A Smart Control Method for Filter Presses Based on Multi-Source Data

Abstract

This invention relates to an intelligent control method for a filter press based on multi-source data, belonging to the field of sludge treatment technology. The method constructs a theoretical instantaneous filtration rate that accurately reflects the state of the filter press operation scenario and uses it as a benchmark parameter. Then, it determines the correlation between the values ​​of each state parameter and the benchmark parameter at each sampling moment, as well as the time lag in their value changes. This quantifies the dynamic weight of each state parameter during multi-source data fusion in the continuous control of the filter press, ensuring that the most effective and timely responding state parameter is prioritized to control the filter press when operating conditions change. This effectively captures the timing characteristics of abnormal pressure data preceding flow data, such as when the filter cloth is clogged, fundamentally solving the rigidity of fixed weight allocation in filter press scenarios and improving the response speed and control accuracy of the filter press.

Description

Technical Field

[0001] This invention relates to the field of sludge water treatment technology, and in particular to an intelligent control method for filter presses based on multi-source data. Background Technology

[0002] Filter presses, as core equipment for solid-liquid separation, are widely used in wastewater treatment and coal slurry water treatment. Their control precision directly affects the filter cake moisture content and energy consumption. Traditional control methods for filter presses primarily rely on sensor data such as pressure and flow rate, using preset thresholds for start-up and shutdown control. With the development of multi-source data acquisition technology, existing control methods are beginning to integrate data from multiple sensors, including pressure, flow rate, and temperature, employing weighted averaging or fixed rules for decision-making.

[0003] However, in actual dynamic operating conditions, the importance of each data source varies significantly over time. For example, when the filter cloth becomes slightly clogged, pressure data will fluctuate abnormally before flow data. This is because the pressure sensor, which directly monitors the filter chamber pressure, is most sensitive to clogging and often shows abnormal fluctuations first. The flow sensor, installed in the feed pipe, has a certain delay in responding to anomalies. Existing fixed-weight fusion methods completely ignore this crucial characteristic of the time-series differences in anomaly responses from different data sources, still fusing them according to a preset ratio. This leads to a situation where, in the early stages of clogging, when pressure data already shows anomalies, the system fails to adjust in time due to the excessively high weight of flow data. When the clogging becomes severe, the flow data only then shows anomalies, at which point adjustment is too late, ultimately resulting in delayed control decisions. Therefore, the rigid weight allocation problem in existing scenarios not only affects the timeliness of filter press control but also accelerates filter cloth wear and increases energy consumption.

[0004] In summary, existing filter press control methods suffer from technical problems such as slow response speed and insufficient control accuracy. Summary of the Invention

[0005] In view of this, the present invention provides an intelligent control method for filter presses based on multi-source data to solve the technical problems of slow response speed and low control accuracy of existing filter press control methods.

[0006] The present invention provides an intelligent control method for a filter press based on multi-source data, comprising:

[0007] During the operation of the filter press, various state parameters reflecting the filtration state are continuously sampled, and the theoretical instantaneous filtration rate of the filter press at each sampling moment is calculated as the reference parameter at the corresponding sampling moment.

[0008] Using the historical time period with the first duration closest to the current sampling time as the first analysis window, the correlation degree between any state parameter and the benchmark parameter is calculated under the first analysis window to obtain the correlation coefficient between any state parameter and the benchmark parameter at the current sampling time.

[0009] Using the historical time period of the second duration closest to the current sampling time as the second analysis window, the cross-correlation degree between the state parameter and the reference parameter is determined under different time offsets relative to the reference parameter within the second analysis window. The time offset corresponding to the maximum cross-correlation degree is used as the delay time of the state parameter relative to the reference parameter at the current sampling time. Based on the delay time of the state parameter relative to the reference parameter at the current sampling time, a delay time correction factor for the state parameter at the current sampling time is constructed. The time offset can be positive or negative and its absolute value is less than the second duration.

[0010] The dynamic weight of each state parameter at the current sampling time is determined by the correlation coefficient and the delay time correction factor corresponding to each state parameter at the current sampling time, and the intelligent control of the filter press is completed based on the dynamic weight of each state parameter at the current sampling time.

[0011] Furthermore, the state parameters include filter pressure, filter flow rate, and filter temperature.

[0012] Furthermore, the calculation of the theoretical instantaneous filtration rate of the filter press at each sampling time includes:

[0013] The absolute value of the difference between the upstream pressure and the downstream pressure of the filter cloth of the filter press at any sampling time is calculated as the instantaneous pressure difference at any sampling time, and the cumulative volume of filtrate at any sampling time is the third historical time period closest to the sampling time.

[0014] The theoretical instantaneous filtration rate of the filter press at any given sampling time is directly proportional to the filtration area and the instantaneous pressure difference at any given sampling time, and inversely proportional to the dynamic viscosity, filter cloth resistance, filter cake specific resistance, feed concentration, and cumulative filtrate volume at any given sampling time.

[0015] Furthermore, obtaining the correlation coefficient between any state parameter and the reference parameter at the current sampling time includes:

[0016] The sequence formed by the values ​​of any state parameter under the first analysis window at each sampling time is denoted as the first sequence, and the sequence formed by the values ​​of the benchmark parameter under the first analysis window at each sampling time is denoted as the second sequence. The absolute value of the Pearson correlation coefficient between the first sequence and the second sequence is denoted as the correlation coefficient between any state parameter and the benchmark parameter at the current sampling time.

[0017] Furthermore, calculating the delay time of any state parameter relative to the reference parameter at the current sampling time includes:

[0018] ,

[0019] in, This represents the delay time of the i-th state parameter relative to the reference parameter at the current sampling time. This represents the function that maximizes the parameters, used to find the time offset corresponding to the maximum cross-correlation. Indicates the time offset. This represents the total number of downsampling times in the second analysis window. Indicates the second analysis window The value of the i-th state parameter at each sampling time. Indicates the second analysis window The values ​​of the reference parameters at each sampling time point Indicates request and The inner product of.

[0020] Furthermore, the construction of the delay time correction factor for any state parameter at the current sampling time includes:

[0021] The ratio of the absolute value of the delay time of any state parameter at the current sampling time compared to the reference parameter to the preset maximum acceptable delay time is recorded as the delay level. The negative number of the delay level is input into the natural exponential function, and the output result is used as the delay time correction factor of any state parameter at the current sampling time.

[0022] Furthermore, determining the dynamic weight of each state parameter at the current sampling time includes:

[0023] The correlation coefficient corresponding to any state parameter at the current sampling time is weighted by a first weight to obtain a first weighted value. The delay time correction factor of any state parameter at the current sampling time is weighted by a second weight to obtain a second weighted value. The sum of the first weighted value and the second weighted value is input into the natural exponential function, and the output result is used as the absolute weight of any state parameter at the current sampling time.

[0024] The ratio of the absolute weight of any state parameter at the current sampling time to the sum of the absolute weights of all state parameters at the current sampling time is used as the dynamic weight of any state parameter at the current sampling time.

[0025] Furthermore, the intelligent control of the filter press based on the dynamic weights of each state parameter at the current sampling time includes:

[0026] The value of any state parameter at the current sampling time corresponding to the value of the control parameter to be controlled is recorded as the calculated value of the control quantity of any state parameter at the current sampling time. The product of the dynamic weight of any state parameter at the current sampling time and the calculated value of the control quantity is taken as the weighted calculated value of the control quantity of any state parameter at the current sampling time. The sum of the weighted calculated values ​​of the control quantities of all state parameters at the current sampling time is taken as the control value of the control parameter at the current sampling time.

[0027] Furthermore, the control parameters include the opening degree of the feed valve of the filter press and the filter pressure.

[0028] The advantages of this invention compared to the prior art are:

[0029] This invention constructs a benchmark parameter—the theoretical instantaneous filtration rate—based on the characteristics of filter press operation scenarios, which accurately reflects the scenario state. Then, it determines the correlation and time lag of each state parameter relative to the benchmark parameter at each sampling moment. This quantifies the dynamic weight of each state parameter during multi-source data fusion in the continuous control of the filter press, ensuring that the most effective and timely responding state parameter is prioritized to control the filter press when operating conditions change. This effectively captures the timing characteristics of abnormal pressure data preceding flow data, such as when the filter cloth is clogged. It fundamentally solves the rigidity of fixed weight allocation in current scenarios, improving the response speed and control accuracy of the filter press. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart illustrating an intelligent control method for a filter press based on multi-source data, provided in Embodiment 1 of the present invention. Detailed Implementation

[0032] The overall concept of this invention is as follows:

[0033] This invention selects a theoretical value calculated using the inherent parameters and basic physical model of the filter press equipment. Based on the characteristic that this theoretical value is not affected by the sensor placement and is directly linked to the filter press separation effect or the core energy consumption of the equipment, the correlation strength and relative delay time between various data sources and the selected benchmark parameters are evaluated. Based on the correlation strength and relative delay time, the weights of various data sources are dynamically and adaptively allocated to ensure that the data source with the most timely response is given priority to dominate the control when the operating conditions change, thereby improving the response speed and control accuracy of the filter press.

[0034] To further illustrate the technical solution of the present invention, specific embodiments are described below.

[0035] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a particular feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. Furthermore, a particular feature, structure, or characteristic in one or more embodiments may be combined in any suitable form, and the terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless otherwise specifically emphasized.

[0036] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0037] Method Implementation Examples:

[0038] See Figure 1 This is a flowchart illustrating an intelligent control method for a filter press based on multi-source data, as provided in Embodiment 1 of the present invention. Figure 1 As shown, the method may include the following steps:

[0039] S101, during the operation of the filter press, continuously sample various state parameters reflecting the filtration state, and calculate the theoretical instantaneous filtration rate of the filter press at each sampling time as the reference parameter at the corresponding sampling time.

[0040] To ensure full control of the filter press throughout its operation, it is necessary to deploy relevant sensors and data processing units to continuously collect parameters during operation. The collected parameters should include at least several state parameters reflecting the filter press's filtration status, such as filtration pressure, filtration flow rate, and filtration temperature. The following outlines the necessary configuration and process for acquiring and preprocessing these various state parameters:

[0041] 1) Data Acquisition System Configuration:

[0042] Pressure sensor: A piezoresistive pressure sensor is selected, with a range of 0-2.5MPa, an accuracy of ±0.1%FS, and a sampling frequency of 10Hz. It is installed at key positions at the inlet and outlet of the filter chamber of the filter press to collect the pressure of the filter chamber or the upstream and downstream of the filter cloth.

[0043] Electromagnetic flow meter: measuring range 0-12m³ / h, accuracy ±0.5%, sampling frequency 1Hz, installed in the feed pipe.

[0044] Temperature sensor: PT100 platinum resistance thermometer, measuring range 0-200℃, accuracy ±0.5℃, sampling frequency 1Hz, installed on the surface of the filter plate.

[0045] Data acquisition card: 16-bit resolution, supports multi-channel synchronous sampling, timestamp accuracy 1ms.

[0046] (ii) Data preprocessing workflow:

[0047] Outlier handling: using The criteria remove obviously outlier data, and linear interpolation is used to fill in consecutive outliers.

[0048] Data alignment: Based on the pressure data time axis, multi-source data time synchronization is achieved through cubic spline interpolation. Since the pressure data sampling frequency is 10Hz, the corresponding sampling interval is 0.1s, which means that the time interval between adjacent sampling moments in the data formed after data alignment of any state parameter is 0.1s.

[0049] Noise filtering: Pressure data uses a Butterworth low-pass filter (cutoff frequency 5Hz), while flow and temperature data use a moving average filter (window size 10 points).

[0050] Data normalization: The minimum-maximum normalization method is used to map the data collected by each sensor to the [0,1] interval.

[0051] (iii) Data quality assessment:

[0052] Real-time calculation of the signal-to-noise ratio (SNR) of various data sources, i.e., various state parameters, when SNR A data quality warning is triggered when the value drops by 20dB.

[0053] Monitor the missing data rate, when the missing rate The data reconstruction process will be initiated when the threshold of 5% is reached.

[0054] The above setup, through a rigorous preprocessing procedure, ensures good consistency and reliability of multi-source data, laying the foundation for subsequent analysis.

[0055] Analysis reveals that the occurrence times of anomalies in different data sources, i.e., different state parameters, vary when the operating conditions of the filter press change. This temporal variation reflects the different sensitivities of the data sources to specific faults. Therefore, this embodiment aims to quantify the response characteristics of each data source during the process of changing operating conditions by constructing a temporal sensitivity evaluation model, and to dynamically optimize the weights during the multi-source data fusion process.

[0056] Prior to this, to accurately quantify the time differences and magnitudes of anomaly responses from various data sources, it is first necessary to obtain a valid benchmark parameter for the scenario. In the context of a filter press, a truly effective benchmark parameter should be directly linked to the separation effect or the core energy consumption of the equipment. Therefore, this embodiment selects the "theoretical instantaneous filtration rate" as the benchmark parameter after analysis, as it is the most ideal and physically consistent benchmark parameter. This parameter is not a directly measured value, but a theoretical value calculated based on the inherent parameters and basic physical model of the filter press equipment. Furthermore, due to changes in the state of the filter press equipment and the filtration state, the value of this benchmark parameter, the theoretical instantaneous filtration rate, varies at different sampling times. Therefore, it is necessary to calculate the theoretical instantaneous filtration rate of the filter press at each sampling time, including:

[0057] The absolute value of the difference between the upstream pressure and the downstream pressure of the filter cloth of the filter press at any sampling time is calculated as the instantaneous pressure difference at any sampling time, and the cumulative volume of filtrate at any sampling time is the third historical time period closest to the sampling time.

[0058] The theoretical instantaneous filtration rate of the filter press at any given sampling time is directly proportional to the filtration area and the instantaneous pressure difference at any given sampling time, and inversely proportional to the dynamic viscosity, filter cloth resistance, filter cake specific resistance, feed concentration, and cumulative filtrate volume at any given sampling time.

[0059] As a further preferred option, the theoretical instantaneous filtering rate at each sampling time is:

[0060]

[0061] in, This represents the theoretical instantaneous filtering rate at sampling time t, which is also the baseline parameter at sampling time t, in units of... , The filter area is a constant and a known inherent parameter of the filter press equipment, measured in units of... , This represents the real-time instantaneous pressure difference, specifically the instantaneous pressure difference at sampling time t, measured in Pa (Pascals). The instantaneous pressure difference is obtained by taking the absolute value of the difference between the measurements from two pressure sensors. One sensor is the feed pressure sensor, installed near the filter chamber inlet in the pipeline, ensuring the measurement point is the static pressure on the feed side upstream of the filter cloth. The other sensor is the filter chamber pressure sensor, installed in the filtrate outlet channel or filter plate drainage hole, ensuring the measurement point is the filtrate side pressure downstream of the filter cloth. In other words, the absolute value of the difference between the pressure upstream and downstream of the filter cloth at sampling time t is obtained. This represents dynamic viscosity, in Pa. s=kg / (m This constant is determined by the filter press equipment and the material being filtered, and is obtained through initial settings or calibration. The resistance of the filter cloth, expressed in 1 / m (per meter), is a known constant value determined through initial experiments. Specific resistance of the filter cake, measured in m / kg (meters per kilogram), is a known constant value preset based on the material properties. The mass concentration at sampling time t is the real-time feed concentration, expressed in kg / m³ (kilograms per cubic meter), and is measured by an online density meter. The cumulative filtrate volume is represented by the signal from the integral flow meter. The integral calculation yields: The unit is m³. To ensure the standardization of the calculation of the theoretical instantaneous filtration rate at different sampling times and to avoid analytical errors, the integration time in the calculation of the cumulative filtrate volume at each sampling time is set to be the same, and all of them are the third historical time period closest to the current sampling time.

[0062] The formula for the theoretical instantaneous filtration rate is based on Darcy's law, a fundamental law of seepage mechanics. In this embodiment, the benchmark parameter for the theoretical instantaneous filtration rate is an optimization of Darcy's law based on a specific scenario. Furthermore, corresponding to the normalization of various parameters of the filter press measured by the sensor as described above in this embodiment, the parameters used in the calculation of the theoretical instantaneous filtration rate are also dimensionless to ensure that subsequent calculations are dimensionless and to avoid violating the principle of dimensional consistency.

[0063] S102, taking the historical time period with the first duration closest to the current sampling time as the first analysis window, calculate the correlation degree between any state parameter and the benchmark parameter under the first analysis window, and obtain the correlation degree coefficient between any state parameter and the benchmark parameter at the current sampling time.

[0064] After successfully constructing the baseline parameters at each sampling time, since the correlation strength between each data source and the scene baseline parameters changes dynamically over time, this step uses a sliding window to calculate the dynamic correlation coefficient between each data source and the baseline parameters to capture their correlation characteristics and provide a basis for weight allocation.

[0065] To successfully analyze the correlation between any state parameter and the reference parameter, and to avoid large errors caused by the analysis data deviating too much from the current sampling time, this embodiment sets the first analysis window to the historical time period with the first duration closest to the current sampling time. The first duration can be further combined with the working cycle of the filter press, usually taking 50-100 sampling times, and the duration is 5-10 seconds for a sampling frequency of 10 Hz.

[0066] Based on the first analysis window corresponding to the current sampling time, the correlation coefficient between any state parameter and the benchmark parameter at the current sampling time can be obtained, including:

[0067] The sequence formed by the values ​​of any state parameter under the first analysis window at each sampling time is denoted as the first sequence, and the sequence formed by the values ​​of the benchmark parameter under the first analysis window at each sampling time is denoted as the second sequence. The absolute value of the Pearson correlation coefficient between the first sequence and the second sequence is denoted as the correlation coefficient between any state parameter and the benchmark parameter at the current sampling time.

[0068] Further, as a preferred embodiment, the Pearson correlation coefficient between the first sequence and the second sequence, or in other words, the Pearson correlation coefficient between any state parameter and the reference parameter at the current sampling time, is:

[0069]

[0070] in, This refers to the aforementioned time-series sensitivity assessment model, specifically representing the Pearson correlation coefficient between the i-th state parameter and the baseline parameter at the k-th sampling time. It characterizes the degree of correlation between the two types of parameters at that sampling time, with a value range of [-1, 1]. A larger absolute value indicates a stronger correlation. This represents the baseline parameter at the k-th sampling time. Indicates the first The data source contains sampled values ​​at the k-th sampling time, where the k-th sampling time is a general time index within the first analysis window. This can be understood as a loop variable used to iterate through all sampling points in the first analysis window. Indicates the first analysis window The mean, Indicates the first analysis window The mean, This refers to the first analysis window. It's important to explain the first analysis window in this embodiment. Although this embodiment selects the historical time period closest to the sampling time as the first analysis window corresponding to the sampling time, meaning the first analysis window corresponds to a time period according to this determination method, since the sampling times in this embodiment are obtained at fixed intervals, the first analysis window can be understood as a set of sampling times (or sampling points) containing the corresponding number of sampling times, considering the sampling interval. This can be understood as the total number of sampling times in the first analysis window.

[0071] Finally, after obtaining the aforementioned Pearson correlation coefficient... Then, taking its absolute value yields the correlation coefficient between the i-th state parameter and the reference parameter at time t. .

[0072] By acquiring the correlation coefficient that dynamically changes over time, the correlation between various data sources (i.e., state parameters) and the key states of the filter press can be reflected in real time, providing a quantitative basis for the weight allocation of various state parameters when controlling the filter press subsequently. When the correlation strength between a certain state parameter and the benchmark parameter increases, it indicates that it is more important under the current operating conditions and should be assigned a higher weight.

[0073] S103, using the historical time period of the second duration closest to the current sampling time as the second analysis window, the cross-correlation degree between the state parameter and the reference parameter is determined under different time offsets relative to the reference parameter in the second analysis window. The time offset corresponding to the maximum cross-correlation degree is used as the delay time of the state parameter relative to the reference parameter at the current sampling time. Based on the delay time of the state parameter relative to the reference parameter at the current sampling time, a delay time correction factor for the state parameter at the current sampling time is constructed. The time offset can be positive or negative and its absolute value is less than the second duration.

[0074] Referring to the first analysis window, in order to successfully analyze the delay time of any state parameter in response to an abnormal state compared to the baseline parameter, and to avoid the analysis data deviating too much from the current sampling time and causing large errors, this embodiment sets the second analysis window to the historical time period with the second duration closest to the current sampling time. At the same time, to ensure statistical significance, the second analysis window is set to contain 200-500 sampling times, and the duration is 20-50 seconds corresponding to a sampling frequency of 10 Hz.

[0075] The analysis operation of the delay time of any state parameter in responding to an abnormal state relative to a reference parameter is to calculate the delay time of the given state parameter relative to the reference parameter at the current sampling time, including:

[0076]

[0077] in, This represents the delay time of the i-th state parameter at the current sampling moment compared to the reference parameter, in seconds. A positive value indicates lag, and a negative value indicates lead. This represents the function that maximizes the parameter, used to find the time offset corresponding to the maximum cross-correlation. This time offset is the most likely delay time. This represents the time offset, and its preferred value range is... 30 seconds, or plus or minus 30 seconds, is the time limit for setting the search range. 30 seconds, which can cover the typical latency range of various data sources or various state parameters. This represents the total number of downsampling times in the second analysis window. Indicates the second analysis window The value of the i-th state parameter at each sampling time. Indicates the second analysis window The values ​​of the reference parameters at each sampling time point Indicates request and The inner product of.

[0078] The reason for seeking and The inner product is used because this part calculates the dot product (also known as the inner product) of the two vectors at different time offsets corresponding to their time alignment. The principle behind this is that the dot product is a fundamental indicator of the similarity between two vectors. If the amplitude changes of both vectors at the same position follow the same trend (both large or small), their dot product is positive; if the trends are opposite, the dot product is negative. Therefore, by summing the dot products at all aligned time points, if the sum is a large positive number, it indicates that at the current time offset, the two vectors are very similar in shape or have a high degree of cross-correlation. If the sum is close to zero, it indicates dissimilarity or a low degree of cross-correlation. In other words, The partially completed part is the cross-correlation between any state parameter in the second analysis window and its baseline parameter at different time offsets.

[0079] After obtaining the delay time of any state parameter relative to the reference parameter at the current sampling time, a delay time correction factor for any state parameter at the current sampling time can be further constructed, including:

[0080] The ratio of the absolute value of the delay time of any state parameter at the current sampling time compared to the reference parameter to the preset maximum acceptable delay time is recorded as the delay level. The negative number of the delay level is input into the natural exponential function, and the output result is used as the delay time correction factor of any state parameter at the current sampling time.

[0081] The formula for the delay time correction factor of any state parameter at the current sampling time is as follows:

[0082]

[0083] in, This represents the delay time correction factor for the i-th state parameter at the current sampling time. This represents the natural exponential function with base e. This represents the delay time of the i-th state parameter relative to the reference parameter at the current sampling time. This indicates taking the absolute value, and represents the preset maximum acceptable delay time (preferably 10 seconds in this embodiment, considering general scenarios). The closer a value is to 1, the smaller the delay and the better the timeliness.

[0084] The delay time correction factor can more accurately quantify the response timing characteristics of each data source, i.e., various state parameters, and ensure that data sources with timely responses receive higher weights. Therefore, based on the aforementioned correlation coefficient, the weight allocation of each state parameter involved in the control process of the filter press is also corrected by the delay time correction factor.

[0085] S104, determine the dynamic weight of each state parameter at the current sampling time by using the correlation coefficient and the delay time correction factor corresponding to each state parameter at the current sampling time, and complete the intelligent control of the filter press based on the dynamic weight of each state parameter at the current sampling time.

[0086] Combining the correlation strength coefficient and delay time correction factor calculated based on the time-series sensitivity assessment model, a dynamic weight allocation function for each state parameter can be constructed. This function must ensure that data sources with higher correlation strength and lower delay (larger delay factor) have higher weights, thus guaranteeing that the most relevant and timely data dominates the control of the filter press when operating conditions change. Therefore, the dynamic weights for each state parameter at the current sampling time can be determined, including:

[0087] The correlation coefficient corresponding to any state parameter at the current sampling time is weighted by a first weight to obtain a first weighted value. The delay time correction factor of any state parameter at the current sampling time is weighted by a second weight to obtain a second weighted value. The sum of the first weighted value and the second weighted value is input into the natural exponential function, and the output result is used as the absolute weight of any state parameter at the current sampling time.

[0088] The ratio of the absolute weight of any state parameter at the current sampling time to the sum of the absolute weights of all state parameters at the current sampling time is used as the dynamic weight of any state parameter at the current sampling time.

[0089] Furthermore, the formula for the dynamic weight of each state parameter at the current sampling time is as follows:

[0090]

[0091] in, The dynamic weight of the i-th state parameter at the current sampling time satisfies , This represents a natural exponential function with base e, ensuring that the weights are non-negative and smoothly distributed, avoiding drastic jumps. Indicates the degree of correlation coefficient The first weight is applied to control the correlation coefficient during the filter press control process. Sensitivity The larger the value, the more sensitive the control process is to changes in the degree of correlation. This embodiment preferably uses... A value of 2.0 means For every increase of 0.1, The index component increased by 0.2. Indicates the factor used to correct for delay time. The second weight, applied in the weighting, is a type of delay penalty coefficient used to control the importance of the delay factor. The larger the value, the higher the system's timeliness requirement. Corresponding to the preferred value of the first weight, this embodiment preferably uses... A value of 1.5 can effectively balance relevance and timeliness. This represents the total number of data sources, i.e., the state parameters.

[0092] The dynamic weight calculation formula in this embodiment is essentially a variant of the Softmax function, which uses the natural exponential function. The reasons are twofold: First, it amplifies the sensitivity to differences by giving greater weight to high scores in the exp function, making the superior data sources more prominent and better suited to the technical issues of the current scenario. Second, it ensures gradient smoothness. Since the derivative of the exp function is still the exp function, it means that the weight changes are smooth, thus avoiding weight jumps and helping to control the stability of the filter press in the current scenario.

[0093] After obtaining the dynamic weights of each state parameter at the current sampling time, these dynamic weights can be applied to multi-source data fusion to generate optimal control parameters (such as feed valve opening and filter pressure) for the filter press at the current sampling time. Through feedback, the weight allocation strategy is continuously optimized in subsequent time steps, solving the problems of response lag and insufficient adaptability in traditional control methods. Therefore, intelligent control of the filter press is achieved based on the dynamic weights of each state parameter at the current sampling time, including:

[0094] The value of any state parameter at the current sampling time corresponding to the value of the control parameter to be controlled is recorded as the calculated value of the control quantity of any state parameter at the current sampling time. The product of the dynamic weight of any state parameter at the current sampling time and the calculated value of the control quantity is taken as the weighted calculated value of the control quantity of any state parameter at the current sampling time. The sum of the weighted calculated values ​​of the control quantities of all state parameters at the current sampling time is taken as the control value of the control parameter at the current sampling time.

[0095] The formulaic representation of the control parameter's value at the current sampling time is as follows:

[0096]

[0097] in, This indicates the control value of the control parameter (feed valve opening or filter press pressure) at sampling time t. This indicates that the value of the i-th state parameter at sampling time t corresponds to the value of the control parameter to be controlled, which is also the calculated value of the control quantity of the i-th state parameter at sampling time t. This calculated value of the control quantity is obtained based on the control algorithm (such as PID and its variants) for the corresponding state parameter. The process of obtaining this calculated value of the control quantity is existing technology and will not be described in detail in this embodiment. This represents the dynamic weight of the i-th state parameter corresponding to the current sampling time. It can be set to automatically trigger a filter cloth cleaning warning if a significant change in the weight distribution is detected during the process of intelligent control of the filter press based on the control values ​​of the obtained control parameters (such as a continuous increase in the dynamic weight of the pressure state parameter), thus forming a closed-loop optimization.

[0098] This invention constructs benchmark parameters that accurately reflect the state of a scenario based on its characteristics. Then, by constructing correlation coefficients and delay time correction factors between each state parameter and the benchmark parameters, the degree of correlation between their values ​​and the delay in value changes are accurately quantified. Since the benchmark parameters can accurately characterize the scenario state of the filter press, the correlation coefficients and delay time correction factors corresponding to each state parameter ensure that the state parameter with the most timely response is prioritized to control the filter press when the operating conditions change. This effectively captures the timing characteristics of abnormal pressure data preceding flow data when the filter cloth is clogged, fundamentally solving the rigidity of fixed weight allocation in the current scenario and improving the response speed and control accuracy of the filter press.

[0099] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for intelligent control of a filter press based on multi-source data, characterized in that, The method includes: During the operation of the filter press, various state parameters reflecting the filtration state are continuously sampled, and the theoretical instantaneous filtration rate of the filter press at each sampling moment is calculated as the reference parameter at the corresponding sampling moment. Using the historical time period with the first duration closest to the current sampling time as the first analysis window, the correlation degree between any state parameter and the benchmark parameter is calculated under the first analysis window to obtain the correlation coefficient between any state parameter and the benchmark parameter at the current sampling time. Using the historical time period of the second duration closest to the current sampling time as the second analysis window, the cross-correlation degree between the state parameter and the reference parameter is determined under different time offsets relative to the reference parameter within the second analysis window. The time offset corresponding to the maximum cross-correlation degree is used as the delay time of the state parameter relative to the reference parameter at the current sampling time. Based on the delay time of the state parameter relative to the reference parameter at the current sampling time, a delay time correction factor for the state parameter at the current sampling time is constructed. The time offset can be positive or negative and its absolute value is less than the second duration. The dynamic weight of each state parameter at the current sampling time is determined by the correlation coefficient and the delay time correction factor corresponding to each state parameter at the current sampling time, and the intelligent control of the filter press is completed based on the dynamic weight of each state parameter at the current sampling time. The status parameters include filter pressure, filter flow rate, and filter temperature; Calculating the delay time of any state parameter relative to the reference parameter at the current sampling time includes: ， in, This represents the delay time of the i-th state parameter relative to the reference parameter at the current sampling time. This represents the function that maximizes the parameters, used to find the time offset corresponding to the maximum cross-correlation. Indicates the time offset. This represents the total number of downsampling times in the second analysis window. Indicates the second analysis window The value of the i-th state parameter at each sampling time. Indicates the second analysis window The values ​​of the reference parameters at each sampling time point Indicates the request and The inner product of.

2. The intelligent control method for a filter press based on multi-source data according to claim 1, characterized in that, The calculation of the theoretical instantaneous filtration rate of the filter press at each sampling time includes: The absolute value of the difference between the upstream pressure and the downstream pressure of the filter cloth of the filter press at any sampling time is calculated as the instantaneous pressure difference at any sampling time, and the cumulative volume of filtrate at any sampling time is the third historical time period closest to the sampling time. The theoretical instantaneous filtration rate of the filter press at any given sampling time is directly proportional to the filtration area and the instantaneous pressure difference at any given sampling time, and inversely proportional to the dynamic viscosity, filter cloth resistance, filter cake specific resistance, feed concentration, and cumulative filtrate volume at any given sampling time.

3. The intelligent control method for a filter press based on multi-source data according to claim 1, characterized in that, The process of obtaining the correlation coefficient between any state parameter and the reference parameter at the current sampling time includes: The sequence formed by the values ​​of any state parameter under the first analysis window at each sampling time is denoted as the first sequence, and the sequence formed by the values ​​of the benchmark parameter under the first analysis window at each sampling time is denoted as the second sequence. The absolute value of the Pearson correlation coefficient between the first sequence and the second sequence is denoted as the correlation coefficient between any state parameter and the benchmark parameter at the current sampling time.

4. The intelligent control method for a filter press based on multi-source data according to claim 1, characterized in that, The construction of the delay time correction factor for any state parameter at the current sampling time includes: The ratio of the absolute value of the delay time of any state parameter at the current sampling time compared to the reference parameter to the preset maximum acceptable delay time is recorded as the delay level. The negative number of the delay level is input into the natural exponential function, and the output result is used as the delay time correction factor of any state parameter at the current sampling time.

5. The intelligent control method for a filter press based on multi-source data according to claim 1, characterized in that, The determination of the dynamic weight of each state parameter at the current sampling time includes: The correlation coefficient corresponding to any state parameter at the current sampling time is weighted by a first weight to obtain a first weighted value. The delay time correction factor of any state parameter at the current sampling time is weighted by a second weight to obtain a second weighted value. The sum of the first weighted value and the second weighted value is input into the natural exponential function, and the output result is used as the absolute weight of any state parameter at the current sampling time. The ratio of the absolute weight of any state parameter at the current sampling time to the sum of the absolute weights of all state parameters at the current sampling time is used as the dynamic weight of any state parameter at the current sampling time.

6. The intelligent control method for a filter press based on multi-source data according to claim 1 or 5, characterized in that, The intelligent control of the filter press based on the dynamic weights of each state parameter at the current sampling time includes: The value of any state parameter at the current sampling time corresponding to the value of the control parameter to be controlled is recorded as the calculated value of the control quantity of any state parameter at the current sampling time. The product of the dynamic weight of any state parameter at the current sampling time and the calculated value of the control quantity is taken as the weighted calculated value of the control quantity of any state parameter at the current sampling time. The sum of the weighted calculated values ​​of the control quantities of all state parameters at the current sampling time is taken as the control value of the control parameter at the current sampling time.

7. The intelligent control method for a filter press based on multi-source data according to claim 6, characterized in that, The control parameters include the opening degree of the feed valve of the filter press and the filter pressure.

Images