Submersible sewage pump set centralized control method based on equipment residual life model

Abstract

The invention relates to the technical field of water pump control, in particular to a submersible sewage pump set centralized control method based on an equipment residual life model, and the method comprises the steps: training a long-short-term memory neural network based on deep convolution, and predicting the residual life of a used submersible sewage pump; the fixed-frequency submersible sewage pumps and the variable-frequency submersible sewage pumps are numbered; recording the equipment residual life of the submersible sewage pump in normal use as a first threshold value of the water pump, recording the downtime of the submersible sewage pump after maintenance as a second threshold value of the water pump, and taking the sum of the two threshold values as a judgment standard of the start-stop priority of the water pump, the larger the value is, the higher the start priority is, otherwise, the higher the stop priority is; relevant liquid level threshold values are set according to the size of the adjusting tank and the type and number information of the submersible sewage pumps; according to the invention, unnecessary shutdown and maintenance are avoided, the problem of frequent start and stop is avoided, and the constant liquid level of the adjusting tank is effectively maintained.

Description

Technical Field

[0001] The present invention relates to the technical field of water pump control, and in particular to a centralized control method of a submersible sewage pump group based on a residual life model of equipment. Background Art

[0002] The main measures taken for maintenance of submersible sewage pump units in traditional sewage plants are:

[0003] 1) Regular maintenance: Preventive maintenance based on time or usage cycle, the purpose is to inspect and maintain before a failure occurs to reduce the probability of failure.

[0004] 2) Post-maintenance: Repair and replacement performed after equipment failure, with the aim of restoring the normal operation of the equipment.

[0005] The above methods will cause the unit to stop operating, affect the normal operation of the sewage treatment plant, and also have a great impact on the control strategy of the submersible sewage pump.

[0006] The main measures taken to control the submersible sewage pump units in sewage plants are mostly to unilaterally control the start and stop of the submersible sewage pumps based on the flow and liquid level of the pool. When controlling only based on the flow or liquid level, the disadvantage that the water volume changes quickly may cause the submersible sewage pumps to start and stop frequently.

[0007] The main measures taken for energy-saving control of submersible sewage pump units in sewage treatment plants are to replace the overall hardware facilities with more energy-saving or variable-frequency controlled submersible sewage pumps. Energy saving is achieved from the perspective of the overall pump unit hardware, which is costly. The method of using a variable-frequency submersible sewage pump to cooperate with an old fixed-frequency submersible sewage pump to achieve energy-saving control has not been considered. Summary of the invention

[0008] The present invention provides a centralized control method for a submersible sewage pump group based on an equipment remaining life model, which avoids unnecessary shutdown and maintenance and improves production efficiency. At the same time, through the coordination of a variable frequency submersible sewage pump and a fixed frequency submersible sewage pump, the problem of frequent start and stop when only a fixed frequency submersible sewage pump is used is avoided, and the liquid level in the regulating tank is effectively maintained constant.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] A centralized control method for a submersible sewage pump group based on a residual life model of equipment comprises the following steps:

[0011] S1. Train a remaining life model based on a deep convolutional long short-term memory neural network and use it to predict the remaining life of the submersible sewage pump used;

[0012] S2. Number the fixed-frequency submersible sewage pumps and the variable-frequency submersible sewage pumps, wherein the fixed-frequency submersible sewage pumps are m and the variable-frequency submersible sewage pump is 1;

[0013] S3, the remaining life of the submersible sewage pump during normal use is recorded as the first threshold of the water pump, and the downtime of the submersible sewage pump after maintenance is recorded as the second threshold of the water pump. The sum of the two is the criterion for determining the start and stop priority of the water pump. The larger the value, the higher the start priority, and vice versa.

[0014] S4. Set relevant liquid level thresholds according to the size of the regulating tank, the type and quantity of the submersible sewage pumps;

[0015] S5. When the water level in the regulating tank is lower than the set first threshold, all submersible sewage pumps stop running and an alarm is triggered;

[0016] S6. When the water level in the regulating tank is higher than the set first threshold value and lower than the second threshold value, the variable frequency submersible sewage pump is started and the liquid level is controlled to be stable at the set value;

[0017] S7. When the water level in the regulating tank is higher than the second threshold and lower than the third threshold, the variable frequency submersible sewage pump and a fixed frequency submersible sewage pump are started according to the priority and the variable frequency submersible sewage pump is controlled to stabilize the liquid level at the set value;

[0018] S8, and so on, gradually increase the number of fixed-frequency submersible sewage pumps to be started. When the water level in the regulating tank is higher than the set m+1th threshold and lower than the m+2th threshold, start the variable-frequency submersible sewage pump and m fixed-frequency submersible sewage pumps, and control the variable-frequency submersible sewage pumps to stabilize the liquid level at the set value;

[0019] S9. When the water level in the regulating tank is higher than the m+2th threshold, an alarm will be given.

[0020] Furthermore, it also includes obtaining a degradation simulation data set of the submersible sewage pump over time or usage, including vibration signal time series data simulated from operation to failure under different failure modes, for obtaining training data for the remaining life model.

[0021] Furthermore, the submersible sewage pump adopts time-frequency analysis of its vibration signal, and provides time domain and frequency domain information simultaneously by performing short-time Fourier transform on the vibration signal v(t).

[0022] Furthermore, the time-frequency representation of each data set of the vibration signal is a matrix S i , where i = 1, 2, ..., T, T is the number of data sets, and the remaining life prediction of equipment is based on the sequence matrix S i The prediction at time t.

[0023] Furthermore, the mean square error of the equipment remaining life model is less than 0.1.

[0024] Furthermore, the prediction of the remaining life of the equipment adopts the stochastic gradient descent method to optimize the problem of minimizing the loss function, uses time backpropagation to calculate the gradient to optimize the model parameters, and for the top logical layer of the prediction, uses the loss function to calculate the gradient of the weight and bias.

[0025] Furthermore, the liquid level is stabilized at a set value and a PID controller is used to control a variable frequency submersible sewage pump.

[0026] Furthermore, if the water level in the regulating tank decreases, the number of fixed-frequency submersible sewage pumps started will be shut down one by one, and finally the number of variable-frequency submersible sewage pumps will be shut down. The stopping priority of the submersible sewage pump is opposite to the starting priority.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1) The present invention is easy to implement. The required liquid level threshold can be set according to the historical liquid level information and relevant dimensions of the pool, thereby controlling the submersible sewage pump;

[0029] 2) Adjust the start and stop times of the submersible sewage pump in real time according to the liquid level. Combined with the control of the variable frequency submersible sewage pump, it can avoid the energy waste caused by the frequent start and stop of the fixed frequency submersible sewage pump;

[0030] 3) Using variable frequency submersible sewage pumps to control the liquid level at each stage at a reasonable value can not only ensure that the submersible sewage pump works in an efficient working range, but also reduce the actual operating head of the submersible sewage pump and the actual operating power, thereby achieving the purpose of energy saving;

[0031] 4) By completing the reasonable control and allocation of the submersible sewage pump based on the method of equipment remaining life prediction, the service life of the submersible sewage pump can be effectively extended, unnecessary maintenance costs and downtime can be avoided, and the submersible sewage pump's centralized control system can be better coordinated to meet actual production needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of the method of the present invention.

[0033] Figure 2 It is the PID control flow chart of the present invention. DETAILED DESCRIPTION

[0034] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings:

[0035] See Figure 1 , is a flow chart of the method of the present invention. The present invention provides a submersible sewage pump group centralized control method based on the equipment remaining life model, comprising the following steps:

[0036] S1. Train a remaining life model based on a deep convolutional long short-term memory neural network and use it to predict the remaining life of the submersible sewage pump used;

[0037] Obtain a simulated dataset of degradation of a submersible sewage pump over time or usage, including the vibration signal time series data of the run-to-failure simulation under different failure modes;

[0038] A deep convolution-based long short-term memory network model CLSTM is constructed to predict the remaining life of the equipment. The constructed network performs convolution operations on the LSTM input to state and state-to-state transitions, which includes both time series information and signal time-frequency information.

[0039] Since the vibration signal collected from the submersible sewage pump exhibits non-stationary characteristics, it is necessary to use time-frequency analysis. By performing short-time Fourier transform on the vibration signal v(t), time domain and frequency domain information can be provided simultaneously. The Fourier transform of the vibration signal v(t) on the real symmetric window function η(t) is calculated as follows:

[0040]

[0041] Where η is the modulation frequency; τ is the conversion time; j is the imaginary unit, η(t-τ) is the offset of the real symmetric window function η(t) relative to τ;

[0042] The time-frequency representation of each data set of the acquired vibration signal is represented by a matrix S i (i=1,2,…,T), T is the number of data sets, and the remaining life prediction of equipment is based on the sequence matrix S i The prediction at time t;

[0043] The long short-term memory network based on the deep convolutional network contains input gates, output gates, forget gates, cells and output gates. It is mainly realized by the expansion of forward propagation and back propagation. The equation for calculating the output of the forward process is as follows:

[0044] Input Gate:

[0045] in, is the activation value of the input gate at time t; * is the convolution operator; W si is the weight matrix from the input gate to the input; S t is the input; W hi is the weight matrix from the input gate to the hidden unit; H t-1 is a hidden unit; W ci is the diagonal weight matrix representing the perspective connection; C t-1 is a memory unit; B i is the input gate bias vector;

[0046]

[0047] in, is the output value of the input gate at time t; σ is the sigmoid activation function

[0048] Forget Gate:

[0049] in, is the activation value of the forget gate at time t; is the output value of the forget gate at time t; W sf is the weight matrix from the forget gate to the input; W hf is the weight matrix from the forget gate to the hidden unit; W cf is the diagonal weight matrix representing the perspective connection; B f is the forget gate bias vector;

[0050] Cell:

[0051] in, is the activation value of the cell at time t; W sc is the weight matrix from the cell to the input; W hc is the weight matrix from the cell to the hidden unit; B c is the cell bias vector; is the state of the cell at time t; is the state of the cell at the last time step t-1; φ is the hyperbolic tangent nonlinear function

[0052] Output Gate:

[0053] in, is the activation value of the output gate at time t; is the output value of the output gate at time t; W so is the weight matrix from the output gate to the input; W ho is the weight matrix from the output gate to the hidden unit; W co is the diagonal weight matrix representing the perspective connection; B o is the output gate bias vector;

[0054] Unit output:

[0055] in, is the output of the model unit at time t;

[0056] When making predictions, add a logistic layer on the cell output like this:

[0057]

[0058] Among them, flatten means flattening the matrix into a vector; is the predicted value of the remaining life of the equipment; W hr The weight of the logic layer; b r is the bias of the logic layer;

[0059] In the prediction of the remaining life of equipment, the loss function is defined as:

[0060]

[0061] Among them, r t The remaining life of the real equipment;

[0062] θ={W si ,W hi ,W ci ,W sf ,W hf ,W cf ,W sc ,W hc ,W so ,W ho ,W co ,W hr ,B i ,B f ,B c ,B o ,b r} is the set of all parameters to be learned;

[0063] The stochastic gradient descent method is used to solve the problem of minimizing the optimization loss function: time backpropagation is used to calculate the gradient to optimize the model parameters. For the top logical layer of prediction, the gradient of the loss function with respect to the weights and biases can be calculated as:

[0064]

[0065] Among them, σ′(·) is the derivative of the sigmoid activation function;

[0066] The gradient calculation method for the remaining model parameters is the same as above;

[0067] After calculating the gradient through back propagation, the parameters can be optimized using gradient descent. The parameter update process can be written as:

[0068]

[0069] In the formula, α is the learning rate;

[0070] Input the corresponding operation data of the tested submersible sewage pump into the neural network model to obtain the remaining life of the tested submersible sewage pump, and the mean square error MSE of the remaining life prediction model of the equipment can reach below 0.1.

[0071] S2. Number the fixed-frequency submersible sewage pump and the variable-frequency submersible sewage pump. There are m fixed-frequency submersible sewage pumps and 1 variable-frequency submersible sewage pump. The m fixed-frequency submersible sewage pumps are respectively denoted as A1, A2... A m ; The 1 variable-frequency submersible sewage pump is denoted as B1.

[0072] S3. When the submersible sewage pump is operating normally, predict its service life according to the remaining life model of the equipment, and record this life value as the first threshold L1 of the submersible sewage pump. The shutdown time of the submersible sewage pump since the completion of maintenance is recorded as the second threshold L2 of the submersible sewage pump. Use the sum L3 of the first threshold and the second threshold of the submersible sewage pump as the judgment standard for the startup priority of the submersible sewage pump. The larger L3 is, the higher the priority of the corresponding submersible sewage pump. When needed, it starts first. On the contrary, when L3 is the smallest, the pump is stopped first.

[0073] S4. Obtain the size information of the regulating tank and the types and quantities of the corresponding submersible sewage pumps. The size information of the regulating tank is the height H of the regulating tank. The types of the corresponding submersible sewage pumps include the fixed-frequency submersible sewage pump A and the variable-frequency submersible sewage pump B. The number of fixed-frequency submersible sewage pumps is m, which are respectively denoted as A1, A2... A m , and the 1 variable-frequency submersible sewage pump is denoted as B1. Set appropriate liquid level thresholds x1, x2... x m 、x m+1 、x m+2 according to the original historical liquid level information and adjustment information, where 0 < x1 < x 2< … < x m < x m+1 < x m+2 < H, or it can also be evenly divided according to the size depth of the water tank and the number of submersible sewage pumps. When dividing the liquid level threshold, the lowest threshold and the highest threshold of the liquid level should meet the actual application needs, and the liquid level of the regulating tank cannot be zero to cause the submersible sewage pump to run idly or reach the highest overflow level.

[0074] S5. Denote the real-time liquid level of the water tank as h(t), and denote the mark corresponding to the submersible sewage pump when it stops pumping as 0, and vice versa as 1;

[0075] The control of the submersible sewage pump can be divided into the following working conditions:

[0076] 1) Working condition 1: All water pumps stop running and alarm, as shown in formula (16):

[0077]

[0078] 2) Operating condition two: Only start the only variable-frequency submersible sewage pump, as shown in Equation (17):

[0079]

[0080] At this time, use the PID controller to control the variable-frequency submersible sewage pump to make the water tank level stable at the set value y1 (x1 < y1 < x2);

[0081] 3) Operating condition three: Start one variable-frequency submersible sewage pump and one fixed-frequency submersible sewage pump, as shown in Equation (18):

[0082]

[0083] At this time, use the PID controller to control the variable-frequency submersible sewage pump to make the water tank level stable at the set value y2 (x2 < y2 < x3);

[0084] And so on:

[0085] 4) Operating condition m + 2: Start one variable-frequency submersible sewage pump and m fixed-frequency submersible sewage pumps, as shown in Equation (19):

[0086]

[0087] At this time, use the PID controller to control the variable-frequency submersible sewage pump to make the water tank level stable at the set value y m+1 (x m+1 < y m+1 < x m+2 ).

[0088] The PID controller is composed of a regulator that performs proportional, integral, and differential calculations to make the deviation zero. The regulator accepts the difference between the current liquid level and the set liquid level value corresponding to the current liquid level interval, thereby controlling the frequency conversion device to adjust the variable-frequency submersible sewage pump, and the variable-frequency submersible sewage pump makes the regulating pool liquid level gradually stable at the set value;

[0089] The control principle of the PID control system is as follows: See Figure 2 , there is a deviation signal e(t) between the actual output value h(t) and the set input value y(t) of the PID control system. Then, perform proportional, integral, and differential operations on this signal respectively, and perform a certain linear combination on the results. Finally, the generated control signal u(t) acts on the controlled object;

[0090] Among them, the proportional part of PID is called proportional control. When the error of the system increases, the proportional control will increase the output to reduce the error, and the proportional constant K pIt is used to control the incremental size of the output. The larger its value, the faster the output changes. The product of the proportional constant and the error is called the proportional term proportion, as shown in formula (20):

[0091] proportion=K p e(t) (20)

[0092] Among them, e(t)=y(t)-h(t);

[0093] In the present invention, y(t) is the stable liquid level value set in different liquid level intervals, as shown in formula (21):

[0094]

[0095] The integral part of PID is called integral control: the output is adjusted by accumulating the error, and the integral constant K i It controls the size of the integral. It will gradually increase or decrease the output according to the degree of error accumulation, thereby achieving control of the system. The PID integral term intergration is shown in (22):

[0096]

[0097] Among them, T i is the integration time constant;

[0098] The differential part of PID is called differential control: differential control adjusts the output by detecting the rate of change of error, and the differential constant K d It controls the speed of change of the output. When the error changes rapidly, the differential control will increase or decrease the output to adapt to the changes of the system. The PID differential term is shown in (23):

[0099]

[0100] Among them, T d is the differential time constant;

[0101] Combining the three parts gives the output value of the PID controller:

[0102]

[0103] Simplify the above formula and let:

[0104]

[0105] Finally, the following relationship is obtained between the input and output of the PID controller:

[0106]

[0107] It should be noted that in addition to the shutdown of the submersible sewage pump caused by changes in the pool liquid level, the submersible sewage pump also needs to be shut down for maintenance when the remaining life of the equipment is less than the set threshold L.

[0108] The following examples are implemented on the premise of the technical solution of the present invention, and provide detailed implementation methods and specific operation processes, but the protection scope of the present invention is not limited to the following examples. The methods used in the following examples are conventional methods unless otherwise specified.

[0109] [Example]

[0110] Get a dataset of simulated degradation of a submersible sewage pump over time or usage, containing run-to-failure time series data simulated under different failure modes.

[0111] A long short-term memory (LSTM) model based on a deep convolutional network is constructed to predict the remaining life of the equipment. The constructed network performs convolution operations on the LSTM input to state and state-to-state transitions, which includes both time series information and the time-frequency information of the signal.

[0112] After the rationality of the network is determined, it can be used to predict the remaining life of the submersible sewage pump that needs to be tested.

[0113] The m fixed-frequency submersible sewage pumps in the sewage treatment plant's regulating tank are numbered, and the 1 variable-frequency submersible sewage pump is numbered.

[0114] When the submersible sewage pump operates normally, the remaining life of the equipment is predicted according to the equipment remaining life model, and the life value is recorded as the first threshold of the submersible sewage pump; if the submersible sewage pump needs maintenance after a failure, the downtime of the submersible sewage pump from the completion of the maintenance to the reuse is recorded as the second threshold of the submersible sewage pump; the sum of the first threshold and the second threshold of the submersible sewage pump is used as the judgment standard for the starting priority of the submersible sewage pump. The larger the sum of the two, the higher the priority of the corresponding submersible sewage pump. When needed, it is started first, otherwise it is stopped first.

[0115] The submersible sewage pump needs to be repaired after a failure. The first threshold value of the submersible sewage pump before it is put into use after the repair can be taken as a reference to the service life of the submersible sewage pump when it leaves the factory. The second threshold value is 0 when the submersible sewage pump operates normally.

[0116] In addition to the start and stop of the submersible sewage pump caused by changes in the pool liquid level, the submersible sewage pump also needs to be stopped for maintenance when the remaining life of the equipment is less than the set threshold.

[0117] When the first submersible sewage pump is started, the priority of the variable frequency submersible sewage pump is higher than that of the fixed frequency submersible sewage pump.

[0118] After determining the priority of the submersible sewage pump, the submersible sewage pump can be started and stopped according to the control strategy.

[0119] Design the liquid level threshold for stopping the submersible sewage pump: when the liquid level h < x1, all pumps stop running and an alarm is issued.

[0120] Design the liquid level threshold for starting the first submersible sewage pump: when x1 < h < x2, start the first submersible sewage pump. The submersible sewage pump with frequency conversion function is started first. At this time, use the PID controller to control the frequency conversion submersible sewage pump to keep the water tank liquid level stable at the set value y1 (x1 < y1 < x2).

[0121] Design the liquid level threshold for starting two submersible sewage pumps: when x2 < h < x3, start the second submersible sewage pump and start the fixed-frequency submersible sewage pump. At this time, use the PID controller to control the frequency conversion submersible sewage pump to keep the water tank liquid level stable at the set value y2 (x2 < y2 < x3).

[0122] And so on, design the liquid level threshold for starting m + 1 submersible sewage pumps: when x m+1 < h < x m+2 , start the (m + 1)-th submersible sewage pump. At this time, use the PID controller to control the frequency conversion submersible sewage pump to keep the water tank liquid level stable at the set value y m+1 (x m+1 < y m+1 < x m+2 ).

[0123] When the liquid level h > x m+2 , at this time all submersible sewage pumps have been started, and an alarm prompt is sent outwards.

Claims

1. A centralized control method for a submersible sewage pump group based on a residual life model of equipment, characterized in that: The steps include: S1. Train a remaining life model based on a deep convolutional long short-term memory neural network and use it to predict the remaining life of the submersible sewage pump used; S2. Number the fixed-frequency submersible sewage pumps and the variable-frequency submersible sewage pumps, wherein the fixed-frequency submersible sewage pumps are m and the variable-frequency submersible sewage pump is 1; S3, the remaining life of the submersible sewage pump during normal use is recorded as the first threshold of the water pump, and the downtime of the submersible sewage pump after maintenance is recorded as the second threshold of the water pump. The sum of the two is the criterion for determining the start and stop priority of the water pump. The larger the value, the higher the start priority, and vice versa. S4. Set relevant liquid level thresholds according to the size of the regulating tank, the type and quantity of the submersible sewage pumps; S5. When the water level in the regulating tank is lower than the set first threshold, all submersible sewage pumps stop running and an alarm is triggered; S6. When the water level in the regulating tank is higher than the set first threshold value and lower than the second threshold value, the variable frequency submersible sewage pump is started and the liquid level is controlled to be stable at the set value; S7. When the water level in the regulating tank is higher than the second threshold and lower than the third threshold, the variable frequency submersible sewage pump and a fixed frequency submersible sewage pump are started according to the priority and the variable frequency submersible sewage pump is controlled to stabilize the liquid level at the set value; S8, and so on, gradually increase the number of fixed-frequency submersible sewage pumps to be started. When the water level in the regulating tank is higher than the set m+1th threshold and lower than the m+2th threshold, start the variable-frequency submersible sewage pump and m fixed-frequency submersible sewage pumps, and control the variable-frequency submersible sewage pumps to stabilize the liquid level at the set value; S9. When the water level in the regulating tank is higher than the m+2th threshold, an alarm will be given.

2. A method for centralized control of a submersible sewage pump group based on a residual life model of equipment according to claim 1, characterized in that: It also includes obtaining a degradation simulation data set of a submersible sewage pump over time or usage, including vibration signal time series data of operation to failure simulated under different failure modes, which is used to obtain training data for the remaining life model.

3. The centralized control method of a submersible sewage pump group based on the equipment remaining life model according to claim 1 is characterized in that: The submersible sewage pump adopts time-frequency analysis to analyze its vibration signal, and provides time domain and frequency domain information simultaneously by performing short-time Fourier transform on the vibration signal v(t).

4. A centralized control method for a submersible sewage pump group based on a residual life model of equipment according to claim 3, characterized in that: The time-frequency representation of each data set of the vibration signal is a matrix S i , where i = 1, 2, ..., T, T is the number of data sets, and the remaining life prediction of equipment is based on the sequence matrix S i The prediction at time t.

5. The centralized control method of a submersible sewage pump group based on a remaining life model of equipment according to claim 1 is characterized in that: The mean square error of the equipment remaining life model is less than 0.

1.

6. The method for centralized control of a submersible sewage pump group based on a residual life model of equipment according to claim 1, characterized in that: The prediction of the remaining life of the equipment adopts the stochastic gradient descent method to optimize the problem of minimizing the loss function, uses time back propagation to calculate the gradient to optimize the model parameters, and for the top logical layer of the prediction, uses the loss function to calculate the gradient of the weight and bias.

7. The method for centralized control of a submersible sewage pump group based on a residual life model of equipment according to claim 1, characterized in that: The liquid level is stabilized at a set value and a PID controller is used to control a variable frequency submersible sewage pump.

8. The method for centralized control of a submersible sewage pump group based on a residual life model of equipment according to claim 1, characterized in that: If the water level in the regulating tank decreases, the number of fixed-frequency submersible sewage pumps started will be shut down one by one, and finally the number of variable-frequency submersible sewage pumps will be shut down. The stopping priority of the submersible sewage pump is opposite to the starting priority.

Images