A central air conditioning water system control method based on improved stack optimization algorithm

Abstract

The present invention proposes a central air-conditioning water system control method based on an improved heap optimization algorithm. Aiming at the shortcomings of the HBO algorithm, the method starts from the self-contribution mechanism of the individual position, uses the historical optimal individual to guide the individual at the current position, and speeds up the algorithm's optimization speed in the early stage. At the same time, a segmented nonlinear decreasing strategy is proposed to increase the probability of other update mechanisms being selected in the early stage of iteration, so that the algorithm can achieve a better balance between global exploration and local development; at the same time, a fusion mutation strategy is designed as a direct update mechanism for the optimal individual to enhance the algorithm's ability to jump out of the local optimum. The improved HBO algorithm is used to optimize the various control parameters of the central air-conditioning water system, including the chilled water pump frequency, the chiller outlet water set temperature, the cooling pump frequency, and the cooling tower fan frequency. The operation plan corresponding to the optimal individual is used as the optimal control plan to achieve the minimum total energy consumption of the central air-conditioning water system.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, relates to an intelligent optimization algorithm and operation optimization of a central air-conditioning water system, and specifically relates to a central air-conditioning water system control method based on an improved stack optimization algorithm. Background Art

[0002] The total energy consumption of the central air conditioning water system consists of four parts, including the energy consumption of the chiller, the energy consumption of the refrigeration pump, the energy consumption of the cooling pump and the energy consumption of the cooling tower. The chilled water outlet temperature, the cooling water flow rate and the cooling water inlet temperature jointly determine the operating efficiency of the chiller under the current working conditions; the chilled water temperature difference will affect the energy consumption of the chilled water pump and the chiller at the same time; the cooling water temperature difference will affect the energy consumption of the cooling water pump and the chiller at the same time; the cooling water outlet tower temperature will affect the energy consumption of the cooling tower and the chiller at the same time. Therefore, there are many controllable parameters in the air conditioning cold source system that affect the total energy consumption and efficiency of the system, and the coupling relationship between the parameters is complex, with the characteristics of high complexity and multi-variable coupling. Some classic mathematical methods are difficult to work on this optimization problem.

[0003] In recent years, more and more scholars have begun to use natural heuristic optimization algorithms, an approximate solution method, to deal with such problems. The HBO (Heap-based optimizer) algorithm is a new meta-heuristic optimization algorithm based on human behavior. It uses a heap structure to simulate the hierarchical structure of a company and has been widely used in practical engineering operation optimization problems. However, the traditional HBO algorithm still has problems such as slow initial optimization speed, insufficient global exploration ability, easy to fall into local optimality, and lack of direct update mechanism for optimal individuals. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention proposes a central air-conditioning water system control method based on an improved heap optimization algorithm, which uses the historical optimal individual to guide the individual at the current position to speed up the early optimization speed. At the same time, a segmented nonlinear decreasing strategy is designed to appropriately increase the probability of other update mechanisms being selected in the early stage of iteration, maintain a balance between global exploration and local development, improve the convergence accuracy of the algorithm to a certain extent, and effectively reduce the energy consumption of the central air-conditioning water system.

[0005] A central air conditioning water system control method based on an improved stack optimization algorithm optimizes the energy consumption of the central air conditioning water system by generating control schemes for the frequency of each chilled water pump, the outlet water set temperature of each chiller unit, the frequency of each cooling pump, and the frequency of each cooling tower fan. The method specifically includes the following steps:

[0006] Step 1: According to the actual number of equipment in the project and the specific performance parameters, a central air-conditioning water system model is established, constraints are set, and the objective function is established with the minimum total energy consumption of the system.

[0007] Step 2: Use the improved HBO algorithm to optimize the control parameters of each module in the system. The specific steps are as follows:

[0008] Step 2.1, set the relevant parameters of the improved HBO algorithm, including the number of populations and the maximum number of iterations, and then randomly generate the initial population in the search space. Each individual in the population corresponds to a set of system control parameters, and the population is arranged in a heap according to the fitness value of the individuals in the initial population.

[0009] Add the corresponding penalty function to the objective function established in step 1 to obtain the fitness function :

[0010]

[0011] in, and Indicates the chilled water outlet temperature and inlet temperature. is the upper limit of the chilled water inlet temperature, and Indicates the cooling water outlet temperature and inlet temperature on the cooling tower side. and are the upper and lower limits of the cooling water outlet temperature. is the upper limit of the cooling water inlet temperature, and is the inlet and outlet temperature difference of chilled water and cooling water, represents the cooling load supplied, represents the cooling load demand; Both are penalty factors; It is a judgment function. When the constraint condition is not met, the function value is 1, and when it is met, it is 0.

[0012] Step 2.2: According to the random probability p∈(0,1), select the corresponding mechanism to update the individual position. In order to better balance the global exploration ability and local development ability of the algorithm, a piecewise nonlinear decreasing strategy is defined to attenuate p1:

[0013]

[0014] Among them, t represents the current iteration number, represents the maximum number of iterations, and a represents the number of iterations corresponding to the intersection point of the two segments of the function. This strategy uses a non-linear function with a faster descent rate in the early stage of iteration to appropriately increase the possibility of different mechanisms being selected, thereby enhancing the global exploration ability of the algorithm in the early stage of iteration. In the middle and late stages of iteration, it switches to another non-linear function with a relatively gentle descent rate to ensure that p1 does not approach 0 too early, thus maintaining the local development ability of the algorithm;

[0015]

[0016]

[0017] ① When p < p1, an improved self - contribution mechanism is used for position update. This improved method introduces the historical best individual on the basis of the original self - contribution mechanism to guide the update direction of the current individual. Its specific implementation is as follows:

[0018]

[0019] where represents the coordinate of the historical best individual corresponding to the position of individual i in the hierarchical structure in the j - th dimension, is the position of individual i in the j - th dimension at the t - th iteration, , and t represents the number of iterations.

[0020] ② When p1 < p < p2, a mechanism of interacting with the direct leader is used for position update:

[0021]

[0022] where, represents the position of the direct leader of the current individual in the j - th dimension, is a random number taking values between - 1 and 1.

[0023] ③ When p2 < p < p3, a mechanism of interacting with colleagues is used for position update:

[0024]

[0025] where, represents the position of the randomly selected colleague in the - th dimension; and respectively represent the fitness value of the - th individual and the fitness value of the colleague individual in the - th iteration process;

[0026] Step 2.3: Limit the boundaries of the updated individuals. When the position of a certain dimension exceeds the upper or lower boundary, use the boundary threshold to replace it. Then compare the changes in the individual fitness values ​​before and after the update, retain the individual position with a better fitness value, and adjust the position of the individual in the heap according to the size of the fitness value.

[0027] Step 2.4: After all individuals except the best individual have completed the update, a fusion mutation strategy is used to update the position of the best individual according to the dimension. The specific implementation is as follows:

[0028]

[0029] in To satisfy the mutation operator of Gaussian distribution, it is expected that =0, variance =1, is a mutation operator satisfying the t-distribution with n=t degrees of freedom, It represents the position of the optimal individual in the jth dimension after t iterations, Represents the position of the optimal individual after mutation in the jth dimension.

[0030] Step 2.5 selects and retains the best individuals before and after the mutation.

[0031] Step 2.6: Repeat steps 2.2 to 2.5. When the upper limit of the maximum number of iterations is reached, output the position of the current optimal individual.

[0032] Step 3: Use the optimal individual obtained in step 2 as the optimal control solution to control the module parameters in the central air-conditioning water system and optimize the system energy consumption.

[0033] The present invention has the following beneficial effects:

[0034] In view of some shortcomings of the HBO algorithm itself, this method introduces the historical optimal individual to speed up the optimization speed of the algorithm in the early stage of iteration, defines a piecewise nonlinear decreasing strategy to further balance the global exploration ability and local development ability of the algorithm, and adopts a fusion mutation strategy to make up for the lack of direct update mechanism of the optimal individual in the algorithm, improves the ability of the algorithm to jump out of the local optimum, and improves the convergence accuracy of the algorithm to a certain extent; finally, minimizing the total energy consumption of the system is taken as the goal, and the improved HBO algorithm is used to find a set of operation plans that can both meet the constraints and make the total energy consumption of the system smaller, so as to achieve energy consumption optimization of the central air-conditioning water system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flow chart of a central air conditioning water system control method based on an improved stack optimization algorithm;

[0036] Figure 2 A flowchart of the improved heap optimization algorithm for this method;

[0037] Figure 3 It is the convergence curve of different optimization algorithms on the single-peak test function Sphere in the embodiment;

[0038] Figure 4 It is the convergence curve of different optimization algorithms on the multi-peak test function Rastrigin in the embodiment; DETAILED DESCRIPTION

[0039] The present invention will be further explained below with reference to the accompanying drawings;

[0040] like Figure 1 As shown, a central air conditioning water system control method based on an improved stack optimization algorithm specifically includes the following steps:

[0041] Step 1: According to the actual number of equipment and specific performance parameters in the project, a central air-conditioning water system model is established on the TRNSYS simulation platform. The corresponding modules are selected and relevant specific parameters are set. Each module is correctly connected and its energy consumption is connected to the output module respectively. The following objective function is established with the minimum total energy consumption of the central air-conditioning water system:

[0042]

[0043] In the formula, is the total energy consumption of the central air conditioning water system, It is Energy consumption of a variable frequency chilled water pump, It is Energy consumption of each chiller, is the number of chilled water pumps, is the number of chillers; It is Energy consumption of a cooling tower, The Energy consumption of a variable frequency cooling water pump, is the number of cooling towers, is the number of cooling water pumps;

[0044]

[0045] in, is the frequency of the variable frequency water pump, is the coefficient of each term obtained after polynomial fitting of the energy consumption of the qth variable frequency water pump, is the energy consumption of the qth variable frequency water pump, is the rated energy consumption of the qth variable frequency water pump, , The energy consumption model of the variable frequency water pump mentioned above is used;

[0046]

[0047] in, Representative The capacity of the chiller under the current operating conditions, Representative The calibrated performance coefficient of each chiller, Representative The power factor of a chiller when fully loaded at a certain part load rate;

[0048]

[0049] in, For the Rated power of cooling tower fans, For the The actual air volume of the cooling tower fan, is the rated air volume of the fan, is the polynomial fitting coefficient of the fan energy consumption model, k represents the highest power of the polynomial, and r represents the power index of the polynomial;

[0050] The following constraints are designed for the objective function:

[0051]

[0052] in, , , Respectively represent the frequencies of the chilled water pump, cooling water pump and cooling tower fan, and Indicates the chilled water outlet temperature and inlet temperature. and Indicates the cooling water outlet temperature and inlet temperature on the cooling tower side. represents the cooling load supplied, Represents the cooling load demand.

[0053] Step 2: In view of some defects of the HBO algorithm itself, it is improved based on two criteria: (1) Strengthen the global exploration ability of the algorithm in the early stage, thereby accelerating the optimization speed of the algorithm in the early stage of iteration; (2) Introduce a direct update mechanism for the optimal individual position, thereby improving the convergence accuracy of the algorithm and enhancing the ability of the algorithm to escape from the local optimum. Use the improved HBO algorithm to solve the objective function:

[0054] Step 2.1, set the relevant parameters of the improved HBO algorithm, including the number of populations and the maximum number of iterations, and then randomly generate the initial population in the search space. Each individual in the population corresponds to a set of system control parameters. Add a penalty function to the objective function established in step 1 to establish the following fitness function :

[0055]

[0056] Arrange the population into a heap according to the fitness values ​​of the individuals in the initial population.

[0057] Step 2.2: Select the corresponding mechanism to update the individual position according to the random probability p∈(0,1).

[0058] In the HBO algorithm, p1=1-(t / t Max ), is a linear decay strategy, which results in the consistent decay speed of p1 during the entire iteration process. When the value of p1 is large, the algorithm is likely to choose the self-contribution mechanism. When p1 is small, the algorithm is likely to choose the other two position update mechanisms to update the individual position. Therefore, the speed of p1 decay affects the selection of the main update mechanism of the algorithm. The relatively single update mechanism in the early stage of iteration is not conducive to the algorithm to perform global exploration in the search space. For this reason, the present invention uses To decay p1, where t represents the current number of iterations, Represents the maximum number of iterations. This function increases the possibility of the algorithm selecting different update mechanisms by accelerating the decay rate of p1 in the early stage of iteration, thereby enhancing the global exploration ability of the algorithm in the early stage of iteration. However, if the value of p1 approaches 0 too early, the local development ability of the algorithm will be affected to a certain extent. For this reason, the present invention adopts a smooth nonlinear function with a more uniform decline rate in the middle and late stages of iteration. This function is calculated according to the formula To decay p1 in the middle and late stages, the overall piecewise nonlinear decreasing strategy is:

[0059]

[0060] Among them, a represents the number of iterations corresponding to the intersection of the two functions.

[0061]

[0062]

[0063] ① Since the HBO algorithm is likely to choose the self-contribution mechanism to update the position in the early stage of iteration, the self-contribution mechanism is based on To retain the position of the j-th dimension of individual i, to some extent, it leads to insufficient global exploration ability of the algorithm in the early stage. Therefore, when p < p1, the historical best individual is introduced into the self-contribution mechanism for position update:

[0064]

[0065] where represents the coordinate of the historical best individual at the corresponding position of individual i in the hierarchical structure in the j-th dimension, is the position of individual i in the j-th dimension at the t-th iteration, , and t represents the number of iterations. The improved individual contribution strategy guides the update direction of the current individual by leveraging the historical best individual at the corresponding position in the heap.

[0066] ② When p1 < p < p2, a mechanism of interacting with the direct leader is adopted for position update:

[0067]

[0068] where, represents the position of the direct leader of the current individual in the j-th dimension, is a random number taking values between -1 and 1.

[0069] ③ When p2 < p < p3, a mechanism of interacting with colleagues is adopted for position update:

[0070]

[0071] where, represents the position of the randomly selected colleague individual in the -th dimension; and respectively represent the fitness values of the -th individual and the colleague individual during the -th iteration process;

[0072] Step 2.3. Perform boundary restriction on the updated individual. When the position of a certain dimension exceeds the upper and lower boundaries, use the boundary threshold for replacement. Since the mutated individual is not necessarily better than the original individual, the individuals before and after mutation are selected by preference. By comparing the changes in the fitness values of the individuals before and after update, retain the individual with a better fitness value, and adjust the positions of the individuals in the heap according to the magnitude of the fitness values.

[0073] Step 2.4. When all individuals except the best individual have completed the update, perform position update on the best individual. To enhance the ability of the algorithm to jump out of the local optimum, a mutation strategy that combines Gaussian distribution and t-distribution is proposed to perform position update on the best individual according to the dimension:

[0074]

[0075] in To satisfy the mutation operator of Gaussian distribution, it is expected that =0, variance =1, is a mutation operator satisfying the t-distribution with n=t degrees of freedom, It represents the position of the optimal individual in the jth dimension after t iterations, Represents the position of the optimal individual after mutation in the jth dimension.

[0076] The two mutation operators dynamically adjust their respective proportions in the iteration process through trigonometric functions. Since the t distribution with the degree of freedom being the current number of iterations is more dispersed than the Gaussian distribution, it is more suitable for the global exploration of the algorithm in the early stage, while the more concentrated Gaussian distribution is more suitable for appropriate perturbations of the algorithm in the later stage, improving the algorithm's ability to jump out of the local optimum and improving the convergence accuracy of the algorithm to a certain extent.

[0077] Step 2.5 selects and retains the best individuals before and after the mutation.

[0078] Step 2.6: Repeat steps 2.2 to 2.5. When the upper limit of the maximum number of iterations is reached, output the position of the current optimal individual, such as Figure 2 shown.

[0079] In order to verify the effectiveness of the improved HBO algorithm of this method, four classic unimodal test functions and four classic multimodal test functions were selected for performance verification. The unimodal test functions include Sphere function (F1), Rosenbrock function (F2), Schwefel's 2.21 function (F3) and Schwefel's 2.22 function (F4), and the multimodal test functions include Rastrigin function (F5), Quartic function (F6), Griewangk function (F7) and Ackley function (F8). The heap optimization algorithm and three other typical optimization algorithms were selected for comparison. The typical optimization algorithms include differential evolution algorithm, particle swarm algorithm and gravitational search algorithm. In the test function comparison experiment, each algorithm sets the population size to 50 and the maximum number of iterations to 1000. In order to avoid contingency, each algorithm runs independently 30 times. Figure 3 and Figure 4 These are the convergence curves of the improved heap optimization algorithm and other comparison algorithms on the test functions F1 and F5. Figure 3 and Figure 4It can be seen that compared with the traditional heap optimization algorithm, the improved heap optimization algorithm proposed by this method has a faster optimization speed in the early stage of iteration, and also achieves higher convergence accuracy on the single-peak test function F1. On the multi-peak test function F5, it can jump out of the local optimal value multiple times and find the global optimal value.

[0080] Step 3: Set the cooling load demand in the TRNSYS simulation model, and then run the simulation system to generate a dck file containing the parameter information and connection relationship of each module in the simulation model. Run the above algorithm through the MATLAB program, rewrite the value of the optimization variable in the dck file, so as to provide a control signal for the simulation model in TRNSYS, and call TrnEXE64.exe. With the help of the simulation model in TRNSYS, calculate the energy consumption of each module of the system under this set of operating variables and other simulation results of the system, and record and output them through a txt file. The objective function value under this set of operating variables is obtained by reading and processing the txt file generated in the previous step.

[0081] In order to verify the effectiveness of this method for optimizing the operation of central air-conditioning water system, a central air-conditioning water system with a rated cooling capacity of 4000kW was built using the TRNSYS simulation platform, and variable frequency water pumps and variable frequency fans of appropriate specifications were configured for it. The parameters of the main equipment in the simulation system are shown in Table 1.

[0082] Table 1

[0083] Device Type Device Name Specific parameters Chiller Chiller1, Chiller2Chiller3, Chiller4 Rated cooling capacity 1000kW, rated COP 4.05 Variable frequency chilled water pump Chilled water pump 1, Chilled water pump 2 Chilled water pump 3, Chilled water pump 4 Rated power 28kW, rated flow 184m³ / h Frequency conversion cooling water pump Cooling water pump 1, cooling water pump 2 Rated power 48kW, rated flow 478.4m³ / h Cooling Tower Cooling tower 1, cooling tower 2 Rated power 22kW, rated flow 385000m³ / h

[0084] The same type of equipment used in this simulation all adopt the same specifications and sizes, and the performance of the same type of equipment is assumed to be consistent. The performance curve of the chiller and the power coefficient relative to full load at partial load rate all use the default parameters in the TYPE666 module. These parameters can be modified in combination with the actual project. In addition, the fitting coefficients in the energy consumption formula corresponding to the chilled water pump and the cooling water pump are shown in Table 2, and the fitting coefficients in the energy consumption formula corresponding to the cooling tower fan are shown in Table 3:

[0085] Table 2

[0086] Device Type <![CDATA[a0]]> <![CDATA[a1]]> <![CDATA[a2]]> <![CDATA[a3]]> Chilled water pump 0.476 -0.140 1.419 -0.755 Cooling water pump 0.458 0.633 -0.044 -0.047

[0087] Table 3

[0088] Device Type <![CDATA[b0]]> <![CDATA[b1]]> <![CDATA[b2]]> <![CDATA[b3]]> Cooling tower fan 0.131 0.459 0.648 0.680

[0089] Under different cooling load demands, this method is used to optimize the operating variables in the simulation model of the central air-conditioning water system. The frequencies of each chilled water pump, cooling water pump and cooling tower fan and the chilled water outlet set temperature of each chiller are selected as optimization variables. The population size is set to 20 and the maximum number of iterations is 40. The optimization algorithms are compared and analyzed with several other commonly used optimization algorithms in this operation optimization direction, including particle swarm optimization (PSO), differential evolution algorithm (DE), gravitational search algorithm (GSA) and original HBO algorithm. In addition, the optimization algorithm is compared with the traditional equal load distribution strategy. In order to avoid contingency, each optimization algorithm is run independently 5 times and the best result is recorded. The results are shown in Table 4:

[0090] Table 4

[0091] ;

[0092] (Continued Table 4)

[0093] ;

[0094] From the results in Table 4, it can be seen that the control strategies of each intelligent optimization algorithm for the operation optimization of the central air-conditioning water system have a considerable energy-saving effect compared with the traditional control strategy. Moreover, among the various comparison algorithms, this method, namely the improved heap optimization (IHBO) algorithm, can achieve the best optimization results in most cases. Compared with the traditional control strategy, the IHBO algorithm can save 11.65-62.89kW of energy consumption under different cooling load requirements, and the relative energy saving rate is 3%-5.72%. Compared with the DE algorithm, the IHBO algorithm can save 0.05-11.49kW of energy consumption under different cooling load requirements, and the relative energy saving rate is 0.01%-1.19%. Compared with the GSA algorithm, the IHBO algorithm can save 0.02-18.62kW of energy consumption under different cooling load requirements, and the relative energy saving rate can reach up to 2.11%. Compared with the PSO algorithm and the original HBO algorithm, the IHBO algorithm also shows higher optimization accuracy.

[0095] In summary, this method can find an operating scheme with lower energy consumption in most cases, and the operating schemes it finds can meet the cooling load demand and other constraints. It can reduce system energy consumption to a greater extent while ensuring system stability, and has high energy-saving potential and promotion value.

Claims

1. A control method for the central air-conditioning water system based on an improved heap optimization algorithm, characterized in that: Step 1: Establish a central air-conditioning water system model, with the minimum total energy consumption of the system as the objective function; Step 2: Design a calculation method for segmented probability: p3＝1 Among them, t represents the current iteration number, t Max represents the maximum number of iterations, and a represents the number of iterations corresponding to the intersection of two functions; Construct a fitness function, and use the improved HBO algorithm to optimize the control parameters of each module in the system. The control parameters include the frequencies of each chilled water pump, cooling water pump, and cooling tower fan in the central air-conditioning water system, and the chilled water outlet set temperature of each chiller. The specific optimization steps are as follows: Step 2.1: Set the relevant parameters of the improved HBO algorithm, including the population size and the maximum number of iterations, and then randomly generate an initial population in the search space. Each individual in the population corresponds to a set of system control parameters; arrange the population into a heap according to the fitness values of the individuals in the initial population; Step 2.2: Randomly generate a probability value p, and select the corresponding mechanism to update the individual position according to the random probability p ∈ (0, 1): ① When p < p1, use the improved self-contribution mechanism to update the position; in represents the coordinates of the historical best individual at the corresponding position of individual i in the hierarchical structure in the jth dimension, is the position of individual i in the jth dimension at the tth iteration, γ=|2-4*mod(t,25) / 25|, t represents the current iteration number; ② When p1 < p < p2, use the interaction mechanism with the direct leader to update the position; ③ When p2 < p < p3, use the interaction mechanism with colleagues to update the position; Step 2.3: Perform boundary restriction on the updated individuals. When the position of a certain dimension exceeds the upper and lower boundaries, use the boundary threshold for replacement, and then compare the change in the fitness values of the individuals before and after the update, retain the position of the individual with a better fitness value, and adjust the position of the individuals in the heap according to the fitness values; Step 2.4: When all individuals except the optimal individual have completed the update, perform a fusion mutation operation on the optimal individual, and then select the best individuals before and after the mutation; the position update method for the optimal individual is: Among them, t represents the current iteration number, t Max represents the maximum number of iterations, Gauss(μ,σ 2 ) is a mutation operator that satisfies the Gaussian distribution, with the expected μ = 0 and the variance σ 2 =1, trnd(n) is the mutation operator of the t distribution with the degree of freedom n=t, It represents the position of the optimal individual in the jth dimension after t iterations, Indicates the position of the optimal individual in the jth dimension after mutation; Step 2.5: Repeat steps 2.2 to 2.

4. When the upper limit of the maximum number of iterations is reached, output the position of the current optimal individual; Step 3: Use the optimal individual obtained in step 2 as the optimal control scheme to control the module parameters in the central air-conditioning water system and optimize the system energy consumption.

2. A central air conditioning water system control method based on an improved stack optimization algorithm as claimed in claim 1, characterized in that: According to the actual equipment quantity and specific performance parameters in the project, establish a central air-conditioning water system model on the TRNSYS simulation platform, and establish the following objective function with the minimum total energy consumption of the central air-conditioning water system: Among them, P sys is the total energy consumption of the central air conditioning water system, is the energy consumption of the n1th variable frequency chilled water pump, is the energy consumption of the n2th chiller, N1 is the number of chilled water pumps, and N2 is the number of chillers; is the energy consumption of the n3th cooling tower, is the energy consumption of the n4th variable frequency cooling water pump, N3 is the number of cooling towers, and N4 is the number of cooling water pumps; Among them, f is the frequency of the variable frequency water pump, a 0,q 、a 1,q 、a 2,q 、a 3,q is the coefficient of each item obtained after polynomial fitting of the energy consumption of the qth variable frequency water pump, P pump,q is the energy consumption of the qth variable frequency water pump, P pump,q,0 Indicates the rated energy consumption of the qth variable frequency water pump, The energy consumption model of the variable frequency water pump mentioned above is used in both cases; in, represents the capacity of the n2th chiller under the current operating conditions, represents the calibrated performance coefficient of the n2th chiller, represents the power factor of the n2th chiller when fully loaded at a certain part load rate; in, is the rated power of the n3rd cooling tower fan, is the actual air volume of the n3th cooling tower fan, is the rated air volume of the fan, is the polynomial fitting coefficient of the fan energy consumption model, k represents the highest power of the polynomial, and r represents the power index of the polynomial; Design the following constraint conditions for the objective function: In the formula, Respectively represent the frequency of the chilled water pump, cooling water pump and cooling tower fan, m1, m3, m5 represent the lower limit of each frequency value, m2, m4, m6 represent the upper limit of each frequency value, T chw,out and T chw,in Indicates the chilled water outlet temperature and inlet temperature. m7 and m8 are the upper and lower limits of the chilled water outlet temperature. m9 is the upper limit of the chilled water inlet temperature. T cw,out and T cw,in Indicates the cooling water outlet temperature and inlet temperature on the cooling tower side, m 10 and m 11 is the upper and lower limits of the cooling water outlet temperature, m 12 is the upper limit of the cooling water inlet temperature, m 13 and m 14 is the inlet and outlet temperature difference of chilled water and cooling water, CL s Indicates the cooling load supplied, CL n Represents the cooling load demand.

3. A central air conditioning water system control method based on an improved stack optimization algorithm as claimed in claim 2, characterized in that: In the objective function min P sys On the basis of , the corresponding penalty function is added as the fitness function f(·) of the improved HBO algorithm: f(·)=P sys +ω1φ(T chw,in <m9)+ω2φ(m 10 <T cw,out <m 11 )+ω3φ(T cw,in <m 12 )+ω4φ(T chw,in -T chw,out <m 13 )+ω5φ(T cw,in -T cw,out <m 14 )+ω6φ(CL s =CL n ) Among them, ω1 to ω6 are all penalty factors; φ( ) is a judgment function, and the function value is 1 when the constraint condition is not satisfied and 0 when it is satisfied.

4. A central air conditioning water system control method based on an improved stack optimization algorithm as claimed in claim 1, characterized in that: The position update method for interacting with the direct leader is: Among them, B j represents the position of the current individual’s direct leader in the jth dimension, λ j A random number between -1 and 1.

5. A central air conditioning water system control method based on an improved stack optimization algorithm as claimed in claim 1, characterized in that: The position update method for interacting with colleagues is: in, represents the position of a randomly selected colleague in the jth dimension; f(X i (t)) and f(S r ) represent the fitness value of the i-th individual and the fitness value of the colleague individual in the t-th iteration process respectively.

Images