electronic device, including: processor; Memory used to store processor-executable instructions; The processor is configured to execute instructions to implement the steps in the above method.

[0014] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0015] This disclosure also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the method described above.

[0016] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects: In some embodiments of this disclosure, the original production process monitoring parameters of the coal slurry sedimentation stage are obtained; data cleaning is performed on the original production process monitoring parameters to obtain production process monitoring input parameters; the production process monitoring input parameters are input into a spatial domain dosage prediction model to obtain a spatial domain predicted dosage; the production process monitoring input parameters are input into a frequency domain dosage prediction model to obtain a frequency domain predicted dosage; and the target dosage prediction value is determined based on the spatial domain predicted dosage, a first weight, a frequency domain predicted dosage, and a second weight. This disclosure uses the production process monitoring parameters of the coal slurry sedimentation stage, and utilizes both spatial domain and frequency domain dosage prediction models to automatically and accurately predict the dosage of coal slurry, thereby improving the accuracy of coal slurry dosage.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0019] Figure 1 A flowchart illustrating a multi-model collaborative prediction method for chemical dosage in coal slurry water treatment, provided as an exemplary embodiment of this disclosure; Figure 2 A schematic diagram of a multi-model collaborative prediction device for chemical dosage in coal slurry water treatment, provided as an exemplary embodiment of this disclosure; Figure 3 A schematic diagram of the structure of an electronic device provided for an exemplary embodiment of this disclosure. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0021] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0022] It should be noted that the user information involved in this disclosure includes, but is not limited to, user device information and user personal information; the collection, storage, use, processing, transmission, provision and disclosure of user information in this disclosure all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0023] To address the aforementioned technical problems, in some embodiments of this disclosure, the original production process monitoring parameters of the coal slurry sedimentation stage are obtained; data cleaning operations are performed on the original production process monitoring parameters to obtain production process monitoring input parameters; the production process monitoring input parameters are input into a spatial domain dosage prediction model to obtain a spatial domain predicted dosage; the production process monitoring input parameters are input into a frequency domain dosage prediction model to obtain a frequency domain predicted dosage; and the target dosage prediction value is determined based on the spatial domain predicted dosage, a first weight, a frequency domain predicted dosage, and a second weight. This disclosure uses the production process monitoring parameters of the coal slurry sedimentation stage, and utilizes both spatial domain and frequency domain dosage prediction models to automatically and accurately predict the dosage of coal slurry, thereby improving the accuracy of coal slurry dosage.

[0024] The technical solutions provided by the embodiments of this disclosure are described in detail below with reference to the accompanying drawings.

[0025] Figure 1 This is a schematic flowchart illustrating a multi-model collaborative prediction method for chemical dosage in coal slurry water treatment, provided as an exemplary embodiment of this disclosure. Figure 1 As shown, the method includes: S101: Obtain the original production process monitoring parameters for the coal slurry water settling stage; S102: Perform data cleaning on the original production process monitoring parameters to obtain the production process monitoring input parameters; S103: Input the production process monitoring input parameters into the spatial domain dosing prediction model to obtain the spatial domain predicted dosing amount; S104: Input the production process monitoring input parameters into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage; S105: Determine the target dosage prediction value based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight. The first weight is the weight corresponding to the spatial domain predicted dosage, and the second weight is the weight corresponding to the frequency domain predicted dosage.

[0026] In this embodiment, the subject executing the above method is a terminal device or a server.

[0027] The terminal device includes, but is not limited to, mobile stations (MS), mobile terminals, mobile phones, handsets, and portable equipment. This terminal device can communicate with one or more core networks via a radio access network (RAN). For example, the terminal device can be a mobile phone (or "cellular" phone), a computer with wireless communication capabilities, a computer with wireless transceiver capabilities, a virtual reality (VR) terminal device, an AR terminal device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical care, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. The operating systems installed on the terminal device include, but are not limited to, iOS, Android, Windows, Linux, and Mac OS. In different networks, terminals may be called by different names, such as: user equipment, mobile station, user unit, station, cellular phone, personal digital assistant, wireless modem, wireless communication device, handheld device, laptop, cordless phone, wireless local loop station, television, etc. For ease of description, this embodiment will simply refer to it as terminal device.

[0028] In this embodiment, the implementation form of the server is not limited. For example, the server can be a conventional server, a cloud server, a cloud host, a virtual center, or other server devices. The server mainly consists of a processor, hard disk, memory, system bus, and other common computer architecture types.

[0029] In this embodiment, the original production process monitoring parameters of the coal slurry sedimentation stage are obtained; data cleaning is performed on the original production process monitoring parameters to obtain the production process monitoring input parameters; the production process monitoring input parameters are input into the spatial domain dosage prediction model to obtain the spatial domain predicted dosage; the production process monitoring input parameters are input into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage; the target dosage prediction value is determined based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight. This disclosure uses the production process monitoring parameters of the coal slurry sedimentation stage, and utilizes the spatial domain dosage prediction model and the frequency domain dosage prediction model to automatically and accurately predict the dosage of coal slurry, thereby improving the accuracy of coal slurry dosage.

[0030] In some embodiments of this disclosure, the multi-model collaborative prediction system for coal slurry water treatment dosage includes: an intelligent sensing module, a cleaning and standardization module, a dosage prediction module, and a dosage weight adjustment module based on a Bayesian model. The intelligent sensing module primarily monitors the entire production process of coal slurry water settling in coal preparation plants. It collects core process parameters in real time by deploying high-precision detection equipment such as concentration meters, flow meters, and interface meters at key nodes. Monitoring indicators include, but are not limited to: feed flow rate, feed concentration, pH value, underflow flow rate, underflow concentration, overflow turbidity, settling interface depth, temperature, and the concentration and real-time dosage of flocculants and coagulants. The cleaning and standardization module is responsible for preprocessing the raw data collected by the intelligent sensing module. The system sequentially performs data missing value completion, outlier correction, and data standardization operations according to predetermined rules, and rigorously verifies and outputs the preprocessing results, providing high-quality data support for subsequent prediction models. The dosage prediction module, centered on the monitored effective dosage, constructs independent prediction algorithms in both the spatial and frequency domains. Spatial domain prediction: First, the collected monitoring information is decomposed into three parts based on the STL algorithm: trend sequence, seasonal sequence, and residual sequence. Then, a feature dataset is constructed based on the characteristics of each sequence, and trained and predicted using the LightGBM model. Frequency domain prediction: The data is decomposed based on variational mode decomposition, and deep feature extraction and prediction are performed using a temporal convolutional network. The dosage weight adjustment module based on a Bayesian model dynamically adjusts and fuses the predicted dosage in the spatial and frequency domains using a Bayesian model, ultimately outputting a more accurate and robust dosage prediction result.

[0031] It should be noted that the original production process monitoring parameters include: feed flow rate, feed concentration, pH value, underflow flow rate, underflow concentration, overflow turbidity, settling interface depth, flocculant concentration, flocculant dosage, coagulant concentration, and coagulant dosage.

[0032] In some embodiments of this disclosure, raw production process monitoring parameters for the coal slurry settling stage are obtained. One possible approach is to collect the feed flow rate using a flow monitoring device installed on the feed pipe of the thickener; to collect the feed concentration using a concentration detection device installed on the feed pipe of the thickener; to collect the pH value using a pH detector installed in the middle of the thickener; to collect the underflow flow rate using a flow monitoring device installed on the underflow pipe at the bottom of the thickener; to collect the underflow concentration using a concentration detection device installed on the underflow pipe at the bottom of the thickener; to collect the overflow turbidity using a turbidimeter installed in the overflow trough of the thickener; and to collect the settling interface depth using a sludge interface meter installed in the middle of the thickener. Specifically, the feed flow rate is collected by installing a flow monitoring device on the feed pipe of the thickener. Typically, an ultrasonic flow meter is installed when the pipe is full, and an open channel flow meter is installed when the pipe is not full, to achieve real-time monitoring of the feed flow rate. The feed concentration is collected by installing a concentration detection device (usually an ultrasonic concentration meter) on the feed pipe of the thickener, achieving real-time monitoring of the feed concentration. The pH value is collected by installing a pH detector in the middle of the thickener, achieving real-time monitoring of the pH value of the coal slurry water. The settling interface depth of coal sludge in the thickener is collected by installing a sludge interface meter in the middle of the thickener to monitor the settling interface depth in real time. The underflow flow rate is collected by installing a flow monitoring device on the underflow pipe at the bottom of the thickener. Since the underflow pipe is usually full, an ultrasonic flow meter is generally used for monitoring. The underflow concentration is collected by installing a concentration detection device (usually an ultrasonic concentration meter) on the underflow pipe at the bottom of the thickener to achieve real-time monitoring of the underflow concentration. The overflow turbidity is collected by installing a turbidity meter in the overflow trough of the thickener to achieve real-time monitoring of the overflow turbidity. Chemical concentrations and dosages include: flocculant concentration, flocculant dosage, coagulant concentration, and coagulant dosage. These parameters are usually directly based on the monitoring parameters of the chemical dosing equipment; if the equipment's own monitoring function is incomplete, accurate monitoring can be achieved by adding additional concentration meters and flow meters.

[0033] In some embodiments of this disclosure, data cleaning operations are performed on the original production process monitoring parameters to obtain production process monitoring input parameters. Specifically, this includes the following steps: I. Time Series Data Integrity Check. This mainly checks whether the time periods are continuous, without repetition, and without out-of-order data, and whether the data collection frequency is fixed. If data is missing, the missing time period is located, and the values ​​are checked, analyzing the missing value rate and determining the extent of the missing data. For short-term missing data, methods such as the mean method and median method can be used to fill in the missing data. For long-term missing data, the linear interpolation method is used to fill in the missing data.

[0034] II. Threshold Legality Check. This mainly involves checking the range of values ​​for the monitored data. Data that clearly exceeds the normal range is removed, and the data is supplemented based on the time sequence. The main value ranges of the data involved in this disclosure are as follows: The feed flow rate typically ranges from 0 to 3000 m³ / h, depending on the system's production capacity.

[0035] The feed concentration range is typically 20-120 g / L.

[0036] The pH value is usually 5-9.

[0037] The range of bottom flow rate is usually 0-1500 m³ / h.

[0038] The underflow concentration is typically in the range of 400-800 g / L.

[0039] The overflow turbidity value requires that the suspended solids concentration not exceed 50 mg / L (approximately equivalent to 30-50 NTU). Advanced processes can reduce the turbidity to below 10 NTU, typically 0-75 NTU.

[0040] The range of the sedimentation interface depth depends on the depth of the thickener, and is usually 0 minus the depth of the thickener.

[0041] Flocculant concentration, flocculant dosage, coagulant concentration, and coagulant dosage are usually monitored by the reagent addition equipment. If the equipment's monitoring function is incomplete, it can be achieved by adding a concentration meter and flow meter. The concentration is usually 0.1-0.3%, and the flow rate is 0-10 m³ / h.

[0042] III. Outlier Inspection. This mainly involves checking the reasonableness of the monitoring data. In actual production, the concentration, flow rate, dosage, and turbidity of coal slurry water approximately conform to a normal distribution. Therefore, the Median Absolute Deviation (MAD) method is used to remove outliers. The core idea is to use the median to measure the data center and the absolute deviation of the data from the median to measure the degree of dispersion. Values ​​exceeding a threshold are considered outliers. The inspection process is as follows: (1) Calculate the median of the data.

[0043] (2) Calculate the absolute deviation of each data point from the median: |x-med|.

[0044] (3) Calculate the median of the absolute deviation, i.e., MAD: MAD = median(|x-med|).

[0045] (4) Calculate the outlier threshold (2.5 is commonly used). MAD or 3 MAD).

[0046] (5) Exceeding [med-k] MAD,med+k Values ​​within the MAD range are outliers.

[0047] (6) Replace the detected outliers according to the short-term missing value completion method and the long-term missing value completion method.

[0048] IV. Repeatability Check. In time-series data, repeatability checks primarily address two high-frequency issues: time repetition and global repetition. Time repetition manifests as the same timestamp appearing ≥2 times, typically caused by repeated sensor reporting or repeated splicing during data merging, disrupting temporal continuity and leading to interpolation and modeling errors. Global repetition refers to an entire row being completely identical, usually caused by repeated storage during data acquisition or repeated writing during data transformation, increasing data redundancy and raising statistical indicators. Data with both time repetition and global repetition must be removed.

[0049] V. Data Standardization. When standardizing data, Z-score standardization is used for each individual time series data to transform the data into dimensionless values ​​centered on the mean and with the standard deviation as the unit. This preserves the relative trends between parameters (e.g., an increase in concentration corresponds to an increase in the standardized value) and maps all data to the [-1,1] interval (when the distribution is relatively concentrated), which is suitable for the data scale requirements of modeling.

[0050] After completing the data standardization process, the data is saved and used as accumulated data for subsequent drug dosage prediction.

[0051] In some embodiments of this disclosure, production process monitoring input parameters are input into a spatial domain dosing prediction model to obtain the spatial domain predicted dosing amount. The production process monitoring input parameters are input into the spatial domain dosing prediction model to obtain trend prediction results, seasonal prediction results, and residual prediction results; based on the trend prediction results, seasonal prediction results, and residual prediction results, the spatial domain predicted dosing amount is calculated. Specifically, in the spatial domain-based coal slurry water treatment dosing prediction, the STL decomposition algorithm (Seasonal and Trend decomposition using Loess) can decompose the time-series data related to coal slurry water dosing amount into trend terms (…). ), seasonal items ( ), residual terms ( The system consists of three parts: spatial domain prediction of drug dosage. Furthermore, LightGBM is used to accurately predict these three components separately. Finally, the prediction results of the three components are added together to obtain the final predicted dosage value, which can significantly reduce the prediction error caused by the strong nonlinearity and strong disturbance of the coal slurry water system.

[0052] In the above embodiments, the production process monitoring input parameters are input into the spatial domain dosage prediction model to obtain trend prediction results, seasonal prediction results, and residual prediction results. One possible approach is to perform centralized preprocessing on the production process monitoring input parameters based on the cycle length, seasonal smoothing window width, trend smoothing window width, and low-pass filter window width to obtain preprocessed data; then, enter the outer loop iteration, initialize data weights, and execute the inner loop; the inner loop includes: smoothing the detrended sequence by cycle to obtain a preliminary seasonal term, smoothing the deseasonal sequence to update the trend term, and performing low-pass filtering correction on the trend term until the trend term and seasonal term converge; in the outer loop, calculate the absolute median difference of the residuals and update the data weights based on the weight function until a preset number of iterations is reached or the weights converge; finally, output the trend prediction results, seasonal prediction results, and residual prediction results obtained from the decomposition.

[0053] Specifically, STL decomposition is an iterative optimization process, which is mainly divided into two parts: the inner loop and the outer loop. The overall process is as follows: Step 1: Initialize parameters and preprocess. Key parameters to determine include: period length p (e.g., p=12 for monthly data, p=7 for daily data), Loess smoothing window, seasonal smoothing window width ns (must be odd, and ns>p), trend smoothing window width nt (must be odd, and nt>ns), and low-pass filter window width nl (usually nl=1+4×ceil(p / 2)), where ceil represents rounding up.

[0054] For the original sequence Centralized preprocessing was performed (Z-score normalization was used for preprocessing in the aforementioned embodiments, which is for dosage data) to eliminate the influence of extreme values.

[0055] Step 2: Outer loop (robust iteration).

[0056] The core of the outer loop is to reduce the impact of outliers through weighted algorithms, typically iterating 0 to 15 times (depending on actual needs). Specifically, the weights are initialized. =1 (all data weights are equal); execute the inner loop to obtain the current trend item. Seasonal items residuals ; Calculate the robustness weights of the residuals: Calculate the absolute median difference (MAD) of the residuals; Weight update: ; ; in, It is the bisquare weight function.

[0057] Repeat the above steps until the preset number of iterations or weight convergence is reached.

[0058] Step 3: Inner loop (core decomposition, executed once per outer loop iteration).

[0059] The inner loop is the core of STL, and its goal is to gradually optimize the trend and seasonal items. The process is as follows: (1) Detrending: Calculate the detrending sequence Yt-Tt(k-1) (Tt(0) is initially 0).

[0060] (2) Periodic subsequence extraction: The detrended sequence is grouped by period to obtain p subsequences (e.g., monthly data is divided into subsequences for January, February...December).

[0061] (3) Seasonal smoothing: For each periodic subsequence, Loess smoothing (window ns) is used to obtain the preliminary seasonal term St(k).

[0062] (4) Deseasonalization: Calculate the deseasonalized sequence Yt-St(k).

[0063] (5) Trend smoothing: For the deseasonal series, smooth it with Loess (window nt) and update the trend term Tt(k).

[0064] (6) Low-pass filter adjustment: Apply low-pass filter (window nl) to the trend term Tt(k) to correct the excessive fluctuation of the seasonal term. Repeat the above steps until the trend term and seasonal term converge (usually the number of inner loop iterations is 1 to 5).

[0065] Step 4: Final Decomposition Result. After the outer loop completes, the final three components are obtained: Trend (Trend) Long-term variation patterns of the coal slurry water system (e.g., slow increase in feed concentration, and increasing dosage due to equipment aging). Seasonal Periodic fluctuations (such as fluctuations in dosage caused by the feeding rhythm per shift / day and equipment start-up / shutdown cycles); Residual term ): Random disturbances (such as instantaneous changes in coal slime particle size, fluctuations in reagent concentration, and measurement noise).

[0066] In some embodiments of this disclosure, to improve the accuracy of the prediction algorithm, a combination of time-series features and process features is adopted to construct trend feature sets, seasonal feature sets, and residual feature sets, respectively. Training and testing sets are then used for model training and accuracy verification. Specifically, based on the data collected in this example, when constructing the trend prediction feature set, the following can be used: the moving average of feed concentration (5min / 10min / 30min / 1h / 3h), the cumulative value of feed flow rate (5min / 10min / 30min / 1h / 3h / 8h), the sedimentation interface depth of the thickener, the overflow turbidity, the moving average of underflow concentration (5min / 10min / 30min / 1h / 3h), and the historical dosing trend (5min / 10min / 30min / 1h average, etc.). This constructs a dataset with 10-15 dimensions for trend prediction. When constructing the seasonal feature set, a dataset with 8-12 dimensions can be built using hourly features, shift features (morning / noon / evening), historical values ​​of the same period (dosage administered at the same time the previous day), and fluctuation range within the period, to perform seasonal prediction, mainly reflecting cyclical patterns. When constructing the residual feature set, data such as residual fluctuations in the previous 5 minutes, sudden changes in feed concentration, drug concentration deviations, and measurement equipment error range markings can be used. Since residuals are considered noise, the feature dimensions should be kept between 5-8 dimensions, not too many, mainly to capture random perturbations.

[0067] In some embodiments of this disclosure, LightGBM coal slurry water dosing prediction modeling is used. Modeling dataset partitioning: When partitioning the dataset, it is strictly done in chronological order, with a ratio of 70% training set, 20% validation set, and 10% test set. Random partitioning is prohibited to avoid data leakage. Model training: During model training, the principle of using different parameters for different components is followed. LightGBM models are trained separately for each of the three components for dosing prediction.

[0068] In the actual prediction process, based on the trained model, predictions are made from three perspectives: trend, seasonality, and residual. The three prediction results are added together to obtain the final spatial domain predicted dosage.

[0069] In some embodiments of this disclosure, the production process monitoring input parameters are input into a frequency domain dosing prediction model to obtain the frequency domain predicted dosing amount. One possible approach is to use variational mode decomposition (VMD) to decompose the production process monitoring input parameters in the frequency domain, obtaining multiple stationary mode components with finite bandwidth; based on a temporal convolutional network (TCN), multi-input prediction sub-models are established for each of the multiple stationary mode components to output multiple sub-model prediction results; the prediction results of the multiple sub-models are reconstructed to obtain the frequency domain predicted dosing amount. Specifically, when performing frequency domain-based dosing prediction for coal slurry water treatment, considering the nonlinear, non-stationary, and strongly hysteretic characteristics of coal slurry water dosing, VMD is used to decompose the dosing time series into multiple stationary mode components in the frequency domain; then, a TCN temporal convolutional network is used to establish a multi-input prediction model for each mode component; finally, the prediction results of each mode are reconstructed to obtain the final predicted dosing amount.

[0070] Specifically, VMD solves variational problems to... It is decomposed into K modal components IMFk(t), and each IMF (the decomposed frequency data) corresponds to a center frequency. It satisfies the formula: ; ; The above formula represents the objective function and constraints of variational mode decomposition.

[0071] Let be a set of modal components to be solved (i.e., the IMF components decomposed by VMD, corresponding to IMFk(t)). There are K components in total; Given a set of center frequencies to be solved, each modal component... The corresponding center angular frequency; Represents the set of all possible modal components and center frequency set Find the minimum value of the objective function. Summation term This means calculating the target term for each of the K modal components and then summing the results. The core idea of ​​VMD is to decompose the signal into K narrowband modes. Here, the "bandwidth" index of each mode is summed and then minimized as a whole.

[0072] norm , which represents the square of the L2 norm (also called the energy norm). It is used to measure the energy of a signal. In VMD, it is used to constrain the bandwidth of each mode, so that the decomposed modes are as narrowband signals as possible.

[0073] Time partial derivative , indicating time Find the partial derivative, that is After differentiating the signal, the energy of the high-frequency components is amplified. Therefore, the squared L2 norm of the signal after differentiation is essentially a measure of the signal's "bandwidth" (frequency spread), making the modes narrower.

[0074] Analytical signal construction: This part is a key step in VMD, used to construct the analytic signals of the modal components. : Unit impulse function (Dirac delta function). Imaginary unit ( ). : Kernel function of Hilbert transform. Convolution operation.

[0075] Overall meaning: ,in It's the Hilbert transform, and the result is... The analytical signal (transforming a real signal into a complex signal, retaining only the positive frequency components).

[0076] Frequency modulation term , a complex exponential signal, is used to frequency-shift analytic signals. It shifts the frequency spectrum of an analytic signal from an exponential frequency to a frequency-shifting frequency. The center position is moved to near zero frequency. After this processing, the L2 norm of the derivative can be calculated to accurately measure the position of the mode at the center frequency. The bandwidth in the vicinity.

[0077] Constraints middle Given the original input signal to be decomposed, all the decomposed modal components... The sum of must equal the original signal. This is used to ensure that the decomposition is distortion-free and that the information of the original signal is not lost.

[0078] The goal of VMD is to find a set of modal components and their center frequencies under the constraint that "the sum of all modal components equals the original signal", such that the sum of the bandwidths of each modal component near its center frequency (measured by the square of the L2 norm of the derivative after shifting the analytic signal to zero frequency) is minimized, thus obtaining a set of narrowband, non-overlapping modal components.

[0079] When performing VMD decomposition on time-series data of pesticide dosage, the core parameters are selected as follows: K is the number of modes, representing the number of modes into which the original signal needs to be decomposed. The dosage of chemical dosing in coal slurry mainly includes low-frequency trend, mid-frequency period, and high-frequency noise. K is optimally selected as 5, and 4-6 is usually recommended.

[0080] The rationality verification of IMF components includes frequency domain verification and stationarity verification. Frequency domain verification involves performing an FFT (Fourier Transform) on each IMF to check if the center frequencies are distributed from low to high (IMF1 = low frequency, IMF5 = high frequency). If this is not met, the decomposition is unreasonable and the parameters need to be reset for further decomposition. Stationarity verification involves testing the ADF (Advanced Derivative Function) of each IMF component after VMD decomposition. If the p-value is less than 0.05, it indicates that each component is a stationary sequence, meeting the time series prediction model's requirement for data stationarity, and can be used for subsequent TCN (Tracking Channel Network) modeling. If this is not met, the parameters need to be reset for further decomposition.

[0081] Because the TCN model is sensitive to the dimensions of the data, the monitoring data (concentration, flow rate, interface height, etc.) and the dosage data (flow rate) need to be normalized. To reduce the computational load, the dosage data normalization is performed after VMD decomposition, using the extreme value normalization method, and retaining scaler_imf (scale information, i.e., the extreme values ​​of the monitoring data and individual components, used for subsequent inverse normalization operations).

[0082] The TCN model used in this disclosure is constructed according to the structure and parameters shown in Table 1 below:

[0083] Table 1 The training process of the frequency domain dosage prediction model is as follows: Modeling dataset partitioning: When partitioning the dataset, strictly follow the time order and divide it into a ratio of 70% training set, 20% validation set, and 10% test set. Random partitioning is prohibited to avoid data leakage.

[0084] Model Training: During model training, the principle of using different parameters for different IMF components is followed. The core training parameters are set as follows: When training on the low-frequency IMF (IMF1 / 2), the learning rate is set to Epochs are set to 100 and early stop patience is set to 10, because low-frequency signals represent trend and periodic components and require stable fitting.

[0085] When training on high-frequency IMFs (IMF3 / 4 / 5), the learning rate is set to 5×. The epochs were set to 50 and the early stopping patience was set to 5, because high-frequency signals represent noise and overfitting needs to be strictly controlled.

[0086] When performing submodal prediction, the test set is input into the TCN to obtain normalized IMF prediction values. These are then denormalized using the saved scaler_imf to recover the original IMF amplitudes, i.e., each IMF component. In actual production, the monitored data is processed according to data processing rules and input into the trained TCN model to obtain the predicted dosage for each component. By summing the prediction values ​​of all IMFs, the final frequency domain predicted dosage is obtained.

[0087] In some embodiments of this disclosure, the target dosage prediction value is determined based on the spatial domain predicted dosage, a first weight, the frequency domain predicted dosage, and a second weight. One possible approach is to divide historical measured dosage data into rate period data and forecast period data; based on the measured values ​​of the rate period data, the spatial domain predicted dosage, and the frequency domain predicted dosage, calculate the first marginal likelihood corresponding to the spatial domain dosage prediction model and the second marginal likelihood corresponding to the frequency domain dosage prediction model; calculate the posterior probability of the spatial domain dosage prediction model based on the first marginal likelihood and the first prior probability corresponding to the spatial domain dosage prediction model, and use it as the first weight; calculate the posterior probability of the frequency domain dosage prediction model based on the second marginal likelihood and the second prior probability corresponding to the frequency domain dosage prediction model, and use it as the second weight; and perform a weighted average based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight to obtain the target dosage prediction value.

[0088] Specifically, the core of the Bayesian model averaging method is to assign reasonable weights to different models and obtain more robust prediction results by weighted averaging. In the prediction of coal slurry sedimentation dosage in this disclosure, two models, "STL+LightGBM" and "VMD+TCN", are established. Weights are assigned to them according to their prediction accuracy on historical data (the higher the accuracy, the greater the weight). The final prediction value is obtained by weighted summation of the prediction values ​​of the two models.

[0089] Specifically, the data is first divided into rate period data (used to calculate model weights) and validation / forecast period data (used to validate the BMA effect and output the final forecast; this data was prepared in step 2). Rate period data M: Select a complete historical data segment, denoted as M. ; in, Let be the measured value of the dosage at time t. The total number of samples for periodic data.

[0090] Forecast period data: New samples that require prediction of dosage (such as coal slurry water characteristic data for a future period of time).

[0091] Obtain the forecast results of the two models at the rate period.

[0092] Using the two pre-trained models, predict the dosage for each input feature at the rate interval, and obtain the dosage prediction sequence for each model: Spatial domain dosing prediction model (i.e., Model 1): Prediction sequence ,in This is the predicted dosage of medicine from Model 1 at time t; Frequency domain dosage prediction model (i.e., Model 2): ​​Prediction sequence ,in It is the predicted dosage of medicine in Model 2 at time t.

[0093] In practical applications, it is necessary to ensure that the prediction results of the two models correspond one-to-one with the measured values ​​(time / sample dimensions are completely matched) to avoid misalignment.

[0094] Furthermore, the marginal likelihood of the two models is calculated. Marginal likelihood is "how well the model fits the measured data." In the scenario of pesticide dosage forecasting, it is preferentially calculated using the mean absolute error (MAE), and the steps are as follows: Calculate the MAE for each model. The smaller the MAE, the more accurate the model prediction, and the higher the marginal likelihood. The calculation formula is as follows:

[0095]

[0096] Transform MAE into marginal likelihood .

[0097] The marginal likelihood needs to be a positive value and negatively correlated with MAE (the smaller the MAE, the larger the value). It is usually processed using a common exponential transformation.

[0098] in, (Corresponding to two models); α is the adjustment coefficient (it is recommended to start with 1. If the MAE value is too small / too large, it can be adjusted to 0.1 or 10. For example, take 1 if the MAE is between 0.1 and 1, and take 0.1 if the MAE is between 10 and 100). It is a natural exponential function.

[0099] Set the prior probabilities of the model Prior probability is "the degree to which the model is considered reliable when there is no actual measured data." If there is no industry experience / model preference, take the equal prior probability (most commonly used): Prior probabilities of Model 1: ; Prior probabilities of Model 2: .

[0100] Calculate the weights (posterior probabilities) of the two models. According to Bayes' theorem, the weight is "the probability that the model is the optimal model given the actual measured data", and the formula is as follows:

[0101] The formula for calculating the first weight is as follows: ; The formula for calculating the second weight is as follows: ; By using the weights of the two models, a weighted average is calculated on the predicted dosage values ​​for the forecast period to obtain the final target dosage prediction value: .

[0102] in, Predict the dosage of pesticide in the spatial domain of the forecast period sample; Predict the dosage of pesticide for the frequency domain of the forecast period sample; : Target dosage prediction value of the BMA weighted model.

[0103] The prediction data obtained here This is the final predicted data, which can be used to adjust the dosage in actual production.

[0104] Figure 2 This is a schematic diagram of the structure of a multi-model collaborative prediction device 20 for chemical dosage in coal slurry water treatment, provided as an exemplary embodiment of this disclosure. Figure 2 As shown, the multi-model collaborative prediction device 20 for coal slurry water treatment dosage includes: an acquisition module 21, a cleaning module 22, a first prediction module 23, a second prediction module 24, and a determination module 25.

[0105] Among them, the acquisition module 21 is used to acquire the original production process monitoring parameters of the coal slurry water settling stage; The cleaning module 22 is used to perform data cleaning operations on the original production process monitoring parameters to obtain the production process monitoring input parameters; The first prediction module 23 is used to input the production process monitoring input parameters into the spatial domain dosing prediction model to obtain the spatial domain predicted dosing amount. The second prediction module 24 is used to input the production process monitoring input parameters into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage. The determination module 25 is used to determine the target dosage prediction value based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight, wherein the first weight is the weight corresponding to the spatial domain predicted dosage, and the second weight is the weight corresponding to the frequency domain predicted dosage.

[0106] Optionally, the original production process monitoring parameters include: feed flow rate, feed concentration, pH value, underflow flow rate, underflow concentration, overflow turbidity, and settling interface depth; when acquiring the original production process monitoring parameters of the coal slurry water settling stage, the acquisition module 21 is used for: The feed flow rate is collected by a flow monitoring device installed on the feed pipeline of the thickener. The feed concentration is collected by a concentration detection device installed on the feed pipe of the thickener. pH values ​​were collected by installing a pH detector in the middle of the thickening tank; The underflow flow rate is collected by installing a flow monitoring device in the underflow pipe at the bottom of the thickener. The concentration of the underflow is collected by installing a concentration detection device in the underflow pipe at the bottom of the thickener. Turbidity of overflow water was collected by installing a turbidity meter in the overflow trough of the thickening tank. The depth of the settling interface was collected by setting up a sludge interface meter in the middle of the thickening tank.

[0107] Optionally, when the first prediction module 23 inputs the production process monitoring input parameters into the spatial domain dosage prediction model to obtain the spatial domain predicted dosage, it is used for: By inputting the production process monitoring input parameters into the spatial domain dosing prediction model, trend prediction results, seasonal prediction results, and residual prediction results are obtained. Based on the trend forecast, seasonal forecast, and residual forecast results, the spatial domain predicted dosage is calculated.

[0108] Optionally, when the first prediction module 23 inputs the production process monitoring input parameters into the spatial domain dosage prediction model and obtains trend prediction results, seasonal prediction results, and residual prediction results, it is used for: Based on the cycle length, seasonal smoothing window width, trend smoothing window width, and low-pass filter window width, the input parameters for production process monitoring are centrally preprocessed to obtain preprocessed data. Enter the outer loop iteration, initialize the data weights, and execute the inner loop; the inner loop includes: smoothing the detrended sequence by periodic grouping to obtain the preliminary seasonal term, smoothing the deseasonal sequence to update the trend term, and performing low-pass filtering correction on the trend term until the trend term and seasonal term converge. In the outer loop, the absolute median difference of the residuals is calculated, and the data weights are updated based on the weight function until the preset number of iterations is reached or the weights converge. The final output includes trend forecasts, seasonal forecasts, and residual forecasts.

[0109] Optionally, when the second prediction module 24 inputs the production process monitoring input parameters into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage, it is used for: Variational mode decomposition is used to perform frequency domain decomposition on the input parameters for production process monitoring, resulting in multiple stationary mode components with finite bandwidth. Based on a temporal convolutional network, a multi-input prediction sub-model is established for multiple stationary mode components to output the prediction results of multiple sub-models. The prediction results of multiple sub-models are reconstructed to obtain the frequency domain predicted dosage.

[0110] Optionally, when determining the target dosage prediction value based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight, the determining module 25 is used to: Historical measured dosage data are divided into rate-period data and forecast-period data; Based on the measured values ​​of the periodic data, the spatial domain predicted dosage and the frequency domain predicted dosage, the first marginal likelihood corresponding to the spatial domain dosage prediction model and the second marginal likelihood corresponding to the frequency domain dosage prediction model are calculated. Based on the first marginal likelihood and the first prior probability corresponding to the spatial domain dosage prediction model, the posterior probability of the spatial domain dosage prediction model is calculated and used as the first weight. Based on the second marginal likelihood and the second prior probability corresponding to the frequency domain dosage prediction model, the posterior probability of the frequency domain dosage prediction model is calculated and used as the second weight. The target dosage prediction value is obtained by weighting the dosage predicted in the spatial domain, the first weight, the dosage predicted in the frequency domain, and the second weight.

[0111] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0112] Figure 3 This is a schematic diagram of the structure of an electronic device provided as an exemplary embodiment of the present disclosure. For example... Figure 3 As shown, the electronic device includes a memory 31 and a processor 32. Additionally, the electronic device also includes a power supply component 33 and a communication component 34.

[0113] Memory 31 is used to store computer programs and can be configured to store various other data to support operation on the electronic device. Examples of this data include instructions for any application or method used to operate on the electronic device.

[0114] The memory 31 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0115] Communication component 34 is used for data transmission with other devices.

[0116] The processor 32 can execute computer instructions stored in the memory 31 to: acquire the original production process monitoring parameters of the coal slurry sedimentation stage; perform data cleaning operations on the original production process monitoring parameters to obtain the production process monitoring input parameters; input the production process monitoring input parameters into the spatial domain dosage prediction model to obtain the spatial domain predicted dosage; input the production process monitoring input parameters into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage; and determine the target dosage prediction value based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight, wherein the first weight is the weight corresponding to the spatial domain predicted dosage, and the second weight is the weight corresponding to the frequency domain predicted dosage.

[0117] Accordingly, embodiments of this disclosure also provide a computer-readable storage medium storing a computer program. When the computer-readable storage medium stores a computer program, and the computer program is executed by one or more processors, it causes one or more processors to perform... Figure 1 Each step in the method embodiment.

[0118] Accordingly, embodiments of this disclosure also provide a computer program product, which includes a computer program / instructions that are executed by a processor. Figure 1 Each step in the method embodiment.

[0119] The above Figure 3 The communication component is configured to facilitate wired or wireless communication between the device containing the communication component and other devices. The device containing the communication component can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component also includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID), Infrared Data Association (IrDA) technology, Ultra-Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0120] The above Figure 3 The power supply component provides power to the various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which it resides.

[0121] The aforementioned electronic devices also include a display screen and audio components.

[0122] The display includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of touch or swipe actions, but also the duration and pressure associated with the touch or swipe operation.

[0123] An audio component may be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, or voice recognition mode. The received audio signals may be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0124] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0126] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0127] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0128] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0129] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0130] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0131] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0132] The above are merely specific embodiments of this disclosure, enabling those skilled in the art to understand or implement this disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-model collaborative prediction method for chemical dosage in coal slurry water treatment, characterized in that, include: Obtain the original production process monitoring parameters for the coal slurry water settling stage; Data cleaning is performed on the original production process monitoring parameters to obtain the production process monitoring input parameters; The production process monitoring input parameters are input into the spatial domain dosing prediction model to obtain the spatial domain predicted dosing amount; The production process monitoring input parameters are input into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage. The target dosage prediction value is determined based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight, wherein the first weight is the weight corresponding to the spatial domain predicted dosage, and the second weight is the weight corresponding to the frequency domain predicted dosage.

2. The method according to claim 1, characterized in that, The original production process monitoring parameters include: feed flow rate, feed concentration, pH value, underflow flow rate, underflow concentration, overflow turbidity, and settling interface depth; the original production process monitoring parameters for obtaining the coal slurry water settling stage include: The feed flow rate is collected by a flow monitoring device installed on the feed pipeline of the thickener. The feed concentration is collected by a concentration detection device installed on the feed pipe of the thickener. The pH value was collected by installing a pH detector in the middle of the thickening tank; The underflow flow rate is collected by installing a flow monitoring device in the underflow pipe at the bottom of the thickener. The concentration of the underflow is collected by installing a concentration detection device in the underflow pipe at the bottom of the thickener. The turbidity of the overflow water was collected by installing a turbidity meter in the overflow trough of the thickening tank. The settling interface depth was collected by installing a sludge interface meter in the middle of the thickening tank.

3. The method according to claim 1, characterized in that, The step of inputting the production process monitoring input parameters into the spatial domain dosing prediction model to obtain the spatial domain predicted dosing includes: The production process monitoring input parameters are input into the spatial domain dosing prediction model to obtain trend prediction results, seasonal prediction results, and residual prediction results. Based on the trend prediction results, the seasonal prediction results, and the residual prediction results, the spatial domain predicted dosage is calculated.

4. The method according to claim 3, characterized in that, The process involves inputting the production process monitoring input parameters into a spatial domain dosing prediction model to obtain trend prediction results, seasonal prediction results, and residual prediction results, including: Based on the cycle length, seasonal smoothing window width, trend smoothing window width, and low-pass filter window width, the production process monitoring input parameters are centrally preprocessed to obtain preprocessed data. Enter the outer loop iteration, initialize the data weights, and execute the inner loop; the inner loop includes: smoothing the detrended sequence by periodic grouping to obtain the preliminary seasonal term, smoothing the deseasonal sequence to update the trend term, and performing low-pass filtering correction on the trend term until the trend term and seasonal term converge. In the outer loop, the absolute median difference of the residuals is calculated, and the data weights are updated based on the weight function until the preset number of iterations is reached or the weights converge. The final decomposition results are output as follows: the trend prediction result, the seasonal prediction result, and the residual prediction result.

5. The method according to claim 1, characterized in that, The step of inputting the production process monitoring input parameters into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage includes: Variational mode decomposition is used to perform frequency domain decomposition on the input parameters of the production process monitoring to obtain multiple stationary mode components with finite bandwidth. Based on a temporal convolutional network, a multi-input prediction sub-model is established for each of the multiple stationary mode components to output the prediction results of multiple sub-models. The prediction results of multiple sub-models are reconstructed to obtain the frequency domain predicted dosage.

6. The method according to claim 1, characterized in that, The step of determining the target dosage prediction value based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight includes: Historical measured dosage data are divided into rate-period data and forecast-period data; Based on the measured values ​​of the periodic data, the spatial domain predicted dosage, and the frequency domain predicted dosage, calculate the first marginal likelihood corresponding to the spatial domain dosage prediction model and the second marginal likelihood corresponding to the frequency domain dosage prediction model. Based on the first marginal likelihood and the first prior probability corresponding to the spatial domain dosage prediction model, the posterior probability of the spatial domain dosage prediction model is calculated and used as the first weight. Based on the second marginal likelihood and the second prior probability corresponding to the frequency domain dosage prediction model, the posterior probability of the frequency domain dosage prediction model is calculated and used as the second weight. The target dosage prediction value is obtained by weighting the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight.

7. A multi-model collaborative prediction device for chemical dosage in coal slurry water treatment, characterized in that, include: The acquisition module is used to acquire the original production process monitoring parameters of the coal slurry water settling stage; The cleaning module is used to perform data cleaning operations on the original production process monitoring parameters to obtain the production process monitoring input parameters; The first prediction module is used to input the production process monitoring input parameters into the spatial domain dosing prediction model to obtain the spatial domain predicted dosing amount. The second prediction module is used to input the production process monitoring input parameters into the frequency domain dosage prediction model to obtain the frequency domain predicted dosage. The determination module is used to determine the target dosage prediction value based on the spatial domain predicted dosage, the first weight, the frequency domain predicted dosage, and the second weight, wherein the first weight is the weight corresponding to the spatial domain predicted dosage, and the second weight is the weight corresponding to the frequency domain predicted dosage.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to execute instructions to implement the steps of the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

10. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-6.

Images