Intelligent pressure control method for coal mine emulsion pump

Abstract

The present application relates to a kind of coal mine emulsion pump intelligent pressure control method.The technical problems to be solved by the present application are to provide a kind of coal mine emulsion pump intelligent pressure control method.The technical scheme adopted by the present application comprises the following steps: setting pressure sensor at the outlet of emulsion pump, real-time acquisition emulsion pump outlet pressure data;The pressure data is uploaded to cloud server by communication module, and cloud server is used to store and analyze historical pressure data;Based on the historical pressure data and in combination with the characteristics of mine load working condition, an emulsion pump outlet pressure prediction model is constructed in the cloud, which is used to predict the emulsion pump outlet pressure at future time.The present application has the advantages of: having predictive control capability, significantly improving response speed, compared with traditional PID control, pressure response delay is reduced by about 42%, effectively improving the dynamic response speed of emulsion pump;Adaptive algorithm is used to optimize PID control, and control precision and stability are enhanced.

Description

Technical Field

[0001] This invention relates to an emulsion pump control technology, specifically to an intelligent pressure control method for an emulsion pump used in coal mines. Background Technology

[0002] The underground working environment in coal mines is complex and variable, with frequent pressure fluctuations and significant transient pressure changes. Emulsion pump stations are responsible for providing stable hydraulic support to ensure the normal operation of support equipment. Currently, emulsion pump control mostly adopts constant pressure control or simple pressure feedback control methods, which cannot accurately and quickly adapt to drastic changes in mine load, resulting in slow response, serious energy waste, shortened equipment lifespan, and potential safety hazards. Furthermore, traditional control methods lack effective analysis and prediction capabilities for real-time data, making it difficult to cope with sudden pressure fluctuations and posing serious potential risks to mine operation safety and equipment stability. Moreover, existing technologies still have significant shortcomings in the flexibility and adaptability of control algorithms, especially when facing frequent nonlinear disturbances and sudden load changes in underground conditions. Conventional control strategies cannot effectively handle complex multi-source information, making it difficult to guarantee system stability and response speed under uncertain conditions. At the same time, traditional emulsion pump control systems have limited information exchange capabilities with the upper-level scheduling system, resulting in low data utilization efficiency and failing to achieve overall optimization from data acquisition and predictive analysis to closed-loop control. Therefore, there is an urgent need to develop an intelligent control method with predictive, adaptive, and remote information interaction capabilities to address the poor adaptability and lag in existing emulsion pump control systems.

[0003] This is followed by issues such as high energy consumption. Summary of the Invention

[0004] To address the above problems, the technical problem to be solved by the present invention is to provide an intelligent pressure control method for emulsion pumps used in coal mines.

[0005] The technical solution of the intelligent pressure control method for emulsion pumps used in coal mines of this invention is characterized by the following steps:

[0006] 1) Install a pressure sensor at the outlet of the emulsion pump to collect the outlet pressure data of the emulsion pump in real time;

[0007] 2) The pressure data is uploaded to a cloud server via a communication module, and the cloud server is used to store and analyze historical pressure data;

[0008] 3) Based on the historical pressure data and combined with the characteristics of mine load conditions, a prediction model for emulsion pump outlet pressure is constructed in the cloud to predict the emulsion pump outlet pressure at future moments.

[0009] 4) Based on the deviation between the predicted future pressure value and the target pressure, a pressure error prediction value is generated, wherein the target pressure is automatically set according to the historical pressure statistical characteristics of the emulsion pump under stable operating conditions.

[0010] 5) Construct a non-uniformly weighted PID controller based on the predicted pressure error value, assign different weights to the proportional control term, integral control term, and derivative control term, and generate the basic control quantity;

[0011] 6) Based on the real-time change amplitude of the emulsion pump outlet pressure, a disturbance modulation mechanism is introduced to modulate the basic control quantity to obtain the final control quantity;

[0012] 7) Adjust the output frequency of the frequency converter according to the final control quantity to achieve dynamic adjustment of the emulsion pump outlet pressure;

[0013] 8) Feed back the control process data and operation results to the cloud server for closed-loop updating and adaptive optimization of the pressure prediction model and control parameters.

[0014] The pressure prediction model constructs time series input samples based on historical pressure data and automatically selects the optimal prediction model under the current working conditions based on the model performance evaluation results.

[0015] The target pressure is automatically determined based on the average pressure and pressure fluctuation of the emulsion pump during stable operation, and a safety margin is introduced to cope with load changes.

[0016] In the non-uniformly weighted PID controller, the weight of the proportional control term is higher than the weight of the integral control term, and the weight of the integral control term is higher than the weight of the derivative control term, so as to enhance the system's ability to respond quickly to load changes.

[0017] The weights of the proportional, integral, and derivative control terms are automatically optimized and determined by comprehensively evaluating the system response time, overshoot rate, and steady-state error.

[0018] The disturbance modulation mechanism nonlinearly modulates the control quantity based on the change in outlet pressure of the emulsion pump, so as to improve the control response strength when the pressure disturbance increases and suppress unnecessary control output when the pressure disturbance decreases.

[0019] The disturbance threshold and modulation sensitivity parameters in the disturbance modulation mechanism are automatically determined based on the statistical characteristics of historical pressure disturbance data.

[0020] The cloud server periodically evaluates the pressure prediction error during system operation. When the prediction error exceeds a preset threshold, it updates the pressure prediction model and control parameters to achieve long-term adaptive control.

[0021] The advantages of the intelligent pressure control method for emulsion pumps used in coal mines in this invention are as follows: 1. It has predictive control capability, which significantly improves the response speed. Compared with traditional PID control, the pressure response delay is reduced by about 42%, effectively improving the dynamic response speed of the emulsion pump; 2. It adopts an adaptive algorithm to optimize PID control, enhancing control accuracy and stability; 3. It introduces a disturbance modulation factor θ(t). By defining the disturbance modulation factor θ(t) = 1 / [1 + exp(−k·(D_norm − D_th))], the controller can automatically adjust the control gain according to the real-time disturbance intensity, thereby maintaining a stable output during sudden pressure fluctuations. Attached Figure Description

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0023] Figure 1 This is a schematic diagram of the overall structure of the intelligent pressure control system for coal mine emulsion pumps of the present invention.

[0024] Figure 2 This is a comparison curve of the predicted and actual outlet pressure values ​​of the emulsion pump in this invention.

[0025] Figure 3 This is a flowchart illustrating the intelligent pressure control method for emulsion pumps used in coal mines according to the present invention.

[0026] Figure 4 This is a schematic diagram comparing the present invention with the traditional PID control method under pressure fluctuation conditions.

[0027] Figure 5 This is a schematic diagram of the relationship between the disturbance modulation factor and the intensity of pressure disturbance in this invention. Detailed Implementation

[0028] like Figure 1-5 As shown, the intelligent pressure control method for emulsion pumps used in coal mines involved in this invention comprises the following specific steps:

[0029] (1) A high-precision pressure sensor is installed at a special position at the outlet of the emulsion pump. The sensor has extremely high sensitivity and fast response characteristics to ensure that the emulsion pump outlet pressure data P(t) is collected in real time and accurately. This data can accurately reflect the small changes in the underground pressure of the mine.

[0030] (2) Using IoT technology and a high-speed and stable communication module, the collected real-time pressure data is uploaded to the cloud server. The cloud server has massive storage capacity to efficiently store and manage historical pressure data and ensure the real-time nature and integrity of the data.

[0031] (3) Based on massive historical pressure data stored in the cloud, and combined with actual mine conditions, such as the pressure fluctuation characteristics caused by face advancement and dynamic adjustment of hydraulic supports, a high-precision emulsion pump outlet pressure prediction model is constructed: P_pred(t) = f_AA ( P(t), P(t-1), … , P(tn) ),

[0032] Where P_pred(t) represents the predicted pressure value at the next sampling time.

[0033] P(t-1) to P(tn) represent historical pressure data arranged in chronological order of sampling periods.

[0034] n is the number of historical data points, and fAA is the learning function;

[0035] The steps to determine fAA are as follows:

[0036] 1. Establish a model library

[0037] The system comes pre-installed with various classic time series forecasting algorithms, including but not limited to: ARIMA (Autoregressive Moving Average), LSTM (Long Short-Term Memory), GPR (Gaussian Process Regression), and SVR (Support Vector Regression). Each model has an independent training interface and can autonomously optimize parameters and train the model based on historical stress data in a cloud environment.

[0038] 2. Constructing training samples

[0039] Input-output sample pairs are generated using the sliding window method:

[0040] Input sample: Xi = [P(i), P(i−1), …, P(i−n)];

[0041] Output sample: Yi = P(i+1).

[0042] A sample set D = {Xi, Yi} is formed by continuous sampling, and each candidate model is trained based on this set.

[0043] 3. Model Performance Evaluation and Optimal Model Determination

[0044] In the cloud server, the system calculates the prediction error of each model under the current working conditions based on the validation set, and determines the model with the smallest prediction error as the current pressure prediction model f_AA. The prediction error is the mean square error of the model on the validation dataset.

[0045] Calculate the future pressure error prediction value e_pred in real time and assess the risk of pressure fluctuation:

[0046] e_pred = P_set - P_pred,

[0047] Where e_pred is the predicted pressure error,

[0048] P_pred is the predicted pressure value for the next moment.

[0049] P_set is the preset target pressure for the safe operation of the pump station, based on data from the long-term stable operation of the emulsion pump system. During the initial deployment phase, the system automatically enters an initial learning period.

[0050] The system performs statistical analysis on the actual pressure data under historical stable operating conditions and extracts:

[0051] The average pressure during the normal pressure supply phase, P_avg; its operating standard deviation, P_std;

[0052] Then, the following empirical formula is used for dynamic setting:

[0053] P_set = P_avg + k_s × P_std

[0054] Where P_avg is the average pressure value during the stable operation phase.

[0055] P_std is the corresponding pressure fluctuation amplitude indicator.

[0056] k_s is the safety margin coefficient;

[0057] k_s is the safety margin factor, with a recommended value of 0.5 to 1.0, used to cope with pressure drops caused by slight load fluctuations.

[0058] Utilizing the specific load characteristics of the mine and combining the real-time error prediction value e(t+Δt), a PID control algorithm is designed: U_base = a*Kp*e_pred + b*Ki*e_sum + c*Kd*e_diff

[0059] Where e_pred is the predicted pressure error;

[0060] e_sum is the cumulative value of pressure error over a period of time;

[0061] e_diff represents the change in pressure error within adjacent sampling periods;

[0062] a, b, and c are the weighting coefficients for the proportional, integral, and differential terms.

[0063] Automatically optimize PID control parameters (Kp, Ki, Kd). The specific optimization process is as follows:

[0064] Initial sample weight space setting

[0065] Set the weight search space of a, b, and c to satisfy the constraint a + b + c = 1;

[0066] Divide the discrete sampling point set into a, b, and c, such as a ∈ [0.4, 0.7], with a step size of 0.05, and adjust other parameters proportionally;

[0067] The combinations form m candidate weight combinations (aᵢ, bᵢ, cᵢ), which constitute the parameter space.

[0068] 1. Quantitative evaluation of response performance:

[0069] For each set of parameters, the operating system is simulated or tested in the field under standard test disturbances (such as a typical 0.2 MPa step disturbance);

[0070] Record its key indicators such as response time Tr, overshoot rate Mp, and steady-state error Ess;

[0071] 2. Construct a comprehensive performance evaluation function:

[0072] J = w1*Tr + w2*Mp + w3*Ess, where w1, w2, and w3 are normalized performance metric weights (e.g., 0.4, 0.3, and 0.3), ensuring that the performance cost function is minimized.

[0073] 3. Optimal combination selection and correction:

[0074] Find the set that minimizes J as the optimization result;

[0075] If the actual disturbance frequency on site is higher than the setting during training, the weight of the a value can be further adjusted upward by 5% to improve the response speed of the proportional term;

[0076] The final optimization results a = 0.6, b = 0.3, c = 0.1 are used as instance values.

[0077] To enhance the controller's adaptability to different disturbance types, a disturbance modulation factor θ is introduced, which is calculated using the following formula:

[0078] θ = 1 / ( 1 + exp( -k*(D_norm - D_th) ) )

[0079] Where: D_norm = abs( P_now - P_prev ) / P_ref

[0080] P_now is the current sampling pressure value, and P_prev is the pressure value from the previous sampling period.

[0081] P_ref is the system reference pressure, D_th is the disturbance threshold, and k is the sensitivity factor.

[0082] P_ref is the system's "normal" operating pressure level, which is the average pressure of the emulsion pump over a period of time under stable operating conditions.

[0083] P_ref = ( P1 + P2 + … + Pn ) / n

[0084] Among them, P1 to Pn are the pressure values ​​continuously collected during the stable operation period.

[0085] D_th is dynamically determined using a confidence interval fitting method:

[0086] Calculate the mean D_avg and standard deviation D_std of the disturbance amplitude D_norm over a recent period (e.g., 600 seconds).

[0087] D_th = D_avg + z * D_std

[0088] Where z is the standard normal distribution coefficient at the selected confidence level, and D_th is repeatable, objective, and can automatically adapt to different operating conditions and disturbance characteristics, avoiding manual setting.

[0089] The method for determining k is as follows:

[0090] Collect disturbance data over a continuous operating cycle on a cloud platform and calculate the disturbance growth rate for each disturbance occurrence: G_i = ΔD_norm / Δt_i

[0091] Where ΔD_norm represents the magnitude of change in the normalized value of the perturbation intensity, and Δt_i is the minimum time interval before and after the perturbation abruptly occurs.

[0092] The disturbance growth rate was statistically analyzed, and a representative value was selected as the typical disturbance growth rate G_typ.

[0093] Based on the slope properties of the logic function curve, this invention automatically determines the sensitivity factor k using the following formula: k = C / (G_typ - D_th)

[0094] Where C is a constant coefficient used to control the width of the transition interval of the modulation function, preferably ln(19), so that the perturbation modulation factor can achieve a smooth transition from low response to high response near the perturbation threshold.

[0095] The derivation of this formula is based on the width control logic of the sigmoid function in the interval θ=0.05→0.95, and ln(19) ≈ 2.9444 corresponds to the change in the threshold range.

[0096] Using the above method, the sensitivity coefficient k can be automatically estimated based on historical disturbance gradients, ensuring that the modulation curve is neither too sluggish nor too sensitive. It has advantages such as data-driven, adaptive, and no subjective intervention required, which meets the needs of industrial intelligent control.

[0097] (7) The final control output is obtained: U = θ * U_base

[0098] (8) The control signal U is used to adjust the output frequency of the frequency converter: F = F0 + U, so as to realize the dynamic adjustment of the emulsion pump pressure.

[0099] (9) All data is uploaded to the cloud and the system performs closed-loop feedback to ensure the long-term adaptability and self-healing of the system.

[0100] This invention incorporates a high-precision pressure sensor at the outlet of an emulsion pump and constructs a time-series prediction model based on historical data from the cloud. This model can predict future pressure change trends before pressure surges occur, enabling feedforward compensation before load disturbances. By introducing deep learning models such as LSTM, the system can output a correction signal approximately 0.8 seconds before load changes, significantly reducing hysteresis. Experimental results show that compared to traditional PID control, the pressure response delay is reduced by approximately 42%, effectively improving the dynamic response speed of the emulsion pump.

[0101] This invention introduces non-uniform weighting factors a, b, and c for the proportional (P), integral (I), and derivative (D) terms into the traditional PID control structure, and automatically optimizes them through a cloud-based performance function J_i, enabling the PID parameters to adapt to different disturbance characteristics. This method avoids the misalignment problem caused by fixed weights in traditional control, and can automatically correct the balance between control force and feedback speed, ensuring that the system maintains low overshoot and fast convergence under different load changes.

[0102] This invention defines a disturbance modulation factor θ = 1 / [1 + exp(−k·(D_norm − D_th))], enabling the controller to automatically adjust the control gain based on the real-time disturbance intensity, thereby maintaining stable output during sudden pressure fluctuations. Both D_th and k are automatically determined by a statistical model, requiring no manual setting. The system can correct its sensitivity in real time according to operating conditions, possessing data-driven self-learning and disturbance rejection capabilities. Through disturbance modulation, the system achieves stable pressure control within a range of ±0.15 MPa, improving system stability by approximately 62% compared to the ±0.4 MPa fluctuation range of traditional PID controllers.

[0103] This invention enables data to be uploaded to the cloud via IoT communication. The cloud server trains and updates historical data, periodically optimizing the prediction model and PID parameters, forming a closed-loop system of "data acquisition—predictive analysis—control execution—feedback correction". During long-term equipment operation, the control accuracy will not decrease with environmental changes, possessing self-correction and self-evolution characteristics, effectively extending the equipment's lifespan and reducing manual intervention.

[0104] Because the control system of this invention can predict and adjust before pressure changes occur, it avoids frequent start-stop and excessive output, thereby reducing pump energy consumption by about 18% to 25%. At the same time, the pump station operates at a more stable pressure, significantly reduces hydraulic shock, effectively reduces pipeline fatigue and hydraulic component wear, extends equipment maintenance cycles, and improves mine operation safety.

[0105] The control module of this invention has a standard communication interface, enabling data interconnection with the mine's central dispatch platform and supporting remote monitoring and self-diagnosis. This feature allows the system to serve as a core node in an intelligent mine hydraulic control network, providing a scalable foundation for subsequent intelligent mine construction.

[0106] Specific Embodiment: The intelligent pressure control system for coal mine emulsion pumps of the present invention includes: an emulsion pump 1, an outlet high-precision pressure sensor 2, a communication module 3, a cloud server 4, a controller 5, a frequency converter 6, and a hydraulic support execution unit 7. The pressure sensor 2 is installed at the outlet main pipeline of the emulsion pump, with a sampling frequency of 1 Hz and a sampling accuracy of ±0.01 MPa. The communication module 3 adopts an industrial-grade Ethernet or 5G mining communication module to upload the collected data to the cloud server 4. The cloud server 4 is used to store, analyze, and predict pressure data. The controller 5 receives the control signal returned by the cloud server 4 and adjusts the speed of the motor of the emulsion pump 1 through the frequency converter 6, thereby realizing dynamic pressure adjustment.

[0107] The emulsion pump station was selected with a rated output pressure of 10 MPa and a flow rate of 80 L / min. In the initial stage of system operation, pressure sensor 2 continuously collected pump outlet pressure data with a sampling period of 1 s and uploaded the collected pressure data to cloud server 4.

[0108] The cloud server 4 performs initial training on 6,000 sets of continuously collected stress data to build a stress prediction model.

[0109] Training samples are constructed using the sliding window method, with a window length of n=10:

[0110] (1) Constructing the sample set:

[0111] Input vector Xi = [P(i), P(i-1), ..., P(i-10)]

[0112] Output quantity Yi = P(i+1)

[0113] (2) Establish a model library:

[0114] Cloud server 4 calls the following prediction models to form the model library:

[0115] ARIMA model (3,1,2);

[0116] LSTM neural network model (64 hidden layer units, learning rate 0.001).

[0117] GPR model (radial basis kernel function);

[0118] SVR model (radial basis kernel, C=10, gamma=0.1).

[0119] All models are trained on the same training dataset, and the mean squared error of the pressure prediction over the next 10 seconds is used as the performance evaluation metric. Based on the principle of minimizing error, the system automatically selects the LSTM model as the current pressure prediction model f_AA.

[0120] (3) Pressure prediction and target pressure setting

[0121] During the initial learning phase, the system performs statistical analysis on the pressure data under stable operating conditions to obtain the average pressure value P_avg and the pressure fluctuation index P_std during the stable operating phase.

[0122] Statistical analysis yields P_avg = 9.6 MPa, P_std = 0.3 MPa, and the safety margin coefficient k_s is set to 0.6. Therefore, the target set pressure is: P_set = 9.6 + 0.6 * 0.3 = 9.78 MPa

[0123] Controller 5 executes a non-uniformly weighted PID control algorithm based on the predicted pressure error E_pred, and its control output is:

[0124] U_base = a * Kp * E_pred+ b * Ki * E_sum+ c * Kd * E_diff

[0125] Where E_sum is the cumulative value of the prediction error over multiple sampling periods, E_diff is the change in prediction error between adjacent sampling periods, and a, b, and c are the weighting coefficients of the proportional, integral, and derivative terms, and satisfy a + b + c = 1.

[0126] The initial weight search range is set as follows:

[0127] a ∈ [0.4, 0.7]

[0128] b ∈ [0.2, 0.4]

[0129] c ∈ [0.1, 0.2]

[0130] Simulation tests were conducted on various weight combinations by applying a 0.2 MPa step disturbance. The rise time Tr, overshoot Mp, and steady-state error Ess of the system were recorded, and a comprehensive performance evaluation function was constructed.

[0131] J = w1 * Tr + w2 * Mp + w3 * Ess

[0132] Among them, w1, w2, and w3 are 0.4, 0.3, and 0.3, respectively.

[0133] Based on the comprehensive performance evaluation results, the optimal weight combination was finally determined as follows:

[0134] a = 0.6, b = 0.3, c = 0.1

[0135] Corresponding performance indicators: Tr = 0.85 s, Mp = 3.5%, Ess = 0.02 MPa.

[0136] Compared with traditional constant pressure PID control (Tr = 1.8 s, Mp = 10%, Ess = 0.08 MPa), the response speed is improved by about 53% and the steady-state error is reduced by 75%.

[0137] To further suppress false triggering caused by high-frequency disturbances, a disturbance modulation factor θ is introduced, which is calculated as follows:

[0138] θ = 1 / ( 1 + exp( -k * ( D_norm - D_th ) ) )

[0139] Where: D_norm = abs( P_now - P_prev ) / P_ref

[0140] P_now is the current sampled pressure value, P_prev is the pressure value of the previous sampling period, and P_ref is the average pressure value of the system during the stable operation phase.

[0141] The system statistically obtains the average value of D_norm, D_avg = 0.016, and the standard deviation, D_std = 0.004, based on 600 seconds of historical data, and determines the perturbation threshold according to the following formula:

[0142] D_th = D_avg + z * D_std

[0143] Where z is 1.96

[0144] The result is Dth = 0.0238.

[0145] Statistical analysis of historical disturbance gradient data yielded a typical disturbance growth rate G_typ = 0.045, and the sensitivity factor k was calculated using the following formula:

[0146] k = ln(19) / (Gtyp−Dth) = 139.7

[0147] When θ(t)≈0.95, the control output almost completely responds to the disturbance; when θ(t)≈0.05, the system ignores slight disturbances and achieves stable operation.

[0148] Field tests show that in working conditions with frequent load fluctuations, the pressure fluctuation range of the system is reduced from ±0.4 MPa to ±0.15 MPa, and energy consumption is reduced by about 11%.

[0149] The final control output is:

[0150] U = θ * U_base

[0151] And then converted to the motor drive frequency by frequency converter 6:

[0152] F = F0 + U

[0153] Field test results show that under working conditions with frequent load fluctuations, the system pressure fluctuation amplitude was reduced from ±0.4 MPa to ±0.15 MPa, and the system response speed and stability were significantly improved.

[0154] All real-time running data is periodically uploaded to the cloud server 4. When the prediction error exceeds the set threshold, the system automatically triggers the model retraining and parameter update mechanism to achieve long-term stable operation.

[0155] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention are included within the protection scope of the present invention.

Claims

1. A method for intelligent pressure control of an emulsion pump used in coal mines, characterized in that: Includes the following steps: 1) Install a pressure sensor at the outlet of the emulsion pump to collect the outlet pressure data of the emulsion pump in real time; 2) The pressure data is uploaded to a cloud server via a communication module, and the cloud server is used to store and analyze historical pressure data; 3) Based on the historical pressure data and combined with the characteristics of mine load conditions, a prediction model for emulsion pump outlet pressure is constructed in the cloud to predict the emulsion pump outlet pressure at future moments. 4) Based on the deviation between the predicted future pressure value and the target pressure, a pressure error prediction value is generated, wherein the target pressure is automatically set according to the historical pressure statistical characteristics of the emulsion pump under stable operating conditions. 5) Construct a non-uniformly weighted PID controller based on the predicted pressure error value, assign different weights to the proportional control term, integral control term, and derivative control term, and generate basic control quantities, wherein the weight of the proportional control term is higher than the weight of the integral control term, and the weight of the integral control term is higher than the weight of the derivative control term. By comprehensively evaluating the system response time, overshoot rate, and steady-state error, the weights of the proportional control term, integral control term, and derivative control term are automatically optimized and adjusted. 6) Based on the real-time change amplitude of the emulsion pump outlet pressure, a disturbance modulation mechanism is introduced to nonlinearly modulate the basic control quantity, so as to increase the control response strength when the pressure disturbance increases and reduce the control output when the pressure disturbance decreases, so as to obtain the final control quantity. 7) Adjust the output frequency of the frequency converter according to the final control quantity to achieve dynamic adjustment of the emulsion pump outlet pressure; 8) Feed back the control process data and operation results to the cloud server for closed-loop updating and adaptive optimization of the pressure prediction model and control parameters.

2. The intelligent pressure control method for emulsion pumps used in coal mines according to claim 1, characterized in that: The pressure prediction model constructs time series input samples based on historical pressure data and automatically selects the optimal prediction model under the current working conditions based on the model performance evaluation results.

3. The intelligent pressure control method for emulsion pumps used in coal mines according to claim 1, characterized in that: The target pressure is automatically determined based on the average pressure and pressure fluctuation of the emulsion pump during stable operation, and a safety margin is introduced to cope with load changes.

4. The intelligent pressure control method for emulsion pumps used in coal mines according to claim 1, characterized in that: The disturbance threshold and modulation sensitivity parameters in the disturbance modulation mechanism are automatically determined based on the statistical characteristics of historical pressure disturbance data.

5. The intelligent pressure control method for emulsion pumps used in coal mines according to claim 1, characterized in that: The cloud server periodically evaluates the pressure prediction error during system operation. When the prediction error exceeds a preset threshold, it updates the pressure prediction model and control parameters to achieve long-term adaptive control.

Images