Intelligent scheduling method, system, equipment and medium for joint operation of variable-speed multi-pump parallel system

Abstract

The invention relates to the technical field of centrifugal pump optimization control, in particular to an intelligent scheduling method, system and equipment for joint operation of a variable-speed multi-pump parallel system and a medium. The method comprises the following steps: firstly, calling a numerical method to simulate and calculate performance characteristics of parallel centrifugal pumps and pipeline loss characteristics of a variable-speed multi-pump parallel system to obtain a digital characteristic curve; secondly, by taking the minimum total shaft power as a target function, establishing an intelligent regulation and control mathematical model of the variable-speed multi-pump parallel system under a given variable-speed multi-pump parallel system demand target and various constraint conditions; then improving a particle swarm optimization algorithm according to a golden sine algorithm, and solving to obtain operation data; calling a feedforward artificial neural network to establish a proxy model, and performing regression prediction on operation data; and finally, according to a prediction result of the agent model, an optimal regulation and control scheme of the variable-speed multi-pump parallel system is obtained, so that the energy efficiency and the stability of the variable-speed multi-pump parallel system are improved while more accurate prediction of a scheduling result is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of centrifugal pump optimization control, and in particular to an intelligent scheduling method, system, equipment and medium for coordinated operation of a variable speed multi-pump parallel system. Background Art

[0002] At present, there are mainly two adjustment methods for variable speed multi-pump parallel system in fluid transportation: 1. Manual adjustment based on historical data and manual experience: Manual adjustment generally includes speed adjustment and valve adjustment. During the operation of the equipment, according to the demand for pumped fluid, the speed or valve opening is adjusted to change the flow distribution of the multi-pump parallel system to respond to different task requirements; 2. Automatic control and scheduling using intelligent algorithms: The automatic control and scheduling method using intelligent algorithms uses computers to collect and learn historical data to provide control decision-making solutions for the operation of multi-pump parallel systems.

[0003] The existing technical means and control methods are described in detail as follows: (1) Speed ​​regulation method: The speed regulation method can be divided into variable speed regulation method and throttling regulation method. The variable speed regulation method adjusts the pump operating conditions by changing the speed of the water pump prime mover. This method benefits from the advancement of frequency conversion technology, making variable speed regulation more widely used in water pump operating condition regulation. During variable speed regulation, the executive personnel determine the speed of the centrifugal pump and the changes in the demand for water supply pressure and flow rate, so that the water pump can maintain a high operating efficiency within a certain range. The throttling regulation method is relatively simple. The flow rate is changed by manually adjusting the opening of the centrifugal pump discharge valve. As the valve opening becomes smaller, the flow rate and flow velocity decrease accordingly, but this regulation method will result in large throttling losses and poor economy. If the valve is closed too much and the working flow rate is very small, it will also cause safety hazards such as heating of the pump body and cavitation damage; (2) Valve adjustment method: The valve adjustment method mainly changes the output pressure of the water pump by manually adjusting the opening of the outlet valve. This method is simple and economical. By changing the opening of the outlet valve, the output pressure of the water pump can be effectively adjusted. This method is suitable for occasions where water pressure needs to be adjusted quickly. However, due to excessive reliance on personal experience during manual operation and lack of quantitative execution means, this method will result in unstable system operation and no obvious efficiency improvement when applied; (3) Automatic control and scheduling using intelligent algorithms: In traditional algorithmic control, computer data processing and intelligent algorithms are mainly used to optimize and adjust the operating parameters of the control system. By optimizing the operating parameters based on historical data, a control solution that improves system efficiency and stability is obtained. At present, this method usually includes steps such as data collection, data processing, establishing optimization goals and optimization models. The model is solved by intelligent algorithms to find the optimal operating parameters and achieve the best performance of the system. However, it is difficult to adapt to complex systems with multiple pumps in parallel by relying solely on historical system operation data. The existing intelligent algorithms have poor adaptability in the pumping and transportation links, and the accuracy of the optimization and control decisions provided is low.

[0004] Manual adjustment methods rely on the experience and judgment of operation and maintenance personnel, lack quantitative analysis methods during execution, and are greatly affected by changes in the external environment. It is difficult to make accurate judgments based on the characteristics of different types of equipment units when making allocation decisions based on historical data, resulting in insufficient energy saving and low efficiency improvement.

[0005] Existing dispatching and control technologies based on intelligent algorithms often establish constraints by collecting and analyzing historical data, and use different optimization strategies to adjust low-energy consumption and high-efficiency working configurations. However, the common intelligent optimization methods currently lack algorithm design for fluid machinery such as centrifugal pumps. They only build function models based on historical data, while ignoring the research on the parallel operation characteristics of centrifugal pumps. The applicability of the optimization methods is poor, and the accuracy of the prediction analysis results is low.

[0006] In addition, among the optimization algorithms in existing engineering applications, the genetic algorithm has strong global search capabilities, good parallelism, and strong adaptability, but has the disadvantages of high computational complexity, complex programming implementation, and poor final search capabilities; the artificial bee colony algorithm has the advantages of strong global optimization capabilities, wide application range, and strong robustness, but has the problem of being easily trapped in local optimality and the limitation of being only applicable to global optimization of continuous functions. In engineering practice, it does not contribute much to the scheduling optimization of the overall system, and the response time of the optimization strategy is too long. Summary of the invention

[0007] Aiming at the problems that the existing multi-pump parallel system operation optimization methods have difficulty in making quantitative decisions, inaccurate performance predictions, and the existing optimization algorithms have poor adaptability and slow response speed, the present invention proposes an intelligent scheduling method, system, equipment and medium for the coordinated operation of a variable speed multi-pump parallel system; the method first uses a numerical method to simulate and calculate the performance characteristics of parallel centrifugal pumps and the pipeline loss characteristics of the variable speed multi-pump parallel system to obtain a digital characteristic curve; secondly, the minimum total shaft power is used as the objective function, and under a given variable speed multi-pump parallel system demand target, that is, the demand flow Q a and required lift Ha , and under various constraints, a mathematical model of intelligent control of variable speed multi-pump parallel system is established; then the particle swarm optimization algorithm is improved according to the golden sine algorithm to solve the operation data; and the feedforward artificial neural network is called to establish an agent model to regress and predict the operation data; finally, according to the prediction results of the agent model, the optimal control scheme of the variable speed multi-pump parallel system is obtained, which not only realizes more accurate prediction of the scheduling results, but also improves the energy efficiency and stability of the variable speed multi-pump parallel system.

[0008] The specific implementation contents of the present invention are as follows: An intelligent scheduling method for the coordinated operation of a variable speed multi-pump parallel system, firstly, a numerical method is used to simulate and calculate the variable speed performance characteristics of the parallel centrifugal pumps and the pipeline loss characteristics of the variable speed multi-pump parallel system, and a digital characteristic curve is obtained; secondly, according to the digital characteristic curve, the minimum total shaft power is taken as the objective function, and the required flow rate is given under the given variable speed multi-pump parallel system demand target. Q a and required lift H a , and under various constraints, a mathematical model of intelligent control of a variable speed multi-pump parallel system is established; then, the particle swarm optimization algorithm is improved according to the golden sine algorithm to solve the mathematical model of intelligent control of a variable speed multi-pump parallel system and obtain the operating data; and a feedforward artificial neural network is called to establish an agent model to regress and predict the operating data; finally, according to the prediction results of the agent model, the optimal control scheme of the variable speed multi-pump parallel system is obtained.

[0009] In order to better implement the present invention, further, the variable speed multi-pump parallel system cooperative operation intelligent scheduling method specifically includes the following steps: Step S1: using a numerical method to simulate and calculate the variable speed performance characteristics of parallel centrifugal pumps and the pipeline loss characteristics of a variable speed multi-pump parallel system to obtain a digital characteristic curve; Step S2: According to the digital characteristic curve, taking the minimum total shaft power as the objective function, at a given variable speed multi-pump parallel system demand target, that is, the demand flow Q a and required lift H a , and under various constraints, a mathematical model for intelligent control of variable speed multi-pump parallel system is established; Step S3: solving the intelligent control mathematical model of the variable speed multi-pump parallel system according to the golden sine algorithm, the adaptive strategy and the global optimal guided dynamic reverse learning improved particle swarm optimization algorithm to obtain the operation data; Step S4: calling a feedforward artificial neural network to establish a proxy model, regressing and predicting the running data, and obtaining the prediction results; Step S5: According to the prediction results of the agent model, the optimal variable speed multi-pump parallel system coordinated operation scheduling scheme is obtained in combination with the scheduling decision variables.

[0010] In order to better implement the present invention, further, the step S1 specifically includes the following steps: Step S11: According to the factory characteristic curve of a single centrifugal pump, or by establishing a three-dimensional model of a single centrifugal pump, the characteristic curve of a single centrifugal pump is calculated by using a numerical simulation method based on performance prediction. The characteristic curve of a single centrifugal pump is fitted by a polynomial fitting method to obtain the flow rate of a single pump. Q - Lift H , Single pump flow Q - Shaft power P And single pump flow Q -efficiency η Mathematical expression of characteristic curve; Step S12: According to the centrifugal pump similarity theory, calculate the characteristic curve of a single centrifugal pump under variable speed conditions, and establish the single pump flow rate under variable speed conditions of a single centrifugal pump Q - Lift H , Single pump flow Q - Shaft power P And single pump flow Q -efficiency η Mathematical expression of characteristic curve; Step S13: Calculate the characteristic curve of the variable speed multi-pump parallel system based on the flow superposition and head invariance characteristics of the parallel centrifugal pumps, and establish the total flow of the variable speed multi-pump parallel system Q t - Total shaft power P t And the total flow Q t - Parallel pump head H t Mathematical expression of characteristic curve; Step S14: construct a pipeline loss characteristic curve according to the pipeline net head, the pipe resistance coefficient of the branch pipe before the parallel node, the pipe resistance coefficient of the main pipeline of the variable speed multi-pump parallel system, the flow rate of the branch pipeline, and the total flow rate of the parallel centrifugal pump.

[0011] In order to better implement the present invention, further, the step S2 specifically includes the following steps: Step S21: constructing an objective function according to a characteristic curve of a single centrifugal pump, a characteristic curve of a single centrifugal pump under variable speed conditions, a characteristic curve of a variable speed multi-pump parallel system, a pipeline loss characteristic curve, a total shaft power of a variable speed multi-pump parallel system, a shaft power of each centrifugal pump, a state factor of the centrifugal pump, a speed ratio of the centrifugal pump, the number of parallel centrifugal pumps in operation, and a single pump flow rate; Step S22: constructing constraint conditions; the constraining conditions include: constructing a variable speed multi-pump parallel system head constraint according to the pipeline loss characteristic curve, the parallel pump head characteristic curve and the required head, constructing a start-stop number constraint according to the maximum number of running units, constructing a variable speed multi-pump parallel system flow constraint according to the water supply demand flow, constructing a speed ratio constraint according to the rated speed of the centrifugal pump, and constructing a centrifugal pump operating condition constraint according to the rated flow of the centrifugal pump and the set constraint interval; Step S23: Establishing a mathematical model for intelligent scheduling of a variable speed multi-pump parallel system according to the objective function and the constraints.

[0012] In order to better implement the present invention, further, the step S3 specifically includes the following steps: Step S31: constructing a position update strategy according to the golden sine algorithm; the position update strategy is used to update the position of the particle; Step S32: constructing an adaptive strategy according to the set weight factors and learning factors; the adaptive strategy is used to adaptively update the weight factors and learning factors; Step S33: According to the first u Individuals in v The reverse solution of the value on the dimension is used to construct a dynamic reverse learning strategy guided by the global optimality; the dynamic reverse learning strategy guided by the global optimality is used to complete the dynamic reverse learning according to the dynamic reverse learning strategy guided by the global optimality after the particle swarm algorithm iteration is completed, and the top N individuals with the best fitness are selected, and the reverse population is merged with the initial population to form a new population; Step S34: improving the particle swarm optimization algorithm according to the position update strategy, the adaptive strategy, and the global optimal guided dynamic reverse learning strategy, solving the intelligent scheduling mathematical model of the variable speed parallel pump group according to the improved particle swarm optimization algorithm, and obtaining the operation data.

[0013] In order to better implement the present invention, further, the step S4 specifically includes the following steps: Step S41: calling a single hidden layer feedforward neural network to construct a proxy model; Step S42: Adjust the regression coefficient of the proxy model, use the running data as the input layer, and obtain the prediction result through single hidden layer training.

[0014] In order to better implement the present invention, further, the objective function established in step S21 is:

[0015] in, P t is the total shaft power of the variable speed multi-pump parallel system; P i For thei The shaft power of the centrifugal pump; W i For the i The condition factor of a centrifugal pump; b 1, b 2, b 3. b 4 is the fitting coefficient of the polynomial fitting of the flow-power curve of a single centrifugal pump; k is the speed ratio of the centrifugal pump; m o The number of centrifugal pumps in operation in the variable speed multi-pump parallel system; Q The flow rate is for a single pump.

[0016] Based on the above-mentioned variable speed multi-pump parallel system cooperative operation intelligent scheduling method, in order to better realize the present invention, further, a variable speed multi-pump parallel system cooperative operation intelligent scheduling system is proposed, which is used to execute the above-mentioned variable speed multi-pump parallel system cooperative operation intelligent scheduling method; including a simulation unit, a construction unit, a solution unit, a prediction unit, and an output unit; The simulation unit is used to call a numerical method to simulate and calculate the variable speed performance characteristics of the parallel centrifugal pumps and the pipeline loss characteristics of the variable speed multi-pump parallel system to obtain a digital characteristic curve; The construction unit is used to determine the required flow rate of a given variable speed multi-pump parallel system based on the digital characteristic curve and taking the minimum total shaft power as the objective function. Q a and required lift H a , and under various constraints, a mathematical model for intelligent control of variable speed multi-pump parallel system is established; The solving unit is used to solve the intelligent control mathematical model of the variable speed multi-pump parallel system according to the improved particle swarm optimization algorithm based on the golden sine algorithm to obtain the operation data; The prediction unit is used to call a feedforward artificial neural network to establish a proxy model and regressively predict the operating data; The output unit is used to obtain the optimal control scheme of the variable speed multi-pump parallel system according to the prediction results of the agent model.

[0017] Based on the above-proposed intelligent scheduling method for the coordinated operation of a variable speed multi-pump parallel system, in order to better realize the present invention, further, an electronic device is proposed, including a memory and a processor; a computer program is stored on the memory; when the computer program is executed on the processor, the above-proposed intelligent scheduling method for the coordinated operation of a variable speed multi-pump parallel system is implemented.

[0018] Based on the above-mentioned intelligent scheduling method for the coordinated operation of a variable speed multi-pump parallel system, in order to better realize the present invention, further, a computer-readable storage medium is proposed, on which computer instructions are stored; when the computer instructions are executed on the above-mentioned electronic device, the above-mentioned intelligent scheduling method for the coordinated operation of a variable speed multi-pump parallel system is implemented.

[0019] The present invention has the following beneficial effects: (1) The variable speed centrifugal pump performance characteristic curve simulation and fitting method proposed in the present invention accurately constructs the performance characteristic curve of the variable speed multi-pump parallel system. Compared with the prior art method of constructing a function model based only on historical data, it improves the applicability of the optimization method and the accuracy of the result prediction analysis; (2) The present invention analyzes the constraints of the variable speed multi-pump parallel system operating in a specific high-efficiency range, establishes a mathematical model for the coordinated operation optimization of the variable speed multi-pump parallel system, and solves the problem of the operating conditions of the variable speed multi-pump parallel system and the demand, i.e., the demand flow. Q a , Required lift H a Mismatch issues; (3) The present invention takes the minimum total shaft power as the objective function and establishes an accurate optimization model for finding the best scheduling scheme for the variable speed multi-pump parallel system under different flow requirements under various constraints; (4) The improved particle swarm algorithm based on golden sine (GPSO) proposed in this paper is used to calculate the objective function of a variable speed multi-pump parallel system under multiple constraints, solving the problems of long calculation time, low calculation accuracy and slow convergence speed of traditional algorithms in the optimization and scheduling of pump stations. (5) The present invention integrates the various parameters obtained by the algorithm to form a database, constructs an artificial neural network for predicting allocation results, and quickly generates the optimal variable speed multi-pump parallel system operation scheduling plan based on real-time input data, thereby solving the problem that the traditional variable speed multi-pump parallel system control relies on the experience of the staff, resulting in a large amount of electricity waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flow chart of the coordinated operation optimization method of the variable speed multi-pump parallel system provided by the present invention.

[0021] Figure 2 Flowchart for solving the mathematical model.

[0022] Figure 3 Schematic diagram of the structure of a single hidden layer feedforward artificial neural network.

[0023] Figure 4 This is a schematic diagram of the proxy model.

[0024] Figure 5 This is a bar chart of the search speed of some improved particle swarm optimization algorithms GPSO and other optimization algorithms.

[0025] Figure 6 Iterative convergence curves of some improved particle swarm optimization algorithms GPSO and other optimization algorithms.

[0026] Figure 7 Box plots of statistical indicators of some improved particle swarm optimization algorithm GPSO and other optimization algorithms.

[0027] Figure 8 Schematic diagram of the structure of a single hidden layer feedforward artificial neural network in Example 3.

[0028] Fig. 9 This is the regression prediction curve of the surrogate model.

[0029] Fig.10 This is a curve chart of the maximum water supply demand in the region and the required net head in Example 3. DETAILED DESCRIPTION

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. It should be understood that the described embodiments are only part of the embodiments of the present invention, not all of the embodiments, and therefore should not be regarded as limiting the scope of protection. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technical personnel in this field without making creative work are within the scope of protection of the present invention.

[0031] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "disposed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0032] Embodiment 1: This embodiment proposes a method of first using a numerical method to simulate and calculate the variable speed performance characteristics of parallel centrifugal pumps and the pipeline loss characteristics of a variable speed multi-pump parallel system to obtain a digital characteristic curve; secondly, based on the digital characteristic curve, taking the minimum total shaft power as the objective function, under a given variable speed multi-pump parallel system demand target, that is, the demand flow Q a and required lift Ha , and under various constraints, a mathematical model of intelligent control of a variable speed multi-pump parallel system is established; then, the particle swarm optimization algorithm is improved according to the golden sine algorithm to solve the mathematical model of intelligent control of a variable speed multi-pump parallel system and obtain operation data; and a feedforward artificial neural network is called to establish an agent model to regress and predict the operation data; finally, according to the prediction results of the agent model, the optimal control scheme of the variable speed multi-pump parallel system is obtained; specifically, the following steps are included: Step S1: Calling a numerical method to simulate and calculate the variable speed performance characteristics of the parallel centrifugal pumps and the pipeline loss characteristics of the variable speed multi-pump parallel system to obtain a digital characteristic curve.

[0033] The step S1 specifically includes the following steps: Step S11: According to the factory characteristic curve of a single centrifugal pump, or by establishing a three-dimensional model of a single centrifugal pump, the characteristic curve of a single centrifugal pump is calculated by using a numerical simulation method based on performance prediction. The characteristic curve of a single centrifugal pump is fitted by a polynomial fitting method to obtain the single pump flow Q-head H , Single pump flow Q - Shaft power P And single pump flow Q-efficiency η Mathematical expression of characteristic curve; Step S12: According to the centrifugal pump similarity theory, calculate the characteristic curve of a single centrifugal pump under variable speed conditions, and establish the single pump flow Q-head under variable speed conditions of a single centrifugal pump H , Single pump flow Q - Shaft power P And single pump flow Q -efficiency η Mathematical expression of characteristic curve; Step S13: Calculate the characteristic curve of the variable speed multi-pump parallel system based on the flow superposition and head invariance characteristics of the parallel centrifugal pumps, and establish the total flow of the variable speed multi-pump parallel system Q t - Total shaft power P t And the total flow Q t - Parallel pump head H t Mathematical expression of characteristic curve; Step S14: construct a pipeline loss characteristic curve according to the pipeline net head, the pipe resistance coefficient of the branch pipe before the parallel node, the pipe resistance coefficient of the main pipeline of the variable speed multi-pump parallel system, the flow rate of the branch pipeline, and the total flow rate of the parallel centrifugal pump.

[0034] Step S2: According to the digital characteristic curve, taking the minimum total shaft power as the objective function, at a given variable speed multi-pump parallel system demand target, that is, the demand flowQ a and required lift H a , and under various constraints, a mathematical model of intelligent control of variable speed multi-pump system is established.

[0035] The step S2 specifically includes the following steps: Step S21: constructing an objective function according to a characteristic curve of a single centrifugal pump, a characteristic curve of a single centrifugal pump under variable speed conditions, a characteristic curve of a variable speed multi-pump parallel system, a pipeline loss characteristic curve, a total shaft power of a variable speed multi-pump parallel system, a shaft power of each centrifugal pump, a state factor of the centrifugal pump, a speed ratio of the centrifugal pump, the number of parallel centrifugal pumps in operation, and a single pump flow rate; Step S22: constructing constraint conditions; the constraining conditions include: constructing a variable speed multi-pump parallel system head constraint according to the pipeline loss characteristic curve, the parallel pump head characteristic curve and the required head, constructing a start-stop number constraint according to the maximum number of running units, constructing a variable speed multi-pump parallel system flow constraint according to the water supply demand flow, constructing a speed ratio constraint according to the rated speed of the centrifugal pump, and constructing a centrifugal pump operating condition constraint according to the rated flow of the centrifugal pump and the set constraint interval; Step S23: Establishing a mathematical model for intelligent scheduling of a variable speed multi-pump parallel system according to the objective function and the constraints.

[0036] Step S3: According to the golden sine algorithm, the adaptive strategy and the global optimal guided dynamic reverse learning improved particle swarm optimization algorithm, the intelligent control mathematical model of the variable speed multi-pump parallel system is solved to obtain the operation data.

[0037] The step S3 specifically comprises the following steps: Step S31: constructing a position update strategy according to the golden sine algorithm; the position update strategy is used to update the position of the particle; Step S32: constructing an adaptive strategy according to the set weight factors and learning factors; the adaptive strategy is used to adaptively update the weight factors and learning factors; Step S33: According to the first u Individuals in v The reverse solution of the value on the dimension is used to construct a dynamic reverse learning strategy guided by the global optimality; the dynamic reverse learning strategy guided by the global optimality is used to complete the dynamic reverse learning according to the dynamic reverse learning strategy guided by the global optimality after the particle swarm algorithm iteration is completed, and the top N individuals with the best fitness are selected, and the reverse population is merged with the initial population to form a new population; Step S34: improving the particle swarm optimization algorithm according to the position update strategy, the position update strategy, and the dynamic reverse learning strategy guided by the global optimality, solving the intelligent scheduling mathematical model of the variable speed parallel pump group according to the improved particle swarm optimization algorithm, and obtaining the operation data.

[0038] Step S4: Call the feedforward artificial neural network to establish a proxy model, regress and predict the running data, and obtain the prediction results.

[0039] The step S4 specifically comprises the following steps: Step S41: calling a single hidden layer feedforward neural network to construct a proxy model; Step S42: Adjust the regression coefficient of the proxy model, use the running data as the input layer, and obtain the prediction result through single hidden layer training.

[0040] The objective function established in step S21 is:

[0041] in, P t is the total shaft power of the variable speed multi-pump parallel system; P i For the i The shaft power of the centrifugal pump; W i For the i The condition factor of a centrifugal pump; b 1, b 2, b 3. b 4 is the fitting coefficient of the polynomial fitting of the flow-power curve of a single centrifugal pump; k is the speed ratio of the centrifugal pump; m o The number of centrifugal pumps in operation in the variable speed multi-pump parallel system; Q The flow rate is for a single pump.

[0042] Step S5: According to the prediction results of the agent model, the optimal variable speed multi-pump parallel system coordinated operation scheduling scheme is obtained in combination with the scheduling decision variables.

[0043] Working principle: This embodiment first uses a numerical method to simulate and calculate the variable speed performance characteristics of the parallel centrifugal pump and the pipeline loss characteristics of the variable speed multi-pump parallel system to obtain a digital characteristic curve; secondly, based on the digital characteristic curve, the minimum total shaft power is used as the objective function, and the required flow rate is given under the given variable speed multi-pump parallel system demand target. Q a and required lift H a, and under various constraints, a mathematical model of intelligent control of variable speed multi-pump parallel system is established; then the particle swarm optimization algorithm is improved according to the golden sine algorithm to solve the operation data; and the feedforward artificial neural network is called to establish an agent model to regress and predict the operation data; finally, according to the prediction results of the agent model, the optimal control scheme of the variable speed multi-pump parallel system is obtained, which not only realizes more accurate prediction of the scheduling results, but also improves the energy efficiency and stability of the variable speed parallel pump group.

[0044] Embodiment 2: This embodiment is based on the above embodiment 1. Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 , Fig. 9 , Fig.10 As shown, a specific embodiment is described in detail.

[0045] Step S1: Use numerical methods to perform numerical simulation calculations on the variable speed performance characteristics of the centrifugal pump and the pipeline loss characteristics of the variable speed multi-pump parallel system to predict the variable speed performance of the centrifugal pump and determine the flow rate of a single variable speed centrifugal pump. Q - Lift H , Single pump flow Q - Shaft power P And single pump flow Q -efficiency η Mathematical expression of the characteristic curve.

[0046] The performance of a centrifugal pump is usually measured by a series of key parameters, including speed n , Single pump flow Q , Lift H , shaft power P And efficiency η These parameters provide detailed information about the operating characteristics of a centrifugal pump. Specifically, the characteristic curves of a centrifugal pump are an intuitive way to describe its performance. They reflect the relationship between the flow rate, head, power and efficiency of a centrifugal pump at a specific speed.

[0047] The operating state of the centrifugal pump system will change due to factors such as demand changes and adjustment of the variable speed multi-pump parallel system. Therefore, high-precision digital modeling of the centrifugal pump characteristic curve is a prerequisite for establishing a mathematical model for intelligent control of the variable speed multi-pump parallel system.

[0048] This embodiment is a centrifugal pump Q - H The characteristic curve is fitted by quadratic polynomial as shown in formula (1.1): (1.1) In the formula a 1, a 2, a 3 is the fitting coefficient, and its value is determined by comparing the test results with the fitting curve.

[0049] Q - P The curve fitting adopts cubic polynomial fitting, and its expression is shown in formula (1.2): (1.2) In the formula b 1, b 2, b 3. b 4 is the fitting coefficient.

[0050] Q - η The curve is fitted by cubic polynomial fitting method, and its expression is shown in formula (1.3): (1.3) In the formula c 1, c 2, c 3. c 4 is the fitting coefficient.

[0051] The method of constructing the variable speed digital characteristic curve of a centrifugal pump is as follows: When the speed of a centrifugal pump changes, its characteristic curve will change accordingly. Equation (1.4) reveals the specific proportional relationship between the performance parameters of centrifugal pumps when they are similar in geometry and flow conditions: , , , (1.4) In the formula, n 0 is the rated speed, They are the characteristic parameters of the centrifugal pump at rated speed. η 1 and η 2 are the corresponding efficiencies of the centrifugal pump at two speeds.

[0052] Define the speed ratio k=n / n 0, substitute the above formula into formula (1.1) to formula (1.3), then the digital expression of the characteristic curve of the centrifugal pump under variable speed is shown in formula (1.5)-formula (1.7): (1.5) (1.6) (1.7) The method of constructing the variable speed digital characteristic curve of parallel centrifugal pumps is as follows: when m o When two centrifugal pumps of the same model are operated in parallel, the total flow rate is Q t and total shaft power P t The flow rate of a single centrifugal pump is Q and shaft power P of m o times, that is: (1.8) (1.9) Substituting equation (1.8), equation (1.9) and the speed ratio k Substituting into equation (1.5) and equation (1.6), we can get the variable speed of multiple pumps in parallel: Q t - H t and Q t - P t Digital characteristic curve: (1.10) (1.11) In the formula H t is the head of the parallel centrifugal pumps. At this time, the head of each pump is H t ; P t is the total shaft power of the parallel centrifugal pumps; Q t is the total flow of parallel centrifugal pumps; m o The number of centrifugal pumps in operation in the variable speed multi-pump parallel system; k is the speed ratio.

[0053] The calculation of pipeline loss characteristics is constructed as follows: When constructing the mathematical model of a variable speed multi-pump parallel system, it is necessary not only to consider the variable speed characteristics of the parallel pumps, but also to accurately predict and model the pipeline loss of the variable speed multi-pump parallel system. Assuming that the flow of water in the pipeline is ideal, that is, the flow is relatively stable and incompressible, the relationship between the pipeline flow and the head loss can be expressed as: (1.12) in, hloss is the pipeline head loss, λ is the friction coefficient of the inner wall of the pipeline; L is the length of the long straight pipe, in meters; D is the inner diameter of the long straight pipe, in m; Q g is the volume flow rate in the pipe, in m³ / h; g is the acceleration due to gravity.

[0054] When there is a certain terrain difference between the inlet and outlet water of the variable speed multi-pump parallel system, that is, when the net head of the pipeline is not zero, it is necessary to overcome the net head to do work. Since each pump draws water from the same position and the pipeline layout is the same, ignoring the loss difference of each branch pipe before the parallel node, the pipeline loss characteristic curve model of the variable speed multi-pump parallel system can be written as follows: (1.13) In the formula, H x is the pipe resistance head of the variable speed multi-pump parallel system, in m; H st is the net head of the pipeline, in m; S i is the pipe resistance coefficient of each branch before the parallel node, in units of h 2 / m 5 ; S p is the pipe resistance coefficient of the main line of the variable speed multi-pump parallel system, in units of h 2 / m 5 ; Preferably, in this embodiment, a mathematical model for intelligent scheduling of a variable speed multi-pump parallel system is constructed by constructing a digital model of the variable speed characteristics of the multi-pump parallel system through numerical simulation, establishing an objective function, setting constraints, etc.

[0055] Step S2: Based on the digital characteristic curve obtained by the above solution, the minimum total shaft power is determined as the objective function and the constraint condition of operating in a specific high-efficiency range, and a mathematical model for intelligent control of the variable speed multi-pump parallel system is established. The objective function is established as follows: Different choices of objective functions will have an important impact on the overall difficulty of solving the optimization problem and the solution results. Taking the minimum total shaft power of the multi-pump parallel system as the objective function, therefore: (2.1) In the formula, P t Indicates the total shaft power of the variable speed multi-pump parallel system; P i Indicates iThe shaft power of the centrifugal pump; W i For the i The state factor of a centrifugal pump indicates i Working status of a centrifugal pump (1-working status, 0-not working status); k is the speed ratio of the centrifugal pump; m o The number of centrifugal pumps in operation in the variable speed multi-pump parallel system; Q This is the flow rate of a single pump.

[0056] Preferably, in this embodiment, the constraints may include head constraints of variable speed multi-pump parallel system, constraints on the number of start-stop units, flow constraints of variable speed multi-pump parallel system, speed ratio constraints, pump group working condition constraints, etc. The constraint conditions are set as follows: Setting of head constraint conditions for variable speed multi-pump parallel system: Since the variable speed multi-pump parallel system must overcome the pipeline net head and lost work during operation, and considering the stability of the variable speed multi-pump parallel system, its head constraint is expressed as: (2.2) In the formula, m o The number of centrifugal pumps in operation in the variable speed multi-pump parallel system; The head of each operating centrifugal pump; H a To meet the demand.

[0057] Setting of constraints on the number of start and stop units: (2.3) In the formula, m max The maximum number of pumps designed to operate in a pump station.

[0058] Flow constraint setting for variable speed multi-pump parallel system: (2.4) in, Q a The demand flow.

[0059] Speed ​​ratio constraint setting: Due to the limitations of support, lubrication and heat dissipation, the operating speed of centrifugal pumps and motors generally does not exceed the rated speed. When the operating speed is less than a certain value, the motor efficiency will drop significantly. The centrifugal pump speed ratio constraint can be taken as shown in formula (2.5): (2.5) Setting of operating constraints for parallel centrifugal pumps: When the centrifugal pump is working near the design operating point, the streamline in the pump is relatively stable and has a high efficiency. At this time, the pump has a high operating reliability. When it is seriously off-condition, the efficiency will drop significantly. There are often vortices, cavitation and erosion in the pump that reduce the service life of the centrifugal pump. At this time, the pump has a low operating reliability. QUR r ,1.1 QUR r ] The operating condition constraints of the operating flow range are used to ensure the stability of operation. The operating condition constraints are shown in formula (2.6): (2.6) In the formula, Q r It is the design flow rate at rated speed.

[0060] The total flow range of the variable speed multi-pump parallel system at any speed is: (2.7) Step S3: Combining the characteristics of the particle swarm optimization algorithm and the golden sine algorithm, an improved particle swarm optimization algorithm based on the golden sine, GPSO, is proposed. The mathematical model is solved by introducing adaptive strategies and dynamic reverse learning guided by the global optimal method.

[0061] In this embodiment, the mathematical model is solved by constructing an improved particle swarm algorithm GPSO based on the golden sine, and the performance of global search and local search is balanced by introducing adaptive strategies and dynamic reverse learning guided by the global optimum.

[0062] In this embodiment, the improved particle swarm algorithm is based on the golden sine and integrates the improved particle swarm algorithm GPSO with the dynamic reverse learning strategy of adaptive acceleration and global optimal guidance.

[0063] like Figure 2 As shown in the figure, the specific implementation steps of the improved particle swarm optimization GPSO algorithm are as follows: Step 1: Given a population size N, a maximum number of iterations T, and randomly initialize the position of the population; Step 2: Calculate the fitness value of the individual according to the minimum total shaft power of the objective function; Step 3: Record the global optimal and individual optimal positions and determine whether the termination criteria are met. If the termination criteria are not met, execute step 4. If the termination criteria are met, execute step 8. Step 4: Perform adaptive factor update to change weight factors and learning factors; Step 5: Update the velocity and position of the particle; Step 6: Perform reverse dynamic learning guided by global optimality; Step 7: Solve the golden sine optimization algorithm and process the boundary conditions. After the boundary conditions are processed, return to step 2. Step 8: Output the solution results.

[0064] Among them, the termination criteria include the maximum number of loops and the required accuracy; when the maximum number of loops or the required accuracy is reached, it will be judged that the termination criteria are met, the loop will be stopped, and the solution result will be output.

[0065] In this embodiment, the location update strategy based on the golden sine algorithm is as follows: Based on the particle swarm algorithm, the particles are allowed to traverse all points on the unit circle, so that the individuals tend to the optimal value. Since both the particle swarm optimization algorithm and the golden sine algorithm use particles as individuals, the position update formula of the golden sine algorithm is: (3.1) In the formula, x u, l For iteration l Second time individual u location, P o is its optimal position, R 1 and R 2 is a random number, which determines the moving distance and position update direction of the individual in the next iteration, and .

[0066] θ 1 and θ 2 is the introduced golden section number, as shown in formula (3.2): ; (3.2) Among them, τ is the golden ratio, .

[0067] In this embodiment, the adaptive strategy is as follows: The adaptive strategy is to adjust the various parameters of the algorithm, mainly by adjusting the weight factor (3.3) w， Then balance the algorithm's global search ability and local exploration ability, as well as the adjustment formula (3.4) learning factor c , to affect the local exploration ability of the algorithm. The following is the adjustment formula for the weight factor and the learning factor: (3.3) In the formula, w max =1.2, w min=0.4 are the maximum and minimum values ​​of the selected weight factors, d u,v For the v The distance in latitude, d max,v For the v Maximum distance in latitude; (3.4) In the formula, c 1,u,v is the individual learning factor, c 1,max =2.5, c 1,min =0.5 are the maximum and minimum values ​​of the selected individual learning factors respectively; (3.5) In the formula, c 2,u,v is the overall learning factor, c 2,max =3.0, c 2,min =0.6 are the maximum and minimum values ​​of all selected learning factors respectively.

[0068] In this embodiment, the dynamic reverse learning strategy guided by global optimality is as follows: The reverse learning strategy is to evaluate the fitness values ​​of the current solution and the reverse solution at the same time, and select the better solution as the next generation individual. x u,v The current population u Individual in the v The reverse solution is: (3.6) In the formula, ub is the upper bound of the population value range, lb is the lower bound of the population value range, RAND(0,1) means generating a random number in (0,1).

[0069] In order to improve the global performance of the particle swarm algorithm, after each iteration, the global optimal position guides the population to complete dynamic reverse learning by formula (3.6) and selects the former with better fitness. N Individuals are added to merge the reverse population with the initial population to form a new population. The reverse learning strategy can expand the search space while avoiding the time-consuming blind search, and improve the population distribution while ensuring the convergence of the algorithm.

[0070] Step S4: Use the feed-forward artificial neural network FFNN to perform regression prediction on various parameters and establish a proxy model.

[0071] This embodiment is based on a single hidden layer feedforward artificial neural network to construct a neural network proxy model: When considering practical problems, electricity prices, flow demand, lift, etc. often change in equal time periods. If the parameters are substituted into the optimization algorithm for solution each time, it is often difficult to obtain the required optimization results in a timely manner, that is, the response time is long. For this reason, this embodiment proposes to use a feedforward artificial neural network to establish a proxy model, perform regression prediction on the data, and perform fitting processing to solve the problem of long response time.

[0072] Preferably, in this embodiment, the single hidden layer feedforward artificial neural network used is as follows: (4.1) In formula (4.1), x q is the input variable; y p,q For the p Layer q The output signal of a neuron; w is the weight factor; b p,q For the p Layer q The bias of each neuron; Neu is the total number of neurons in this layer; Neu p-1 For the p -Total number of neurons in layer 1; l is the number of iterations; For the p -1st floor e The neurons are progressing to l The weight factor at the iteration; is the output signal of the e-th neuron in the p-1th layer; f p,q () is the activation function; e is the neuron count variable.

[0073] 70% of the data in the sample database is used to train the model, and 30% is used to evaluate the performance of the model. In order to quantify the fitting effect of the surrogate model, the regression coefficient is used to judge the fitting performance. Adjust the regression coefficient It can represent the regression performance of the model prediction, and its formula is shown in formula (4.2). Adjusting the regression coefficient can ignore the impact of the number of samples on the results. The closer the value is to 1, the better the model predicts the data. >0.9 is a necessary condition for solving engineering optimization problems: (4.2) In the formula, S n is the total number of samples, np is the number of features, y z is the sample target value, is the sample target mean, is the model prediction value, and z is the number of features.

[0074] This network structure enables the network to learn more complex feature representations, thereby having better performance when dealing with complex problems. By introducing a prefix structure, the network can process input data more flexibly and improve the generalization ability of the model.

[0075] Step S5: Obtain the optimal control scheme for the variable speed multi-pump parallel system through the prediction results of the agent model.

[0076] In this embodiment, based on the prediction results of the agent model and combined with the scheduling decision variables, the optimal control scheme of the multi-pump parallel system is obtained as follows: like Figure 3 As shown, the single hidden layer feedforward neural network mainly includes a hidden layer, an activation function, and an output layer. Figure 4 As shown in the figure, the data obtained by solving the mathematical model by the improved particle swarm algorithm GPSO are used as the input layer. After the hidden layer is trained, the final output layer uses the predicted parameter values ​​and combines the scheduling decision variables to obtain the optimal scheduling solution; among which, the scheduling decision variables are Q a , that is, the target flow rate of the variable speed multi-pump parallel system.

[0077] The model quickly generates optimal pump station operation scheduling plans based on real-time input data. These plans include pump start and stop decisions, speed adjustment, etc., aiming to minimize energy consumption and operating costs while meeting water supply needs.

[0078] The other parts of this embodiment are the same as those of the above-mentioned embodiment 1, and thus will not be described in detail.

[0079] Embodiment 3: This embodiment is based on any one of the above embodiments 1 to 2. Figure 5 , Figure 6 , Figure 7 , Figure 8 , Fig. 9 , Fig.10 As shown, a detailed explanation is given by taking four centrifugal pumps of the same model working in parallel as an example.

[0080] A mathematical model of a variable speed multi-pump parallel system is established; in this embodiment, four centrifugal pumps of the same model are connected in parallel as a test scheme, and the scheme before optimization operation always keeps the centrifugal pumps running at the same speed.

[0081] By adopting this implementation mode, by establishing a mathematical model that accurately describes the operating characteristics of the variable speed multi-pump parallel system, the mathematical model is improved, which can better describe the working conditions of the variable speed multi-pump parallel system.

[0082] To sum up, the fitness function is: (4.3) In the formula, fit is the fitness value.

[0083] Performance test of improved particle swarm algorithm GPSO; specifically, to verify the performance of the improved particle swarm algorithm GPSO of the present invention, comparative tests are carried out on 23 test functions with basic particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, genetic algorithm GA and sparrow search algorithm SSA.

[0084] Set relevant parameters for relevant optimization algorithms: c 1=2, c 2=2, set relevant parameters for the adaptive particle swarm algorithm APSO: c 1=2, c 2=2, w max =0.9, w min =0.4, and the relevant setting parameters for the chaotic particle swarm algorithm CPSO are c 1=2, c 2=2, φ = c 1+ c 2. Shrinkage Factor χ for , the relevant setting parameters for GA are P c =0.8, P m = 0.05. The algorithm has the same settings in all test functions, where the population size N is 50, the maximum number of iterations is 1000, and the number of runs is 30.

[0085] When testing the performance of an algorithm, four indicators are often used: optimal value, worst value, median, and standard deviation. However, the persuasiveness of algorithm performance testing based solely on commonly used statistical indicators is insufficient. In optimization algorithms, search speed is an important performance indicator that measures the ability of an algorithm to find the optimal solution within a given time. Improving the search speed usually means that the algorithm can converge to the optimal solution faster, thereby reducing computing time and resource consumption. Therefore, this article also uses CRI to examine the search speed of the algorithm. Figure 5As shown in the figure, (a) is the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, genetic algorithm GA in the test function f 6, where (b) is the search speed bar graph of the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, and genetic algorithm GA in the test function f 13 The search speed bar chart of the above test, where (c) is the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, genetic algorithm GA in the test function f 15 The search speed bar chart of the above test, where (d) is the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, genetic algorithm GA in the test function f 18 The search speed bar graph of the above test.

[0086] From Table 1-1 and Figure 5 It can be seen that the improved particle swarm algorithm GPSO performs well in most of the indicators. Whether it is a low-dimensional problem or a high-dimensional problem, the improved particle swarm algorithm GPSO shows significant advantages. Compared with several other optimization algorithms, the improved particle swarm algorithm GPSO can often quickly converge to the solution. It can be seen that the improved particle swarm algorithm GPSO has a clear advantage in search speed. Even when the number of iterations is high, it still maintains a fast search speed. It is obviously not enough to have an advantage in search speed alone. This test also compares the comprehensive performance of the improved particle swarm algorithm GPSO with other optimizations. The speed of approaching the optimal solution and the speed at which a sequence approaches its limit value are also indicators of the performance of the optimization algorithm.

[0087] like Figure 6 As shown in the figure, (a) is the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, genetic algorithm GA in the test function f 5, where (b) is the iterative convergence curve of the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, and genetic algorithm GA in the test function f 8, where (c) is the iterative convergence curve of the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, and genetic algorithm GA in the test function f14 The iterative convergence curves of the above test, where (d) is the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, genetic algorithm GA in the test function f 19 Iterative convergence curve of the above test. The improved particle swarm optimization algorithm GPSO is significantly better than other optimization algorithms in terms of convergence speed and fitness level.

[0088] like Figure 7 As shown in Figure 1, (a) is a box plot of the statistical indicators of the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, and genetic algorithm GA tested on the test function f7, and (b) is a box plot of the statistical indicators of the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, and genetic algorithm GA tested on the test function f 8 statistical indicators of the test box plot, where (c) is the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, genetic algorithm GA in the test function f 20 The statistical indicators of the above test box plot, where (d) is the particle swarm optimization algorithm PSO, adaptive particle swarm algorithm APSO, chaotic particle swarm algorithm CPSO, improved particle swarm algorithm GPSO, genetic algorithm GA in the test function f 22 The statistical indicators box plot of the above test. At the same time, the optimal value and standard deviation were selected in the statistical indicators, such as Figure 7 As shown in Table 1-2 and Table 1-3, in the statistical indicators, the improved particle swarm optimization algorithm GPSO is significantly better than other optimization algorithms in most cases. At the same time, in order to better verify the test results of solution accuracy, as shown in Table 1-4, the Friedman test with a significance level of 0.05 was used to compare the various optimization algorithms involved in the test. It was found that the improved particle swarm algorithm GPSO has a clear advantage in dealing with this problem, followed by the adaptive particle swarm algorithm APSO, the particle swarm optimization algorithm PSO and the chaotic particle swarm algorithm CPSO have similar solution effects, and the genetic algorithm GA has the worst effect.

[0089] Table 1-1 Comparison of search speed of improved particle swarm algorithm GPSO and other optimization algorithms

[0090] Table 1-2 Comparison of optimal values ​​of improved particle swarm algorithm GPSO and other optimization algorithms

[0091] Table 1-3 Comparison of standard deviations of improved particle swarm optimization algorithm GPSO and other optimization algorithms

[0092] Table 1-4 Improved particle swarm algorithm GPSO and other optimization algorithms Friedman sorting results

[0093] Building an agent model; In this example test, if Figure 8 As shown in the figure, due to the small number of fitting data samples, only a single hidden layer cascade feedforward neural network is used for fitting and prediction, and compared with the actual optimization data in the later stage.

[0094] The simulation training process uses the gradient descent method as the training algorithm. The training goal is that the root mean square error between the predicted value and the actual value is less than 0.001, and the number of training steps is not more than 1000. During the training process, 70% of the data in the sample database is used to train the model, and 30% is used to evaluate the performance of the model. In order to quantify the fitting effect of the surrogate model, the regression coefficient is used to judge the fitting performance.

[0095] Adjusted regression coefficients The regression performance of the model prediction can be expressed by fitting the above method to 100 sets of data obtained when the net head is 10m. The results are as follows: Fig. 9 As shown, (a) is the proxy model Q a - P t Prediction curve; (b) Q a - k Prediction curve. Q a - P t and Q a - k The adjusted regression coefficient They are 0.9768 and 0.9149 respectively. Although the sample data is small, the adjusted regression coefficients are all greater than 0.9. In actual engineering examples, the algorithm can be used to solve the problem in the early stage, a database can be formed after a period of time, and then an agent model can be established to quickly obtain the optimal solution.

[0096] Specifically, based on the above solution, an example application is performed: The scale of a water supply pump station in a certain area of ​​City A is 80,000m 3 / d, the pump station uses 4 double-suction centrifugal pumps of the same model, the pump parameters are shown in Table 2-1, the maximum water supply demand and time-of-use electricity price in a certain area of ​​the city are as follows Fig.10 As shown in the figure, the current dispatching method of the pump station is to increase the number of running pumps only when the existing open water pumps are not enough to meet the water supply demand, and all four pumps are operated as fixed speed pumps. The staff of the pump station will adjust the number of pumps and valve openings of the pump station according to the water supply demand. The motor power consumption cost of the pump station is priced by time according to the national standard, of which the base price of electricity is 0.617 yuan, and the electricity price from 11:00 to 2:00 and 17:00 to 21:00 is 0.952 yuan. The prices at other times are all base prices. The current operation results of the pump station are shown in Table 2-2, and the results of operation optimization using the optimization algorithm in this article are shown in Table 2-3.

[0097] From the results in Table 2-2 and Table 2-3, it can be seen that the current daily operating cost of the pump station is 18614.34 yuan. By using the optimization algorithm of this embodiment to solve, it can be obtained that the daily operating cost of the pump station after optimization is 13742.09 yuan. The power saving rate of the improved optimization scheme is:

[0098] Compared with the existing improved optimization algorithm with a power saving rate of about 10%, this optimization algorithm has lower energy consumption loss, and compared with the current solution, the improved solution reduces the number of pump starts and stops, which improves the service life of the equipment to a certain extent.

[0099] The single hidden layer prefix neural network in the artificial neural network is used to perform regression prediction fitting processing on the optimized data, and verified based on the actual optimization data, and it is found that the prediction effect is good. The improved scheme is applied to the actual engineering example, and it is found that compared with the current operation scheme of the pump station, the improved optimization scheme has a power saving rate of 26.17% and reduces the number of pump starts and stops.

[0100] Table 2-1 Parameters of variable speed double suction pump set

[0101] Table 2-2 Daily operation results of the pump station at the current stage

[0102] Table 2-3 Pump station daily operation optimization results

[0103] The other parts of this embodiment are the same as any of the above-mentioned embodiments 1 to 2, and thus will not be described in detail.

[0104] Embodiment 4: Based on any one of the above-mentioned embodiments 1 to 3, this embodiment proposes an intelligent scheduling system for the coordinated operation of a variable speed multi-pump parallel system, which is used to execute the above-mentioned intelligent scheduling method for the coordinated operation of a variable speed multi-pump parallel system; it includes a simulation unit, a construction unit, a solution unit, a prediction unit, and an output unit; The simulation unit is used to call a numerical method to simulate and calculate the variable speed performance characteristics of the parallel centrifugal pumps and the pipeline loss characteristics of the variable speed multi-pump parallel system to obtain a digital characteristic curve; The construction unit is used to determine the required flow rate of a given variable speed multi-pump parallel system based on the digital characteristic curve and taking the minimum total shaft power as the objective function. Q a and required lift H a , and under various constraints, a mathematical model for intelligent control of variable speed multi-pump parallel system is established; The solving unit is used to solve the intelligent control mathematical model of the variable speed multi-pump parallel system according to the improved particle swarm optimization algorithm based on the golden sine algorithm to obtain the operation data; The prediction unit is used to call a feedforward artificial neural network to establish a proxy model and regressively predict the operating data; The output unit is used to obtain the optimal control scheme of the variable speed multi-pump parallel system according to the prediction results of the agent model.

[0105] This embodiment also proposes an electronic device, including a memory and a processor; a computer program is stored on the memory; when the computer program is executed on the processor, the above-mentioned intelligent scheduling method for coordinated operation of a variable speed multi-pump parallel system is implemented.

[0106] This embodiment also proposes a computer-readable storage medium, on which computer instructions are stored; when the computer instructions are executed on the above-mentioned electronic device, the above-mentioned intelligent scheduling method for coordinated operation of the variable speed multi-pump parallel system is implemented.

[0107] The other parts of this embodiment are the same as any one of the above-mentioned embodiments 1 to 3, and thus will not be described in detail.

[0108] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. An intelligent scheduling method for the coordinated operation of a variable speed multi-pump parallel system, characterized in that: Firstly, numerical methods are used to simulate and calculate the performance characteristics of parallel centrifugal pumps and the pipeline loss characteristics of variable speed multi-pump parallel systems to obtain digital characteristic curves; secondly, based on the digital characteristic curves, the minimum total shaft power is used as the objective function, and according to the set variable speed multi-pump parallel system demand targets and constraints, a mathematical model of intelligent control of the variable speed multi-pump parallel system is established; then, the particle swarm optimization algorithm is improved according to the golden sine algorithm to solve the mathematical model of intelligent control of the variable speed multi-pump parallel system and obtain the operating data; the feedforward artificial neural network is called to establish the agent model to regress and predict the operating data; finally, according to the prediction results of the agent model, the optimal control scheme of the variable speed multi-pump parallel system is obtained.

2. According to claim 1, a variable speed multi-pump parallel system coordinated operation intelligent scheduling method is characterized in that: The intelligent scheduling method for the coordinated operation of a variable speed multi-pump parallel system specifically comprises the following steps: Step S1: using a numerical method to simulate and calculate the variable speed performance characteristics of parallel centrifugal pumps and the pipeline loss characteristics of a variable speed multi-pump parallel system to obtain a digital characteristic curve; Step S2: According to the digital characteristic curve, taking the minimum total shaft power as the objective function, and according to the set variable speed multi-pump parallel system demand target and constraint conditions, a variable speed multi-pump parallel system intelligent control mathematical model is established; Step S3: solving the intelligent control mathematical model of the variable speed multi-pump parallel system according to the golden sine algorithm, the adaptive strategy and the global optimal guided dynamic reverse learning improved particle swarm optimization algorithm to obtain the operation data; Step S4: calling a feedforward artificial neural network to establish a proxy model, regressing and predicting the running data, and obtaining the prediction results; Step S5: According to the prediction results of the agent model, the optimal variable speed multi-pump parallel system coordinated operation scheduling scheme is obtained in combination with the scheduling decision variables.

3. According to claim 2, a variable speed multi-pump parallel system coordinated operation intelligent scheduling method is characterized in that: The step S1 specifically includes the following steps: Step S11: According to the factory characteristic curve of a single centrifugal pump, or by establishing a three-dimensional model of a single centrifugal pump, the characteristic curve of the single centrifugal pump is calculated by a numerical simulation method based on performance prediction; the characteristic curve of the single centrifugal pump is fitted by a polynomial fitting method to obtain the single pump flow Q-head H , Single pump flow Q - Shaft power P and single pump flow Q-efficiency η Mathematical expression of characteristic curve; Step S12: According to the centrifugal pump similarity theory, calculate the characteristic curve of a single centrifugal pump under variable speed conditions, and establish the single pump flow rate under variable speed conditions of a single centrifugal pump Q - Lift H , Single pump flow Q - Shaft power P And single pump flow Q -efficiency η Mathematical expression of characteristic curve; Step S13: Calculate the characteristic curve of the variable speed multi-pump parallel system based on the flow superposition and head invariance characteristics of the parallel centrifugal pumps, and establish the total flow of the variable speed multi-pump parallel system Q t - Total shaft power P t And the total flow Q t - Parallel pump head H t Mathematical expression of characteristic curve; Step S14: construct a pipeline loss characteristic curve according to the pipeline net head, the pipe resistance coefficient of the branch pipe before the parallel node, the pipe resistance coefficient of the main pipeline of the variable speed multi-pump parallel system, the flow rate of the branch pipeline, and the total flow rate of the parallel centrifugal pump.

4. The intelligent scheduling method for coordinated operation of a variable speed multi-pump parallel system according to claim 3 is characterized in that: The step S2 specifically includes the following steps: Step S21: constructing an objective function according to a characteristic curve of a single centrifugal pump, a characteristic curve of a single centrifugal pump under variable speed conditions, a characteristic curve of a variable speed multi-pump parallel system, a pipeline loss characteristic curve, a total shaft power of a variable speed multi-pump parallel system, a shaft power of each centrifugal pump, a state factor of the centrifugal pump, a speed ratio of the centrifugal pump, the number of parallel centrifugal pumps in operation, and a single pump flow rate; Step S22: constructing constraint conditions; the constraining conditions include: constructing a variable speed multi-pump parallel system head constraint according to the pipeline loss characteristic curve, the parallel pump head characteristic curve and the required head, constructing a start-stop number constraint according to the maximum number of running units, constructing a variable speed multi-pump parallel system flow constraint according to the water supply demand flow, constructing a speed ratio constraint according to the rated speed of the centrifugal pump, and constructing a centrifugal pump operating condition constraint according to the rated flow of the centrifugal pump and the set constraint interval; Step S23: Establishing a mathematical model for intelligent control of a variable speed multi-pump parallel system according to the objective function and the constraints.

5. The intelligent scheduling method for coordinated operation of a variable speed multi-pump parallel system according to claim 2 is characterized in that: The step S3 specifically comprises the following steps: Step S31: constructing a position update strategy according to the golden sine algorithm; the position update strategy is used to update the position of the particle; Step S32: constructing an adaptive strategy according to the set weight factors and learning factors; the adaptive strategy is used to adaptively update the weight factors and learning factors; Step S33: According to the first u Individuals in v The reverse solution of the value on the dimension is used to construct a dynamic reverse learning strategy guided by the global optimality; the dynamic reverse learning strategy guided by the global optimality is used to complete the dynamic reverse learning according to the dynamic reverse learning strategy guided by the global optimality after the particle swarm algorithm iteration is completed, and the top N individuals with the best fitness are selected, and the reverse population is merged with the initial population to form a new population; Step S34: improving the particle swarm optimization algorithm according to the position update strategy, the adaptive strategy, and the dynamic reverse learning strategy guided by the global optimum, solving the intelligent control mathematical model of the variable speed multi-pump parallel system according to the improved particle swarm optimization algorithm, and obtaining the operation data.

6. The intelligent scheduling method for coordinated operation of a variable speed multi-pump parallel system according to claim 5 is characterized in that: The step S4 specifically comprises the following steps: Step S41: calling a single hidden layer feedforward artificial neural network to construct a proxy model; Step S42: Adjust the regression coefficient of the proxy model, use the running data as the input layer, and obtain the prediction result through single hidden layer training.

7. The intelligent scheduling method for coordinated operation of a variable speed multi-pump parallel system according to claim 4 is characterized in that: The objective function established in step S21 is: in, P t is the total shaft power of the variable speed multi-pump parallel system; P i For the i The shaft power of the centrifugal pump; W i For the i The condition factor of a centrifugal pump; b 1, b 2, b 3. b 4 is the fitting coefficient of the polynomial fitting of the centrifugal pump flow-power curve; k is the speed ratio of the centrifugal pump; m o The number of centrifugal pumps in operation in the variable speed multi-pump parallel system; Q This is the flow rate of a single pump.

8. An intelligent dispatching system for coordinated operation of a variable speed multi-pump parallel system, used to execute the intelligent dispatching method for coordinated operation of a variable speed multi-pump parallel system as claimed in claim 1; characterized in that: It includes simulation unit, construction unit, solution unit, prediction unit and output unit; The simulation unit is used to simulate and calculate the performance characteristics of the variable speed multi-pump parallel centrifugal pump and the pipeline loss characteristics of the variable speed multi-pump parallel system by numerical methods to obtain a digital characteristic curve; The construction unit is used to determine the required flow rate of a given variable speed multi-pump parallel system based on the digital characteristic curve and taking the minimum total shaft power as the objective function. Q a and required lift H a , and under various constraints, a mathematical model for intelligent control of variable speed multi-pump parallel system is established; The solving unit is used to solve the intelligent control mathematical model of the variable speed multi-pump parallel system according to the improved particle swarm optimization algorithm based on the golden sine algorithm to obtain the operation data; The prediction unit is used to call a feedforward artificial neural network to establish a proxy model and regressively predict the operating data; The output unit is used to obtain the optimal control scheme of the variable speed multi-pump parallel system according to the prediction results of the agent model.

9. An electronic device, characterized in that: It comprises a memory and a processor; a computer program is stored in the memory; when the computer program is executed on the processor, the intelligent scheduling method for coordinated operation of a variable speed multi-pump parallel system as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions; when the computer instructions are executed on the electronic device as described in claim 9, the intelligent scheduling method for coordinated operation of a variable speed multi-pump parallel system as described in any one of claims 1-7 is implemented.

Images