Model prediction and multi-objective optimization based speed regulation method for fully mechanized coal mining liquid supply system

Abstract

The application discloses a kind of based on model prediction and multi-objective optimization's fully mechanized liquid supply system speed regulation method, belongs to fully mechanized liquid supply control technical field, comprising the following steps: S1, the real-time state information of pump station is collected;S2, according to the real-time state information of pump station, the outlet pressure of pump station is predicted using prediction model;S3, multi-objective optimization function is constructed, S4, according to the outlet pressure of predicted pump station and multi-objective optimization function, the outlet pressure of pump station is controlled.The application realizes pump machine health guarantee, by real-time monitoring pump machine health state, and it is included in optimization target, can effectively prolong the service life of pump machine, reduce maintenance cost, improve the reliability and economy of system.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of fully mechanized liquid supply control, and particularly relates to a fully mechanized liquid supply system speed regulation method based on model prediction and multi-objective optimization. BACKGROUND

[0002] In the fully mechanized working face, the emulsion pump station is the power source of the hydraulic support, and the stability and response speed of the liquid supply pressure directly affect the normal work of the hydraulic support and the coal mining efficiency. The traditional pump station speed regulation method mainly adopts PID control to control the outlet pressure by adjusting the frequency of the pump station motor. However, the PID control has the following limitations in practical application: first, parameter tuning is difficult: the PID parameters are sensitive to the change of the system model parameters, and it is difficult to achieve ideal control effect in the actual complex and changeable fully mechanized working face environment; second, the response speed is limited: PID control is essentially a control based on error feedback, and the response to system disturbance lags behind, which is difficult to meet the demand of fast action of the hydraulic support; third, single target control: PID control can usually only optimize a single target (such as pressure deviation), and it is difficult to consider multiple targets such as pressure fluctuation, system energy efficiency and equipment life; fourth, lack of foresight: PID control cannot predict the future state change of the system, and it is difficult to deal with sudden load disturbance.

[0003] In recent years, with the increasing demand for automation and intelligentization of fully mechanized working faces, pump control speed regulation methods based on model predictive control have gradually attracted attention. By using the system model to predict the future state of the system, and optimizing the control input based on the prediction results, more accurate and faster control of the system is realized. However, most of the existing pump station speed regulation methods based on model predictive control are still limited to single target optimization, and the actual working condition characteristics of the fully mechanized working face, such as the randomness and periodicity of support action, the performance degradation caused by long-time operation of the pump, etc., are not fully considered. SUMMARY

[0004] In order to solve the problems of insufficient control accuracy, slow response speed, difficulty in considering multi-objective optimization, lack of foresight and insufficient consideration of pump health status in the existing fully mechanized pump station speed regulation method, a fully mechanized liquid supply system speed regulation method based on model prediction and multi-objective optimization is proposed.

[0005] The technical scheme of the present application is as follows: a fully mechanized liquid supply system speed regulation method based on model prediction and multi-objective optimization comprises the following steps:

[0006] S1, collecting real-time state information of the pump station;

[0007] S2, predicting the outlet pressure of the pump station according to the real-time state information of the pump station by using a prediction model;

[0008] S3, constructing a multi-objective optimization function;

[0009] S4, controlling the pump station outlet pressure according to the predicted pump station outlet pressure and the multi-objective optimization function.

[0010] Further, in S1, the real-time state information includes the pump station outlet pressure, flow, motor frequency, motor current, motor power, emulsion temperature and support action state.

[0011] Further, in S2, the predicted model composed of the mechanism model and the data-driven model is used to predict the pump station outlet pressure.

[0012] Further, the expression of the mechanism model is:

[0013]

[0014] In the formula, Q(t) represents the pump outlet flow, K1 represents the first pump characteristic coefficient, K2 represents the second pump characteristic coefficient, n(t) represents the pump speed, P(t) represents the pump outlet pressure, represents a nonlinear disturbance term, a i represents an autoregressive coefficient of the disturbance term, b j represents a moving average coefficient of the control input, represents a nonlinear disturbance term at time t-i, and u(t-j) represents a control input at time t-j.

[0015] The data-driven model comprises an input layer, a bidirectional LSTM layer, an attention mechanism module and an output layer connected in sequence; and the expression of the loss function L of the data-driven model is:

[0016] ;

[0017] In the formula, λ1 represents a weight coefficient of the mean square error, λ2 represents a weight coefficient of the continuous ranking probability score, and λ3 represents a weight coefficient of the KL divergence. L MSE represents the mean square error, L CRPS represents the continuous ranking probability score, and L KL represents the KL divergence.

[0018] The expression of the predicted model is:

[0019] ;

[0020] ;

[0021] ;

[0022] In the formula, represents the predicted model output, represents a mechanism model output, represents a data-driven model output, e represents a model fusion error, w m (t) represents a dynamic weight of the mechanism model, w d (t) represents a dynamic weight of the data-driven model, k represents an adjustment coefficient, C m (t) represents a confidence of the mechanism model;

[0023] The expression of the residual compensation function of the prediction model is:

[0024]

[0025] In the formula, ΔP(t) represents a pressure prediction residual, β0represents a steady-state residual coefficient, β1represents a dynamic residual coefficient, β2represents a pressure change rate residual coefficient, P(t) represents a system pressure measurement value, represents a derivative of the system pressure measurement value, represents a residual noise term.

[0026] Further, in S3, the multi-objective optimization function includes a pressure deviation minimization function, a frequency adjustment amplitude minimization function, an energy efficiency maximization function, and a health risk minimization function.

[0027] Further, the expression of the pressure deviation minimization function J1 is:

[0028] ;

[0029] In the formula, P(k) represents an actual pressure of a pump station outlet at time k, P set (k) represents a set pressure of the pump station outlet at time k, and N represents a total time.

[0030] The expression of the frequency adjustment amplitude minimization function J2 is:

[0031] ;

[0032] In the formula, τ(k) represents a time response delay of the system at time k, and Δf(k) represents a system frequency deviation;

[0033] The expression of the energy efficiency maximization function J3 is:

[0034] ;

[0035] In the formula, η(k) represents an efficiency coefficient of the system at time k, P(k) represents active power of the system at time k, and Q(k) represents reactive power of the system at time k;

[0036] The expression of the health risk minimization function J4 is:

[0037] ;

[0038] In the formula, T safe represents the device safety temperature threshold, T(k) represents the temperature of the device at time k.

[0039] Further, in S4, based on the predicted pump station outlet pressure, the multi-objective optimization function is solved to obtain the optimal control input sequence, and the optimal control input sequence is output to the pump station to adjust the motor frequency, completing the control of the pump station outlet pressure.

[0040] Further, the expression of the motor frequency f is:

[0041] ;

[0042] In the formula, f base represents the reference frequency, P max represents the maximum allowable outlet pressure, P min represents the minimum allowable outlet pressure, P o represents the actual outlet pressure, Δf max represents the maximum allowable adjustment amount.

[0043] The beneficial effects of the present application are:

[0044] (1) The control precision of the present application is high, and through model-based prediction and multi-objective optimization, more accurate control of the pump station outlet pressure can be realized, effectively reducing the pressure deviation and fluctuation, and improving the stability and reliability of the hydraulic support work;

[0045] (2) The response speed of the present application is fast, and the motor frequency can be adjusted in advance according to the predicted support action state, thereby improving the response speed of the system to load disturbance and meeting the demand of fast action of the hydraulic support;

[0046] (3) The present application performs multi-objective optimization, which can simultaneously consider multiple objectives such as pressure control, energy efficiency optimization, device life and pump health, realize comprehensive performance optimization of the pump station system, and improve the overall operation efficiency of the system;

[0047] (4) The present application has strong adaptability, and can flexibly adjust the weight coefficients in the multi-objective optimization function according to different working conditions and demands, has strong adaptability, and can adapt to the complex and changeable environment of the fully mechanized coal mining face;

[0048] (5) The present application has high intelligence, adopts model predictive control and multi-objective optimization algorithm, has high intelligence, and meets the development trend of automation and intelligence of the fully mechanized coal mining face;

[0049] (6) The present application realizes pump health guarantee, through real-time monitoring of pump health status, and the pump health status is brought into an optimization target, so that the service life of the pump is effectively prolonged, the maintenance cost is reduced, and the reliability and economy of the system are improved. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 A flowchart of the model prediction and multi-objective optimization based fully mechanized coal mining liquid supply system speed regulation method. DETAILED DESCRIPTION

[0051] The embodiments of the present application will be further described below with reference to the accompanying drawings.

[0052] As shown in Figure 1 The present application provides a model prediction and multi-objective optimization based fully mechanized coal mining liquid supply system speed regulation method, comprising the following steps:

[0053] S1, collecting real-time state information of the pump station;

[0054] S2, predicting the pump station outlet pressure by using a prediction model according to the real-time state information of the pump station;

[0055] S3, constructing a multi-objective optimization function;

[0056] S4, controlling the pump station outlet pressure according to the predicted pump station outlet pressure and the multi-objective optimization function.

[0057] In the embodiments of the present application, the pump station system includes multiple pump groups such as emulsion pumps, spray pumps and water pumps, and each pump group is equipped with a frequency converter and corresponding sensors.

[0058] The sensors include pressure sensors, flow sensors, vibration sensors and temperature sensors, etc., for real-time monitoring of the running state of the pump station system.

[0059] The industrial computer is the core of the control system, and is connected with the industrial computer, the frequency converter and the sensors, and is responsible for data acquisition, processing, model prediction, optimization calculation and control output.

[0060] In the embodiments of the present application, in S1, the real-time state information includes the pump station outlet pressure, flow, motor frequency, motor current, motor power, emulsion temperature and support action state.

[0061] The real-time state information is used for monitoring the running state of the pump station system in real time, including but not limited to pump station outlet pressure P, reflecting the output capacity of the pump station; flow Q, reflecting the liquid supply capacity of the pump station; motor frequency f, reflecting the running speed of the pump station; motor current I, reflecting the load condition of the motor; motor power P, reflecting the energy consumption of the pump station; emulsion temperature T, reflecting the viscosity of the emulsion and the running state of the pump station; support action state S, reflecting the change condition of the load, such as column lifting, column lowering and displacement; pump vibration and temperature and other health state indicators H, reflecting the running health condition of the pump.

[0062] In the embodiment of the present application, in S2, the prediction model composed of the mechanism model and the data-driven model is used to predict the pump station outlet pressure.

[0063] In the embodiment of the present application, the expression of the mechanism model is:

[0064]

[0065] In the formula, Q(t) represents the pump outlet flow, K1 represents the first pump characteristic coefficient, K2 represents the second pump characteristic coefficient, n(t) represents the pump speed, P(t) represents the pump outlet pressure, represents a nonlinear disturbance term, a i represents an autoregressive coefficient of the disturbance term, b j represents a moving average coefficient of the control input, represents a nonlinear disturbance term at t-i, u(t-j) represents a control input at t-j; the pump speed is linearly related to the motor frequency. The first pump characteristic coefficient and the second pump characteristic coefficient are calibrated through factory test, and the typical value range is: K1=0.03-0.05, K2=0.15-0.25.

[0066] The data-driven model comprises an input layer, a bidirectional LSTM layer, an attention mechanism module and an output layer connected in sequence; and the expression of the loss function L of the data-driven model is:

[0067] ;

[0068] In the formula, λ1 represents a weight coefficient of mean square error, λ2 represents a weight coefficient of continuous ranking probability score, λ3 represents a weight coefficient of KL divergence, L MSE represents mean square error, L CRPS represents continuous ranking probability score, L KL represents KL divergence; λ1=0.6, λ2=0.3, λ3=0.1.

[0069] The input layer: 8-dimensional vector features [P(t), Q(t), f(t), I(t), T(t), S(t), H(t), t];

[0070] Bidirectional LSTM layer (64 units): capture temporal features;

[0071] GRU layer (32 units): extract nonlinear patterns;

[0072] Attention mechanism module: , where softmax(·) denotes the activation function, h i denotes the i-th hidden state, W denotes the first learnable parameter, V denotes the second learnable parameter, and b denotes the third learnable parameter;

[0073] Output layer: N-step pressure prediction, N = 5 steps, corresponding to the typical response time of the hydraulic system.

[0074] The expression of the prediction model is:

[0075] ;

[0076] ;

[0077] ;

[0078] where, denotes the prediction model output, denotes the mechanism model output, denotes the data-driven model output, e denotes the model fusion error, w m (t) denotes the dynamic weight of the mechanism model, w d (t) denotes the dynamic weight of the data-driven model, k denotes the adjustment coefficient, C m (t) denotes the confidence of the mechanism model.

[0079] Using the real-time state information obtained in step S1, the future pump station outlet pressure is predicted. This model can be established in a mechanism modeling, data-driven modeling, or a combination of the two ways, for example: the mechanism model can be based on the basic principles of fluid mechanics and electromagnetism, to establish a nonlinear mathematical model of the pump station system. The data-driven model uses historical operation data to establish a prediction model of the pump station system using deep learning. The data-driven model can effectively capture the nonlinear relationship and dynamic characteristics of the system.

[0080] The hybrid model can combine the mechanism model and the data-driven model, utilize the mechanism model to provide physical constraints of the system, and utilize the data-driven model to compensate for the influence of inaccurate parameters of the mechanism model or unmodeled dynamics. Meanwhile, the predicted support action state information is also considered. For example, according to the action history data of the support, a pattern recognition or time series analysis method can be used to predict the action mode (such as column lifting, column lowering and pushing) of the support in a future period of time, and estimate the corresponding flow demand, so as to more accurately predict the change of the pump station outlet pressure.

[0081] The expression of the residual compensation function of the prediction model is as follows:

[0082]

[0083] In the expression, ΔP(t) represents a pressure prediction residual, β0 represents a steady-state residual coefficient, β1 represents a dynamic residual coefficient, β2 represents a pressure change rate residual coefficient, P(t) represents a system pressure measurement value, represents a derivative of the system pressure measurement value, represents a residual noise term.

[0084] In the embodiment of the present application, in S3, the multi-objective optimization function includes a pressure deviation minimization function, a frequency adjustment amplitude minimization function, an energy efficiency maximization function and a health risk minimization function.

[0085] In the embodiment of the present application, the expression of the pressure deviation minimization function J1 is as follows:

[0086] ;

[0087] In the expression, P(k) represents an actual pressure of the pump station outlet at time k, P set (k) represents a set pressure of the pump station outlet at time k, and N represents a total time.

[0088] The expression of the frequency adjustment amplitude minimization function J2 is as follows:

[0089] ;

[0090] In the expression, τ(k) represents a time response delay of the system at time k, and Δf(k) represents a system frequency deviation;

[0091] The expression of the energy efficiency maximization function J3 is as follows:

[0092] ;

[0093] In the expression, η(k) represents an efficiency coefficient of the system at time k, P(k) represents an active power of the system at time k, and Q(k) represents a reactive power of the system at time k;

[0094] The expression of the health risk minimization function J4 is:

[0095] ;

[0096] In the expression, T safe represents the device safety temperature threshold, and T(k) represents the temperature of the device at time k.

[0097] In the embodiment of the present application, in S4, the multi-objective optimization function is solved based on the predicted pump station outlet pressure, the optimal control input sequence is obtained, and the optimal control input sequence is output to the pump station to adjust the motor frequency, thereby completing the control of the pump station outlet pressure.

[0098] The optimal solution selection selects a group of optimal control input sequences [f t1 ,f t2 ,…,f tn ] from the Pareto optimal solution set according to actual needs, wherein f tk represents the motor frequency set value to be recommended to be applied in the future k control periods (k = 1, 2,..., n).

[0099] The control instruction sending sends the optimized control input sequence to the frequency converter by the industrial computer, adjusts the frequency of the pump group motor, and realizes the balanced speed regulation of the pump station system.

[0100] The control effect monitoring monitors the running state of the pump station system in real time, evaluates the control effect, and adjusts the control parameters as needed.

[0101] In the embodiment of the present application, the expression of the motor frequency f is:

[0102] ;

[0103] In the expression, f base represents the reference frequency, P max represents the maximum allowed outlet pressure, P min represents the minimum allowed outlet pressure, P o represents the actual outlet pressure, and Δf max represents the maximum allowed adjustment amount.

[0104] Those skilled in the art will appreciate that the embodiments described herein are intended to help the reader understand the principles of the present application and should be understood as not limiting the scope of protection of the present application to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations according to the technical inspirations disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the scope of protection of the present application.

Claims

1. A speed regulation method for a fully mechanized fluid supply system based on model prediction and multi-objective optimization, characterized in that, Includes the following steps: S1. Collect real-time status information of the pumping station; S2. Based on the real-time status information of the pumping station, use a predictive model to predict the outlet pressure of the pumping station; S3. Construct a multi-objective optimization function; S4. Control the pump station outlet pressure based on the predicted pump station outlet pressure and the multi-objective optimization function; In S1, the real-time status information includes pump station outlet pressure, flow rate, motor frequency, motor current, motor power, emulsion temperature, and support operation status. In S2, a prediction model composed of a mechanism model and a data-driven model is used to predict the pump station outlet pressure. The expression for the mechanism model is: In the formula, Q(t) represents the pump outlet flow rate, K1 represents the characteristic coefficient of the first pump, K2 represents the characteristic coefficient of the second pump, n(t) represents the pump speed, and P(t) represents the pump outlet pressure. a represents the nonlinear disturbance term. i b represents the autoregressive coefficient of the disturbance term. j This represents the moving average coefficient of the control input. Let represent the nonlinear disturbance term at time ti, and u(tj) represent the control input at time tj; The data-driven model comprises an input layer, a bidirectional LSTM layer, an attention mechanism module, and an output layer connected in sequence; the expression for the loss function L of the data-driven model is: ； In the formula, λ1 represents the weighting coefficient of the mean square error, λ2 represents the weighting coefficient of the continuous ranking probability score, λ3 represents the weighting coefficient of the KL divergence, and L MSE L represents the mean square error. CRPS LKL represents the probability score of continuous sorting, and LKL represents the KL divergence. The expression for the prediction model is: ； ； ； In the formula, This represents the output of the prediction model. The output represents the mechanism model. This indicates the data-driven model output, where 'e' represents the model fusion error, and 'w' represents the model fusion error. m (t) represents the dynamic weights of the mechanistic model, w d (t) represents the dynamic weights of the data-driven model, k represents the adjustment coefficient, and C m (t) represents the confidence level of the mechanism model; The expression for the residual compensation function of the prediction model is: In the formula, ΔP(t) represents the pressure prediction residual, β0 represents the steady-state residual coefficient, β1 represents the dynamic residual coefficient, β2 represents the pressure change rate residual coefficient, and P(t) represents the system pressure measurement value. The derivative of the system pressure measurement value. Represents the residual noise term; In S3, the multi-objective optimization function includes a pressure deviation minimization function, a frequency adjustment amplitude minimization function, an energy efficiency maximization function, and a health risk minimization function; The expression for the pressure deviation minimization function J1 is: ； In the formula, P(k) represents the actual pressure at the pump station outlet at time k. set (k) represents the set pressure at the pump station outlet at time k, and N represents the total time. The expression for the frequency adjustment amplitude minimization function J2 is: ； In the formula, τ(k) represents the time response delay of the system at time k, and Δf(k) represents the system frequency deviation; The expression for the energy efficiency maximization function J3 is: ； In the formula, η(k) represents the efficiency coefficient of the system at time k, P(k) represents the active power of the system at time k, and Q(k) represents the reactive power of the system at time k. The expression for the health risk minimization function J4 is: ； In the formula, T safe T(k) represents the safe temperature threshold of the equipment, and T(k) represents the temperature of the equipment at time k.

2. The speed regulation method for a fully mechanized fluid supply system based on model prediction and multi-objective optimization according to claim 1, characterized in that, In step S4, based on the predicted pump station outlet pressure, the multi-objective optimization function is solved to obtain the optimal control input sequence, and the optimal control input sequence is output to the pump station to adjust the motor frequency and complete the control of the pump station outlet pressure.

3. The speed regulation method for a fully mechanized fluid supply system based on model prediction and multi-objective optimization according to claim 2, characterized in that, The expression for the motor frequency f is: ； In the formula, f base P represents the reference frequency. max P represents the maximum permissible outlet pressure. min P represents the minimum permissible outlet pressure. o Indicates actual export pressure, Δf max This indicates the maximum allowable adjustment amount.

Images