A method for optimizing the operation control of a centrifugal pump

Abstract

The present invention discloses a method for optimizing the operation control of a centrifugal pump, which includes an offline calculation module and an online calculation module. The offline calculation module establishes a mathematical model of the centrifugal pump system according to the centrifugal pump theory and outputs the offline data of the centrifugal pump system, and constructs an offline database A; through the actually measured head, efficiency and flow rate data, an offline database B is constructed and corrected by an improved BPNN prediction model. The online calculation module, when the output end of the centrifugal pump system changes, triggers the intelligent control system of the centrifugal pump, searches for the control rate online, and judges whether the state parameter threshold is satisfied to perform operation adjustment. The present invention stores the head, flow rate and rotational speed as tree-shaped data, combines the binary tree search algorithm to quickly search for data under different working conditions, realizes real-time monitoring and optimized control; establishes an improved BPNN prediction model for self-data training to correct the problem of inaccurate mathematical model caused by the aging of the centrifugal pump.

Description

Technical Field

[0001] The present invention relates to the field of centrifugal pump control, and particularly to an intelligent centrifugal pump operation optimization control method. Background Art

[0002] With the continuous development of society, the centrifugal pump control system is used more and more frequently in daily life and engineering, and the requirements for it are also getting higher and higher. In order to maximize the control effect of the system, it is necessary to continuously strengthen the optimization of the centrifugal pump control system in the practical application process. The automatic control system of the centrifugal pump can effectively reduce the control cost input and improve the safety control.

[0003] However, in the current centrifugal pump automatic control system, its detection system lacks a certain timeliness, and the output of the centrifugal pump is also output according to the designed data, and cannot be changed according to the specific situation of the centrifugal pump. In this way, the response to emergencies during operation is insufficient, and a certain number of management personnel are required for manual intervention. In reality, the management personnel of the centrifugal pump control system have incomplete knowledge of the control system and cannot make very accurate judgments and operations. Therefore, it not only affects the effect of the centrifugal pump control system, but also wastes a certain amount of energy. Summary of the Invention

[0004] In view of the above technical deficiencies, the present invention provides an intelligent centrifugal pump optimization control method that can monitor the real-time state of the centrifugal pump and perform efficient optimization and adjustment.

[0005] The present invention is realized through the following technical solutions:

[0006] The centrifugal pump operation optimization control method is divided into two main modules: an offline calculation module and an online calculation module.

[0007] The offline calculation module includes: First, according to the centrifugal pump theory, a mathematical model of the centrifugal pump system is established, and the offline data of the centrifugal pump system is output by using the mathematical model of the centrifugal pump system, and an offline database A of the centrifugal pump system is constructed. Second, through the actually measured head, efficiency and flow rate data, an improved backpropagation neural network BPNN prediction model is established for self-data training to construct an offline database B of the centrifugal pump control system.

[0008] The online calculation module triggers the centrifugal pump intelligent control system according to the change of the output end, combines the offline database A and the binary tree search algorithm to search for the control rate online, and during the operation, judges whether the state parameter threshold is satisfied to control the operation and adjustment of the centrifugal pump system.

[0009] The centrifugal pump operation optimization control method specifically includes the following steps:

[0010] Step 1. Offline calculation process:

[0011] 1.1: Establish a mathematical model of the centrifugal pump system based on the theoretical relationship between the head, efficiency, and flow rate of the centrifugal pump;

[0012] 1.2: According to the above mathematical model of the centrifugal pump system, output the discrete point head H corresponding to all flow rate values within the working range of the centrifugal pump i at the flow rate Q i1 、Q i2 and the rotational speeds n0, n m , after sorting the data in a binary tree structure, form an offline database A.

[0013] Step 2. Online calculation process:

[0014] 2.1: Read that the output end of the centrifugal pump system has changed;

[0015] 2.2: The change in the output end triggers the intelligent control system of the centrifugal pump and enters online calculation;

[0016] 2.3: Use the pre-order traversal in the binary tree traversal algorithm to quickly find the required optimal control rate, and the control rate corresponds to the rotational speed;

[0017] 2.4: Adjust the operation of the change in the output end according to the control rate;

[0018] 2.5: Judge whether the adjusted output head and flow rate data meet the set state parameter thresholds. If false, return to step 2.2; if true, enter the next step;

[0019] 2.6: Establish an offline database B; store the head, flow rate, and rotational speed data actually output by the centrifugal pump system in the offline database B;

[0020] 2.7: When the rotational speed data contains the rated rotational speed in the actual output data, enter step 3.

[0021] Step 3. Establish an improved BPNN prediction model and correct the data in the offline database B.

[0022] The specific steps are as follows:

[0023] 3.1: Confirm the number of parameters in the input layer and output layer, and normalize the data in the offline database B to accelerate the convergence ability of the BPNN prediction model;

[0024] 3.2: Determine the number of neurons in the hidden layer;

[0025] 3.3: Build a BPNN neural network prediction model and train it;

[0026] 3.4: Randomly shuffle the array saved in the offline database B in step 2.6 and input it into the established BPNN prediction model;

[0027] 3.5: Using the improved BPNN prediction model, predict the full characteristic curve of the centrifugal pump to obtain flow rate, head, and rotational speed data, and correct the data in the offline database B. The data in the corrected offline database B is used to quickly find the required optimal control rate when the output of the centrifugal pump system changes again.

[0028] The beneficial effects of the present invention are as follows:

[0029] 1. For the centrifugal pump at the starting moment T0, due to the change in rotational speed, the mathematical model from 0 to T0 is inconsistent with the mathematical model under constant rotational speed. The present invention stores the head, flow rate, and rotational speed as tree-like data and combines the binary tree search algorithm to quickly search for the data of the centrifugal pump under different working conditions, realizing real-time monitoring and efficient optimization adjustment of the centrifugal pump.

[0030] 2. An improved BPNN model is established for self-data training, and the mathematical model relationship of head-flow rate-rotational speed under the current working condition is output every set period of time, which can correct the problem of inaccurate mathematical models caused by the aging of the centrifugal pump. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is the optimization control flow chart of the centrifugal pump of the present invention;

[0032] Figure 2 is the tree-structured data of flow rate and rotational speed under the same head H i ;

[0033] Figure 3 is the head H i tree-structured data. DETAILED DESCRIPTION OF THE INVENTION

[0034] The following further describes the specific embodiments of the present invention with reference to the drawings.

[0035] As Figure 1 shown, the optimization control method for the operation of the centrifugal pump of the present invention includes an offline calculation module and an online calculation module.

[0036] The offline calculation module includes: First, according to the centrifugal pump theory, establish a mathematical model of the centrifugal pump system, and use the mathematical model of the centrifugal pump system to output the offline data of the centrifugal pump system and construct an offline database A of the centrifugal pump system; Second, based on the actually measured head, efficiency, and flow rate data, establish an improved backpropagation neural network BPNN prediction model for self-data training to construct an offline database B of the centrifugal pump control system.

[0037] The online calculation module triggers the intelligent control system of the centrifugal pump according to the change of the output end, combines the offline database A and the binary tree search algorithm to search for the control rate online, and judges whether the state parameter threshold is satisfied during operation to control the operation and adjustment of the centrifugal pump system.

[0038] The operation optimization control method of the centrifugal pump of the present invention is as Figure 1 shown, and specifically includes the following steps:

[0039] Step 1: Perform offline calculation of the operation optimization control method of the centrifugal pump:

[0040] 1.1: Establish a mathematical model of the centrifugal pump system according to the theoretical relationship between the head H, efficiency η and flow rate Q of the centrifugal pump.

[0041] Preferably, according to the head-flow relationship diagram of the centrifugal pump, establish a function model of the head H and flow rate Q at the rated speed n0 and the starting moment T0:

[0042]

[0043] T0: H = AQ 2 + BQ + C

[0044] Where: A, B, C are coefficients, n is any speed, n0 is the rated speed, and T0 is the moment after the centrifugal pump is started of.

[0045] Preferably, according to the efficiency-flow relationship diagram of the centrifugal pump, establish a function model of the flow rate Q with respect to the efficiency η at the rated speed n0:

[0046]

[0047] Where: O, P, J are coefficients, n is any speed, and n0 is the rated speed.

[0048] Preferably, the best working point of the centrifugal pump can be calculated according to the relationship curve of the head H, efficiency η, and flow rate Q, and its position of the best working point is constantly changing according to the continuous change of the flow rate Q.

[0049] 1.2: According to the mathematical model of the centrifugal pump system shown in step 1.1, use programming software to output a series of discrete points of the theoretical head H corresponding to all flow rates Q within the working range of the centrifugal pump. The flow rate Q at the discrete point head H i under i1 , Q i2 and the rotational speeds n0, n m are composed of a binary tree structure data as Figure 2 shown.

[0050] Where H iThe head under a certain working condition, n m is any rotational speed, n0 is the rated rotational speed, Q i2 is the flow rate under the rated rotational speed n0, Q i1 is the flow rate under the transient condition of the working point at startup.

[0051] Store the data of the head H i (i = 1, 2, 3, 4, 5…K) as binary tree structure data, where K is the serial number of the highest point of the head of the centrifugal pump. Figure 3 As shown in, taking K = 7 as an example. When the following conditions are met:

[0052] (1) If the left subtree is not empty, then the values of all nodes on the left subtree are less than its root node, that is, H2 < H1, H4 < H2, H6 < H3.

[0053] (2) If the right subtree is not empty, then the values of all nodes on the right subtree are greater than its root node, that is, H3 > H1, H5 > H2, H7 > H3.

[0054] The head data, flow rate data, and rotational speed data after the above processing form an offline database A.

[0055] Step 2: Perform online calculation for the optimal control of the centrifugal pump operation:

[0056] 2.1: It is read that the output end of the centrifugal pump system changes.

[0057] 2.2: The change of the output end triggers the intelligent control system of the centrifugal pump and enters the online calculation.

[0058] 2.3: Use the preorder traversal in the binary tree traversal algorithm to quickly find the required optimal control rate. The control rate corresponds to the rotational speed, specifically as follows:

[0059] Set the head to be searched as H x , and use the preorder traversal to traverse all the head data in the offline database A. Start searching from the root node. If H x is less than the node value, search in the left subtree. If H x is greater than the node value, search in the right subtree.

[0060] 2.4: Adjust the operation according to the control rate for the change of the output end.

[0061] 2.5: Judge whether the adjusted output head and flow rate data meet the set state parameter thresholds. If it is false, return to step 2.2; if it is true, enter the next step.

[0062] 2.6: Outside the original offline database A, another offline database B is established. The head, flow rate, and rotational speed data actually output by the centrifugal pump system are processed according to the data processing method in step 1.2 and then stored in the offline database B.

[0063] 2.7: When the rotational speed data contains n0 in the actually output data, it indicates that the centrifugal pump has reached the rated rotational speed, and then step 3 is entered.

[0064] Step 3: Establish an improved BPNN prediction model and save it in the offline database B.

[0065] 3.1: Confirm the number of parameters in the input layer and output layer, and normalize the data in the offline database B to accelerate the convergence ability of the BPNN prediction model. The formula is as follows:

[0066]

[0067] In the formula, x1 is the normalized data, x is the data in the training set, x min is the minimum value in the training set, and x max is the maximum value in the training set.

[0068] 3.2: Determine the number of neurons in the hidden layer. The formula is as follows:

[0069]

[0070] In the formula, n is the number of neurons in the input layer; m is the number of neurons in the output layer; a is a constant from 1 to 10; p is the number of neurons in the hidden layer.

[0071] 3.3: Build a BPNN neural network prediction model and train it. The specific calculation steps are as follows:

[0072] (1) Select the sigmod activation function. The input of each neuron in the hidden layer is:

[0073]

[0074] In the formula, x2 is the output data of the output layer, a1 is the activation threshold of the neuron in the hidden layer, and w ij is the weight connecting the input layer and the hidden layer. i is the neuron in the input layer, and j is the neuron in the hidden layer.

[0075] (2) The output of each neuron in the hidden layer is:

[0076] y j = f(S j )

[0077] (3) The output of each neuron in the output layer is:

[0078]

[0079] Wherein, a2 is the activation threshold of the output layer neurons, and w jk is the weight value between the hidden layer and the output layer, j is the neuron of the hidden layer, and k are the neurons of the output layer respectively.

[0080] (4) Calculate the output error t k :

[0081]

[0082] Wherein, Z k is the actual output of the k-th neuron during training, is the expected output of the k-th neuron.

[0083] (5) Update each weight threshold by using an optimization algorithm to accelerate the convergence speed of the BPNN model. The steps are as follows:

[0084] Calculate the velocity of momentum:

[0085] v = βv + (1 - β)ds

[0086] Wherein, v is the gradient calculated by exponential weighted average, β is an exponent for gradient accumulation, ds is the original gradient with respect to s, and update the parameter s: s = s - αv, where α is the learning rate.

[0087] 3.4: Randomly shuffle the array saved in the offline database B in step 2.6 and input it into the established BPNN prediction model.

[0088] 3.5: Use the improved BPNN prediction model to predict the full characteristic curve of the centrifugal pump, obtain the flow rate, head and rotational speed data, and correct the data in the offline database B.

[0089] When the output end changes again to trigger the intelligent control system of the centrifugal pump and enters the online calculation, use the pre-order traversal in the binary tree traversal algorithm to quickly find the required optimal control rate in the offline database B.

[0090] Preferably, the centrifugal pump operation optimization control method online collects the real-time head, rotational speed and flow rate data of the centrifugal pump.

[0091] Preferably, the centrifugal pump operation optimization control method monitors and analyzes the real-time data online.

[0092] Preferably, the operation of the centrifugal pump is adjusted according to the control rate, mainly by adjusting the rotational speed of the centrifugal pump impeller to achieve the adjustment of the head and efficiency.

[0093] Optimally, when adjusting the operation of the centrifugal pump, some external factors affecting the actual operation, such as vibration, noise, etc., also need to be considered.

[0094] Optimally, the operation optimization control method of the centrifugal pump has a self-learning function. The output actual data replaces the theoretical data in the offline calculation database. As the amount of data in the offline database increases, the improved backpropagation neural network (BPNN) is used to train its own data, and the head, flow rate, and rotational speed data in the offline database are corrected, thereby continuously improving the accuracy of the database.

[0095] The operation optimization control method of the centrifugal pump described in the present invention can realize the real-time monitoring and efficient optimization adjustment of the centrifugal pump.

[0096] The content described in this embodiment is only an enumeration of the implementation forms of the inventive concept of the present invention. The protection scope of the present invention should not be regarded as limited to the specific forms stated in the embodiments. The protection scope of the present invention also extends to equivalent technical means that those skilled in the art can think of based on the inventive concept of the present invention.

Claims

1. An optimization control method for centrifugal pump operation, including an offline calculation module and an online calculation module, characterized in that: The offline calculation module includes: First, according to the centrifugal pump theory, establish a mathematical model of the centrifugal pump system. Using the mathematical model of the centrifugal pump system, output the offline data of the centrifugal pump system, and construct an offline database A of the centrifugal pump system; Second, through the actually measured head, efficiency, and flow rate data, establish an improved backpropagation neural network BPNN prediction model for self-data training to construct an offline database B of the centrifugal pump control system. The online calculation module triggers the intelligent control system of the centrifugal pump according to the change of the output end of the centrifugal pump system, combines the offline database A and the binary tree search algorithm to search for the control rate online, and during operation, judges whether the state parameter threshold is satisfied to control the operation and adjustment of the centrifugal pump system. Specifically, it includes the following steps: Step 1. Offline calculation process: 1.1: Establish a mathematical model of the centrifugal pump system based on the theoretical relationship between the head, efficiency, and flow rate of the centrifugal pump; 1.2: According to the above mathematical model of the centrifugal pump system, output the discrete point head H corresponding to all flow values within the operating range of the centrifugal pump i at the flow rate Q i1 、Q i2 and the rotational speeds n0, n m , after sorting the data in a binary tree structure, an offline database A is formed; Step 2. Online calculation process: 2.1: It is read that the output end of the centrifugal pump system changes; 2.2: The change of the output end triggers the intelligent control system of the centrifugal pump and enters the online calculation; 2.3: Use the preorder traversal in the binary tree traversal algorithm to quickly find the required optimal control rate, and the control rate corresponds to the rotational speed; 2.4: Adjust the operation according to the control rate for the change of the output end; 2.5: Judge whether the adjusted output head and flow rate data meet the set state parameter thresholds. If it is false, return to step 2.2; if it is true, go to the next step; 2.6: Establish an offline database B; store the head, flow rate, and rotational speed data actually output by the centrifugal pump system in the offline database B; 2.7: When the rotational speed data contains the rated rotational speed in the actual output data, enter step 3; Step 3. Establish an improved BPNN prediction model and correct the data in the offline database B; the data in the corrected offline database B is used to online search for the control rate when the output end changes again; specifically: 3.1: Confirm the number of parameters of the input layer and output layer, and normalize the data in the offline database B to accelerate the convergence ability of the BPNN prediction model; 3.2: Determine the number of neurons in the hidden layer; 3.3: Build a BPNN neural network model and train it; 3.4: Randomly shuffle the array saved in the offline database B in step 2.6 and input it into the established BPNN prediction model; 3.5: Use the improved BPNN prediction model to predict the full characteristic curve of the centrifugal pump, obtain the flow rate, head, and rotational speed data, and correct the data in the offline database B.

2. According to the centrifugal pump operation optimization control method described in claim 1, wherein: In step 2.4, adjusting the operation of the centrifugal pump according to the control rate is to adjust the head and flow rate by adjusting the rotational speed of the centrifugal pump impeller.

Images