Water pump energy-saving optimization method and system for complex water pump deployment form

Abstract

The invention provides a water pump energy-saving optimization method and system for a complex water pump deployment form. The method comprises the following steps: a data acquisition step: acquiring and obtaining operation data of a water pump system; a model training step: preprocessing the data, and training a model based on a preprocessing result to obtain a water pump system model; a parameter optimization step: acquiring real-time operation parameters, and optimizing control parameters of the water pump based on the water pump system model; and a parameter issuing step: issuing and executing the control parameters. By collecting water pump operation data and pipeline data in real time, efficient operation, energy conservation and consumption reduction of a complex water pump system are achieved.

Description

Technical Field

[0001] This invention belongs to the field of energy-saving control technology for water pumps or fluid transport, specifically relating to an energy-saving optimization method and system for water pumps with complex deployment configurations. The complex water pump refers to a water pump system that can be divided into groups a and b, with pumps within each group connected in parallel and pumps between groups connected in series. More specifically, it relates to an energy-saving optimization method for water pump systems based on multi-sensor data acquisition, data cleaning, and Bayesian modeling and optimization algorithms. This method is applicable to complex deployment scenarios such as multiple pumps in parallel, multiple pumps in series, a combination of series and parallel connections, and variable frequency speed control, aiming to achieve efficient operation and energy saving of the water pump system. Background Technology

[0002] In industrial water systems, parallel and series deployment of pumps is a common strategy for addressing variable flow demands and improving system redundancy. However, the coordinated operation of different pump models faces significant challenges, particularly in output distribution and control optimization. Specific issues include: Firstly, uneven output is caused by differences in characteristics. Specifically, the rated parameters of different pump models, such as flow rate, head, and power curves, vary significantly. When dissimilar pumps are connected in parallel, the flow-head curves (QH curves) of each pump are mismatched, causing the flow distribution to deviate from the theoretical optimal value. For example, a high-head pump may "suppress" a low-head pump, causing its operating point to shift to an inefficient zone, or even triggering backflow, resulting in energy waste. In a series configuration, if the head-flow characteristics of the pumps are inconsistent, it may lead to overload of the upstream pump and cavitation of the downstream pump, seriously affecting the stability of the system.

[0003] Secondly, coordination under dynamic operating conditions is difficult. Specifically, in actual systems, load demand fluctuates over time, requiring dynamic adjustment of the pump start-stop combinations and frequencies. Traditional methods often employ fixed priorities or rotation strategies, which cannot respond to changes in operating conditions in real time. For example, in a parallel system, when a new pump is added, the operating point of the existing pumps will drift due to changes in pipeline resistance. If control is delayed, it can easily lead to a decrease in overall efficiency.

[0004] Third, model mismatch and high optimization complexity exist. Specifically, existing control strategies are typically based on simplified models of a single pump, such as the constant efficiency assumption, ignoring the coupling effect of multiple pumps. For combinations of dissimilar pumps, traditional models struggle to accurately predict the correlation between system-level flow rate, head, and power. Furthermore, the optimization problem involves discrete variables, namely the number of pumps starting and stopping, and continuous variables, namely frequency, in mixed-integer nonlinear programming (MINLP), which is difficult to solve and has poor real-time performance.

[0005] The limitations of existing solutions are: 1. Uniform frequency control: Force all water pumps to operate at the same frequency, ignoring differences in characteristics, resulting in some water pumps operating inefficiently for a long time.

[0006] II. Adjustment of empirical rules: Relying on manual setting of priorities or allocation ratios, lacking data-driven optimization, and having poor adaptability.

[0007] 3. Dependence on high-precision models: Some solutions use CFD simulation or high-fidelity models, which have high computational costs and are difficult to apply online.

[0008] Patent document CN111237181A discloses an online identification and optimization control method for the operating characteristics of a water pump system. The method includes: step S100, collecting basic information about the water pump system; step S200, collecting real-time data of the water pump system; step S300, constructing and updating a characteristic model of the water pump system based on the basic information and real-time data; and step S400, controlling the water pump system according to the characteristic model. This method lacks data-driven optimization, has poor adaptability, and high computational cost, making it difficult to apply online.

[0009] This problem urgently needs to be solved. Summary of the Invention

[0010] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for optimizing energy-saving water pumps in complex deployment configurations.

[0011] A water pump energy-saving optimization method for complex water pump deployment configurations provided by the present invention includes: Data acquisition steps: Collect and obtain the operating data of the water pump system; Model training steps: Preprocess the data and train the model based on the preprocessing results to obtain the water pump system model; Parameter optimization steps: Obtain real-time operating parameters and optimize the control parameters of the water pump based on the water pump system model; Parameter distribution steps: The control parameters are distributed and executed.

[0012] Preferably, in the data acquisition step, the operating data includes: frequency, power, flow rate, pump inlet pressure and pump outlet pressure; the frequency of acquiring the operating data is less than or equal to 1 minute.

[0013] Preferably, in the model training step, the preprocessing, i.e., excluding abnormal data in the running data, includes: data with negative power, data with flow rate greater than the rated value, data with negative flow rate, or data with negative pressure difference between the pump outlet pressure and the pump inlet pressure. The pump system model includes: a pump power model, a pump flow rate model, and a pump pressure difference model.

[0014] Preferably, the mathematical expression for the water pump power model is:

[0015] in, Indicates the power of the water pump; Indicates the rated power of the water pump; This indicates the default standard frequency. The similarity index between frequency and power; Indicates frequency; k and These represent one type of trained parameter and another type of trained parameter, respectively. Indicates the water pump flow rate; In the model training step, the pump power model is trained using a Bayesian model, including: Step A1: Setting The rated power of the water pump is 75% to 110%. The range is 2 to 4; k is a positive value; b is a real number; Step A2: Based on the settings in Step A1, evaluate the power training target using a Bayesian algorithm to obtain the trained water pump power model; the power training target includes: k and b; The mathematical expression for the pump flow model is:

[0016] in, , , All parameters are to be trained; P is the pump pressure difference; In the model training step, the pump flow model is trained using a Bayesian model, including: Step B1: Settings It is a positive value. The values ​​are real values, and k3 is 1 to 3; Step B2: Based on the settings in Step B1, evaluate the parameters to be trained using the Bayesian algorithm to obtain the trained water pump flow model; The mathematical expression for the pump pressure difference model is:

[0017] in, , , , b2 and b2 are parameters to be trained; In the model training step, the water pump pressure difference model is trained using a Bayesian model, including: Step C1: Settings , , It is a positive value. b1 to b2 are 1 to 3; Step C2: Based on the settings in Step C1, evaluate the parameters to be trained using the Bayesian algorithm to obtain the trained pump differential pressure model.

[0018] Preferably, in the model training step, the inputs to the water pump system model include: the number and frequency of water pumps being turned on; The outputs of the pump system model include: pump pressure difference, pump power, and pump flow rate; In the parameter sending step, the number of pumps to be turned on, their combination, and the control frequency (Hz) are sent to the PLC or DDC for execution; the combination refers to grouping the pumps of the pump system into groups, with pumps in parallel within each group and pumps in series between groups.

[0019] According to the present invention, a pump energy-saving optimization system for complex pump deployment configurations includes: Data acquisition module: Collects and obtains the operating data of the water pump system; Model training module: preprocesses the data and trains the model based on the preprocessing results to obtain the water pump system model; Parameter optimization module: acquires real-time operating parameters and optimizes the control parameters of the water pump based on the water pump system model; Parameter distribution module: Distributes and executes the control parameters.

[0020] Preferably, in the data acquisition module, the operating data includes: frequency, power, flow rate, pump inlet pressure and pump outlet pressure; the operating data is acquired at a frequency of less than or equal to 1 minute.

[0021] Preferably, in the model training module, the preprocessing, i.e., excluding abnormal data in the running data, includes: data with negative power, data with flow rate greater than the rated value, data with negative flow rate, or data with a negative pressure difference between the pump outlet pressure and the pump inlet pressure. The pump system model includes: a pump power model, a pump flow rate model, and a pump pressure difference model.

[0022] Preferably, the mathematical expression for the water pump power model is:

[0023] in, Indicates the power of the water pump; Indicates the rated power of the water pump; This indicates the default standard frequency. The similarity index between frequency and power; Indicates frequency; k and These represent one type of trained parameter and another type of trained parameter, respectively. Indicates the water pump flow rate; In the model training module, the pump power model is trained using a Bayesian model, including: Module A1: Settings The rated power of the water pump is 75% to 110%. The range is 2 to 4; k is a positive value; b is a real number; Module A2: Based on the settings of Module A1, the power training target is evaluated using a Bayesian algorithm to obtain the trained water pump power model; the power training target includes: k and b; The mathematical expression for the pump flow model is:

[0024] in, , , All parameters are to be trained; P is the pump pressure difference; In the model training module, the pump flow model is trained using a Bayesian model, including: Module B1: Settings It is a positive value. The values ​​are real values, and k3 is 1 to 3; Module B2: Based on the settings of Module B1, the parameters to be trained are evaluated using the Bayesian algorithm to obtain the trained water pump flow model; The mathematical expression for the pump pressure difference model is:

[0025] in, , , , b2 and b2 are parameters to be trained; In the model training module, the water pump pressure difference model is trained using a Bayesian model, including: Module C1: Settings , , It is a positive value. b1 to b2 are 1 to 3; Module C2: Based on the settings of Module C1, the parameters to be trained are evaluated through Bayesian algorithm to obtain the trained water pump differential pressure model.

[0026] Preferably, the inputs to the water pump system model include: the number and frequency of water pumps being turned on; The outputs of the pump system model include: pump pressure difference, pump power, and pump flow rate; In the parameter sending module, the number of pumps to be turned on, their combination, and the control frequency (Hz) are sent to the PLC or DDC for execution; the combination refers to grouping the pumps in the pump system, with pumps in parallel within a group and pumps in series between groups.

[0027] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention solves the problems of low energy efficiency and optimization lag in complex water pump systems by integrating mechanistic models and data-driven methods. It has significant energy-saving effects and strong robustness and real-time response capabilities. It can be widely used in scenarios such as central air conditioning, industrial circulating water, and chemical fluid transportation.

[0028] 2. This invention constructs a pump system model and optimizes energy-saving parameters by collecting real-time pump operation data and pipeline data, thereby achieving efficient operation and energy saving of complex pump systems. Attached Figure Description

[0029] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This invention provides a schematic diagram of the optimized process for a water pump system. Figure 2 A schematic diagram illustrating the calculation process of pressure difference and flow rate under complex series-parallel configurations for the optimized water pump system provided by this invention. Detailed Implementation

[0030] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0031] This invention discloses a pump energy-saving optimization method and system applicable to complex pump deployment configurations, belonging to the field of industrial energy-saving control technology. By collecting real-time pump operation data and pipeline data, a pump system model is constructed and energy-saving parameters are optimized.

[0032] See Figure 1 The method includes the following steps: First, install the necessary sensors and collect the pump operation data in real time via PLC or DDC, including frequency, power, flow rate and inlet and outlet pressure, and store it at a frequency of 1 minute. Secondly, historical data was used and the cleaning and screening were carried out based on the moving window method. A hybrid model was constructed by combining the physical characteristics of the water pump with Bayesian probabilistic inference. The calibration parameters were estimated by maximum a posteriori and the relationship between water pump power, flow rate and pressure difference was accurately characterized. Next, a differential evolution algorithm is used for multivariate optimization, with frequency and number of units in operation as inputs, and power minimization as the objective, while constraining flow rate and pressure difference to be within the upper and lower limits, to solve for the optimal control strategy; Finally, the optimization results are sent to the execution unit via a communication protocol to achieve closed-loop dynamic control.

[0033] In other words, this invention discloses a pump energy-saving optimization method and system suitable for complex pump deployment configurations, comprising the following steps: Sensor installation and data acquisition steps: Install the necessary sensors in the water pump system, connect them to the PLC or DDC controller, and acquire all data from the water pump system from the PLC or DDC. Data is acquired and stored at time intervals of less than or equal to 1 minute.

[0034] Model training steps: First, historical data is preprocessed to remove outliers. Then, following the mixed mechanism model, Bayesian model training methods are used to estimate model parameters. Finally, based on the pump topology, the individual pump models are combined to obtain the pump system model. In other words, Bayesian model training methods are used to estimate model parameters to obtain the pump system model.

[0035] Parameter optimization steps: Obtain the current actual pressure / differential pressure, flow rate and other operating parameters, obtain the maximum / minimum pressure / differential pressure control range, use differential evolution algorithm to optimize the number of pumps in operation, combination and control frequency, and find the pumps in operation, combination and control frequency with the lowest energy consumption under the condition of satisfying the differential pressure constraint range.

[0036] Optimization parameter distribution steps: Distribute the number of water pumps to be turned on, their combination, and control frequency to the PLC or DDC for execution. Perform parameter optimization and optimization parameter distribution every 15 minutes.

[0037] Specifically, during the sensor installation and data acquisition steps, it is necessary to determine whether the following sensors exist; if not, they must be installed: Pressure sensors: used to monitor pipeline pressure. The minimum number required for installation must meet the following conditions: 1. If the water pumps are deployed in parallel, one pressure sensor needs to be installed on each of the main inlet pipeline and the outlet pipeline of the water pumps; 2. If the water pumps are divided into two groups, with parallel connections within each group and series connections between the groups, then one pressure sensor needs to be installed on the inlet main pipeline and one pressure sensor needs to be installed on the outlet pipeline of the water pump system. At the same time, one pressure sensor needs to be installed on the main pipeline between the two groups of water pumps. Specifically, the grouping is based on the pipeline connection method, with pumps that are deployed in parallel within the same pipe section forming a group.

[0038] 3. If the water pumps are divided into multiple groups, with parallel connections within each group and complex series-parallel connections between groups, then one pressure sensor needs to be installed on the inlet main pipeline and the outlet pipeline of each group of water pumps.

[0039] Flow sensors: used to monitor pipeline flow. The minimum number required for installation must meet the following conditions: If the water pumps are deployed in parallel, a flow sensor needs to be installed on the main inlet pipe or the outlet pipe of the water pumps. If the water pumps are divided into two groups, with parallel connections within each group and series connections between the groups, then a flow sensor needs to be installed on the inlet main pipeline or outlet pipeline of the water pump system. Additionally, if there is a bypass branch between the two groups of water pumps, then a flow sensor needs to be installed on that branch. If the water pumps are divided into multiple groups, with parallel connections within each group and complex series-parallel connections between groups, then each group of water pumps needs to have one flow sensor installed on its inlet main pipeline or outlet pipeline.

[0040] Inverter: If the water pump is equipped with an inverter, the frequency control value of the inverter needs to be collected. If no inverter is configured, installing an inverter is the preferred option.

[0041] Electricity meter: Each water pump must be equipped with an electricity meter to collect the instantaneous power of the water pump.

[0042] In one possible implementation, the following data must be collected during the sensor installation and data acquisition steps; none of them can be omitted: 1. Pump inlet pressure, i.e., P_in; 2. Pump outlet pressure, i.e., P_out; 3. Pump operating frequency: If no frequency converter is installed, the default is the mains frequency, i.e., Hz; 4. The real-time power of the water pump, i.e., E.

[0043] Specifically, in the model training step, the first step is data cleaning. Data cleaning uses an outlier elimination method, excluding data that includes: negative power, flow exceeding or being negative, and negative pressure difference ΔP = P_out - P_in.

[0044] After excluding outlier data, a sliding window method is used to perform steady-state filtering on the collected data to remove unstable data.

[0045] The sliding window method smooths data by using a sliding window to eliminate noise caused by short-term fluctuations.

[0046] The typical implementation steps include: parameter setting: determining the window size w and sliding step s based on data characteristics and requirements. Testing showed that w = 10 and s = 1 were suitable.

[0047] Window initialization: Set the window's starting position to the first point of the data sequence.

[0048] Window data processing: Apply window functions to the data within the current window, such as calculating the mean, median, and standard deviation. Tests show that calculating the mean yields the best results. Compare each data point within the window to the mean; points deviating more than 2% from the mean are considered outliers and marked for deletion.

[0049] Move window: Move the window to the right by s, which is 1 data point, and repeat the window data processing steps until the entire dataset has been traversed.

[0050] Results processing: Remove all labeled outliers, and use the resulting subset of data to train the model.

[0051] Specifically, in the modeling process, the second step is model training: 1. Model Form: Firstly, Model 1: Based on the formula of the similarity law of water pumps. Where E is the pump power and Hz is the pump operating frequency, the following model is used to predict the pump power: ,in, k and b are the parameters being trained. The physical meaning is the rated power of the water pump. The physical meaning of is the similarity index between frequency and power. The standard frequency value is 50 by default, and (k*F+b) is the flow rate correction for power. * indicates product.

[0052] Secondly, Model 2: Based on the formula of the similarity law of water pumps. F is the water pump flow rate. The symbol is a direct proportion, Hz is the operating frequency of the water pump, and the following model is used to predict the water pump flow rate: ,in, , , These are the parameters being trained.

[0053] Thirdly, Model 3: Based on the formula of the water pump-phase resistance law. Where P is the pump pressure difference and Hz is the pump operating frequency, the following model is used to predict the pump pressure difference: , in, , , , b2 and b2 are both training parameters.

[0054] 1. Model training methods: To train a Bayesian model, the first step is to determine the prior range of the coefficients using physical principles. Specifically, in Model 1, βE_rated should be 75% to 110% of the rated power value on the pump's nameplate. The value should be between 2 and 4, k should be a positive value, b should be a real number, and all should conform to a normal distribution. Specifically, in Model 2, It should be a positive value. It should be a real value, k3 should be between 1 and 3, and all should conform to a normal distribution; Specifically, in Model 3, a1, a2, and a3 should all be positive values, b1 should be between 1 and 3, b2 should be between 1 and 3, and all should conform to a normal distribution; Then, using the Bayesian method, combined with the prior range of the aforementioned parameters and historical data after at least 14 days of cleaning, the maximum likelihood estimation is performed on the trained parameters to obtain the complete model of the water pump.

[0055] Specifically, in the modeling process, the third step is to build a system model. The inputs to the system model are the number of pumps turned on, i.e., State and the control frequency, i.e. Hz. The outputs of the model are the pressure difference, i.e. P, the flow rate, i.e. F, and the power, i.e. E.

[0056] in: (1) Pressure difference (P) = Model 3 (Hz, F) (2) Flow rate (F) = Model 2 (Hz, P) * State (3) Power (E) = Model 1 (Hz, F) * State Where * represents product.

[0057] Since the model's inputs are only Hz and State, Newton's method is also required. The pressure difference (P) and flow rate (F) must simultaneously satisfy the above formulas (1) and (2). For systems where the pumps are deployed in parallel, the aforementioned method is suitable for the entire system. For the other two types, the following conditions must also be met: When the water pumps are divided into two groups, group a and group b, with parallel connections within each group and series connections between groups, the calculation process also requires additional constraints: the flow rate Fa of group a must be equal to the flow rate Fb of group b, and the total pressure difference P must be the sum of the pressure difference Pa of group a and the pressure difference Pb of group b.

[0058] When water pumps are divided into multiple groups, with parallel connections within each group and complex series-parallel connections between groups, the system should adhere to the law of conservation of water flow mass. The pressure difference increases progressively. The law of conservation of water flow mass means that when the system is decomposed into several series systems, the flow rates of these series systems are equal. The sum of the flow rates of each pump in the parallel pump system within a series system equals the total flow rate of that parallel system. Similarly, the progressively increasing pressure difference means that when the system is decomposed into several series systems, the sum of the pressure differences of each of these series systems equals the total pressure difference of that system. Details regarding the parallel pump system within a series system are as follows... Figure 2 As shown.

[0059] The system model has now been constructed.

[0060] Specifically, in one possible implementation, in the parameter optimization step, the actual pump operating status, control frequency, differential pressure, flow rate, and other parameters in the current system are obtained, and thresholds such as upper and lower limits for differential pressure and upper and lower limits for flow rate are obtained. A differential evolution algorithm is then used to optimize the parameters, with the objective function being... That is, the total power consumption of all water pumps is the lowest, and at the same time, P≥ lower pressure difference limit, P≤ upper pressure difference limit, F≥ lower flow rate limit, and F≤ upper flow rate limit. This indicates the real-time power of the water pump.

[0061] Differential Evolution (DE) is a powerful population-based stochastic optimization algorithm, particularly suitable for solving global optimization problems in continuous spaces. It is simple, robust, and parallelizable, and performs exceptionally well in handling complex optimization problems that are nonlinear, nondifferentiable, multimodal, and noisy. The specific steps of the Differential Evolution algorithm are as follows: Step 1: Initialize the population.

[0062] Setting parameters: NP is the population size, i.e., the number of individuals. Set it to 5 to 10 times the problem dimension D, but this can be adjusted depending on the problem. For water pump optimization problems, the population size usually needs to be set to 500 or more.

[0063] F is the scaling factor, which controls the strength of the difference vector perturbation, typically ranging from [0.4, 1.0]. A larger F value indicates stronger exploratory capabilities; a smaller F value indicates stronger exploitation capabilities. For pump optimization problems, testing suggests that an F value of 0.4-0.5 is preferable.

[0064] CR stands for Crossover Rate. It controls how much of the experimental vector inherits from the mutated vector, ranging from [0, 1]. A larger CR value makes the experimental vector more like the mutated vector; a smaller CR value makes the experimental vector more like the target vector. For pump optimization problems, testing has shown that a CR value of 0.4~0.5 is appropriate.

[0065] G_max represents the maximum number of generations to evolve. For the water pump optimization problem, testing showed that G_max should be set to 200 or higher.

[0066] Randomly generate the initial population: Randomly generate each individual. For each individual i in the population, i = 1, 2, ..., NP: For each dimension j, j = 1, 2, ..., D, randomly generate a value. Specifically, the dimensions here represent: the on / off state of each water pump (state = 0 or 1, 0 indicating off, 1 indicating on); and the control frequency of each water pump, typically within the range [25, 50]. Record this individual as... And set it to 0, that is, G=0.

[0067] Calculate the initial population fitness by inputting the pump control parameter information carried by each individual into the system model to calculate F, P, and E, i.e., F, P, E. f is the system model, and then the calculated E is corrected with a penalty term, i.e. , This is a penalty function that takes effect when F exceeds the upper or lower limits of flow rate or P exceeds the upper or lower limits of differential pressure. The penalty is 50 times the excess value, i.e.: ΔF and ΔP are the absolute values ​​of the differences between F and the upper and lower limits, respectively, after F and P exceed the upper and lower limits.

[0068] Step 2: Main loop, i.e., G=0 to G=G_max-1.

[0069] For each individual in the population This is called the target vector.

[0070] Mutation: the target vector Generate a mutation vector .

[0071]

[0072]

[0073] It is the difference vector, which reflects the direction and magnitude of the difference between the two individuals in the solution space. F scales this difference, controlling the step size of the perturbation. As a basis vector for perturbation. If the mutation vector a certain component Exceeding the boundary If so, its value is restricted to the boundary value. This represents the lowest boundary value of the mutation vector. This represents the highest boundary value of the mutation vector. It represents the product.

[0074] Crossover: Utilizing mutation vectors and target vector Perform cross operations to generate test vectors The aim is to introduce population diversity while preserving some information about the target vector. Binary crossover occurs for each component j of the trial vector, j = 1, 2, ..., D: generating a random number. Generate a random integer. Specifically, ensure that at least one component comes from the mutation vector. D is the maximum value of the component.

[0075] For each component j: if or Specifically, this component of the test vector is taken from the mutation vector; otherwise: Specifically, this component of the test vector is taken from the target vector.

[0076] Selection: In the experimental vector and target vector They make a greedy choice among themselves to determine which one will enter the next generation of the species. .if Specifically, the better experimental vector advances to the next generation. Otherwise, Specifically, the target vector is preserved in the next generation.

[0077] Step 3: Determine the termination condition.

[0078] Check and determine whether the maximum number of generations has been reached. If the result is yes, meaning the termination condition is met, then the algorithm ends and outputs the individual with the best fitness in the current population. This is considered the approximate global optimal solution found. Otherwise, set G = G + 1 and return to the main loop to continue evolution.

[0079] The present invention also provides a pump energy-saving optimization system for complex pump deployment configurations. The pump energy-saving optimization system for complex pump deployment configurations can be implemented by executing the process steps of the pump energy-saving optimization method for complex pump deployment configurations. That is, those skilled in the art can understand the pump energy-saving optimization method for complex pump deployment configurations as a preferred embodiment of the pump energy-saving optimization system for complex pump deployment configurations.

[0080] According to the present invention, a pump energy-saving optimization system for complex pump deployment configurations includes: Data acquisition module: Collects and obtains the operating data of the water pump system; Model training module: preprocesses the data and trains the model based on the preprocessing results to obtain the water pump system model; Parameter optimization module: acquires real-time operating parameters and optimizes the control parameters of the water pump based on the water pump system model; Parameter distribution module: Distributes and executes the control parameters.

[0081] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0082] In the description of this application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0083] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for optimizing energy-saving operation of water pumps in complex deployment configurations, characterized in that, include: Data acquisition steps: Collect and obtain the operating data of the water pump system; Model training steps: Preprocess the data and train the model based on the preprocessing results to obtain the water pump system model; Parameter optimization steps: Obtain real-time operating parameters and optimize the control parameters of the water pump based on the water pump system model; Parameter distribution steps: Distribute and execute the control parameters; The pump system model includes: a pump power model, a pump flow rate model, and a pump pressure difference model; The mathematical expression for the pump power model is: in, Indicates the real-time power of the water pump; Indicates the rated power of the water pump; This indicates the default standard frequency. The similarity index between frequency and power; Indicates frequency; k and These represent one type of trained parameter and another type of trained parameter, respectively. This indicates the water pump flow rate.

2. The pump energy-saving optimization method for complex pump deployment configurations according to claim 1, characterized in that, In the data acquisition step, the operating data includes: frequency, power, flow rate, pump inlet pressure and pump outlet pressure; the operating data is acquired at a frequency of less than or equal to 1 minute.

3. The pump energy-saving optimization method for complex pump deployment configurations according to claim 1, characterized in that, In the model training step, the preprocessing step involves excluding abnormal data from the operating data. The abnormal data includes: data with negative power, data with flow rate greater than the rated value, data with negative flow rate, or data with a negative pressure difference between the pump outlet pressure and the pump inlet pressure.

4. The pump energy-saving optimization method for complex pump deployment configurations according to claim 1, characterized in that, In the model training step, the pump power model is trained using a Bayesian model, including: Step A1: Setting The rated power of the water pump is 75% to 110%. The range is 2 to 4; k is a positive value; b is a real number; Step A2: Based on the settings in Step A1, evaluate the power training target using a Bayesian algorithm to obtain the trained water pump power model; the power training target includes: k and b; The mathematical expression for the pump flow model is: in, , , All parameters are to be trained; P is the pump pressure difference; In the model training step, the pump flow model is trained using a Bayesian model, including: Step B1: Settings It is a positive value. The values ​​are real values, and k3 is 1 to 3; Step B2: Based on the settings in Step B1, evaluate the parameters to be trained using the Bayesian algorithm to obtain the trained water pump flow model; The mathematical expression for the pump pressure difference model is: in, , , , b2 and b2 are parameters to be trained; In the model training step, the water pump pressure difference model is trained using a Bayesian model, including: Step C1: Settings , , It is a positive value. b1 to b2 are 1 to 3; Step C2: Based on the settings in Step C1, evaluate the parameters to be trained using the Bayesian algorithm to obtain the trained pump differential pressure model.

5. The pump energy-saving optimization method for complex pump deployment configurations according to claim 4, characterized in that, In the model training step, the inputs to the water pump system model include: the number and frequency of water pumps being turned on; The outputs of the pump system model include: pump pressure difference, pump power, and pump flow rate; In the parameter sending step, the number of pumps to be turned on, their combination, and the control frequency (Hz) are sent to the PLC or DDC for execution; the combination refers to grouping the pumps of the pump system into groups, with pumps in parallel within each group and pumps in series between groups.

6. A pump energy-saving optimization system for complex pump deployment configurations, characterized in that, include: Data acquisition module: Collects and obtains the operating data of the water pump system; Model training module: preprocesses the data and trains the model based on the preprocessing results to obtain the water pump system model; Parameter optimization module: acquires real-time operating parameters and optimizes the control parameters of the water pump based on the water pump system model; Parameter distribution module: distributes and executes the control parameters; The pump system model includes: a pump power model, a pump flow rate model, and a pump pressure difference model; The mathematical expression for the pump power model is: in, Indicates the power of the water pump; Indicates the rated power of the water pump; This indicates the default standard frequency. The similarity index between frequency and power; Indicates frequency; k and These represent one type of trained parameter and another type of trained parameter, respectively. This indicates the water pump flow rate.

7. The pump energy-saving optimization system for complex pump deployment configurations according to claim 6, characterized in that, In the data acquisition module, the operating data includes: frequency, power, flow rate, pump inlet pressure and pump outlet pressure; the operating data is acquired at a frequency of less than or equal to 1 minute.

8. The water pump energy-saving optimization system for complex water pump deployment configurations according to claim 6, characterized in that, In the model training module, the preprocessing involves excluding abnormal data from the operating data. The abnormal data includes: data with negative power, data with flow rate greater than the rated value, data with negative flow rate, or data with a negative pressure difference between the pump outlet pressure and the pump inlet pressure.

9. The pump energy-saving optimization system for complex pump deployment configurations according to claim 6, characterized in that, In the model training module, the pump power model is trained using a Bayesian model, including: Module A1: Settings The rated power of the water pump is 75% to 110%. The range is 2 to 4; k is a positive value; b is a real number; Module A2: Based on the settings of Module A1, the power training target is evaluated using a Bayesian algorithm to obtain the trained water pump power model; the power training target includes: k and b; The mathematical expression for the pump flow model is: in, , , All parameters are to be trained; P is the pump pressure difference; In the model training module, the pump flow model is trained using a Bayesian model, including: Module B1: Settings It is a positive value. The values ​​are real values, and k3 is 1 to 3; Module B2: Based on the settings of Module B1, the parameters to be trained are evaluated using the Bayesian algorithm to obtain the trained water pump flow model; The mathematical expression for the pump pressure difference model is: in, , , , b2 and b2 are parameters to be trained; In the model training module, the water pump pressure difference model is trained using a Bayesian model, including: Module C1: Settings , , It is a positive value. b1 to b2 are 1 to 3; Module C2: Based on the settings of Module C1, the parameters to be trained are evaluated through Bayesian algorithm to obtain the trained water pump differential pressure model.

10. The pump energy-saving optimization system for complex pump deployment configurations according to claim 9, characterized in that, In the model training module, the inputs to the water pump system model include: the number and frequency of water pumps being turned on; The outputs of the pump system model include: pump pressure difference, pump power, and pump flow rate; In the parameter sending module, the number of pumps to be turned on, their combination, and the control frequency (Hz) are sent to the PLC or DDC for execution; the combination refers to grouping the pumps in the pump system, with pumps in parallel within a group and pumps in series between groups.

Images