Optimization method of balancing valve in heating system based on improved Jaya algorithm

Abstract

The present invention relates to the technical field of optimization and control of urban heating systems, and in particular to a method for optimizing a balancing valve of a heating system based on an improved Jaya algorithm. The method comprises step 1, initializing algorithm parameters; step 2, initializing a population; step 3, using an adaptive search balance strategy, dynamically selecting to execute a benefit-seeking and harm-avoiding strategy and a Q-learning-guided local search to update the solution; step 4, updating the optimal solution in the current population, and judging whether the algorithm termination condition is reached. If so, the algorithm evolution ends, and the global optimal solution found and its corresponding heating energy consumption are output, otherwise step 3 is executed. The present invention solves the problem of optimizing the opening of a balancing valve in a secondary pipe network heating system, and has the positive effects of reducing energy consumption, reducing operating costs, and improving energy efficiency.

Description

Technical Field

[0001] The invention relates to the technical field of urban heating system optimization control, and in particular to a heating system balancing valve optimization method based on an improved Jaya algorithm. Background Art

[0002] With the growth of global energy demand and the advancement of energy conservation and emission reduction policies, optimizing energy consumption has become a focus of international attention. As an important part of urban infrastructure, the energy efficiency optimization of urban heating systems is crucial to achieving sustainable development goals.

[0003] In urban heating systems, the efficiency of the secondary water supply network is affected by many factors, such as pipe network design, operation mode and external climatic conditions. Studies have shown that up to 15% to 30% of energy losses in heating systems are caused by hydraulic imbalance. The typical problem is that users close to the heat exchange station in the system experience higher room temperatures, while remote users are often affected by low temperatures. In order to improve the room temperature of remote users, heating companies usually need to increase the power of the heat exchange station circulation pump to increase the flow of hot water to remote users, but this also causes overheating and energy waste for nearby users. In the past, these problems were usually solved by manually adjusting the balancing valve. However, this method not only relies on the operator's experience, but also becomes increasingly difficult to manage as the heating network becomes more complex, and improper adjustment can also cause excessive energy consumption of the circulation pump.

[0004] Therefore, there is an urgent need to develop a new technical solution to optimize the setting of the balancing valve and maximize the energy efficiency of the heating system. Summary of the invention

[0005] The purpose of the present invention is to propose a heating system balancing valve optimization method based on an improved Jaya algorithm, which solves the problem of heat waste in the heating system and the problem of excessive power consumption of the circulating pump, so as to achieve the purpose of reducing energy consumption and improving heating efficiency.

[0006] The present invention provides a heating system balancing valve optimization method based on an improved Jaya algorithm, which is characterized by comprising the following steps:

[0007] Step 1: Initialize the algorithm parameters, set the population size PS, and the maximum number of iterations during the algorithm evolution process And the threshold of the number of times each individual in the population is allowed to evolve , where the number of solutions contained in the population corresponds to the number of solutions for the configuration of balancing valves in the secondary pipe network heating system; the threshold It represents the upper limit of the number of times that the balancing valve configuration solution fails to reduce the heating energy consumption during the iteration process;

[0008] Step 2, initialize the population, encode it by using the representation method based on the balancing valve opening sequence, mix the random strategy and the linear interval strategy to generate the first solution of the population, and use the scrambling rule to generate the remaining PS-1 solutions of the population, with a total of PS solutions in the population;

[0009] Step 3: Using the adaptive search balance strategy, dynamically choose to execute the strategy of seeking benefits and avoiding harm, and the local search guided by Q-learning to update the solution;

[0010] Step 4: Update the optimal solution in the current population and determine whether the maximum number of iterations has been reached. If so, the evolution ends and the optimal solution and the minimum heating energy consumption corresponding to the optimal solution are output. Otherwise, execute step 3.

[0011] Further, in step 2, the encoding steps are performed based on the representation method of the balance valve opening sequence as follows, each solution consists of a one-dimensional vector Represents a one-dimensional vector The subscript corresponds to the terminal user's ID, and the subscript corresponds to The elements in represent the balance valve opening of the end user, so one solution is used to represent the balance valve opening configuration of all end users. Therefore, a solution can be expressed as , Indicates that the ID is The balancing valve opening for heating the end user.

[0012] Furthermore, in step 2, the random strategy is to randomly select The value between is taken as the balance valve opening. Indicates the minimum opening of the balancing valve. Indicates the maximum opening of the balancing valve; the linear interval strategy is expressed by the formula To generate the balance valve opening, where is the newly generated solution The opening of a balancing valve, Indicates the number of balancing valves supplying heat to end users. Center front The balance valve opening is generated by random strategy. The opening of the balancing valve is generated by alternating between the random strategy and the linear interval strategy, thereby mixing the random strategy and the linear interval strategy to generate the first solution of the population.

[0013] Furthermore, in step 2, the scrambling rules include exchange scrambling, reverse scrambling, and insertion scrambling, wherein:

[0014] The operation process of exchange scrambling is to randomly select the openings of two balancing valves in the solution and exchange them;

[0015] The reverse scrambling operation process is to randomly select an index point and reverse the balancing valve opening sequence from the index point to the end of the solution;

[0016] The operation process of inserting scrambling is to randomly insert a new balance valve opening value into any position in the solution and remove the last balance valve opening value in the solution.

[0017] Furthermore, in step 3, an adaptive search balance strategy is used to dynamically update the solution, specifically including:

[0018] In the early stage, the solution is updated through the strategy of seeking benefits and avoiding harm;

[0019] In the mid-term, the current solution is adjusted by combining the profit-seeking and harm-avoiding strategy and the local search guided by Q-learning. The profit-seeking and harm-avoiding strategy and the local search guided by Q-learning are selected according to the strategy ratio. Mixed use, By formula Generate, where is the current iteration number of the algorithm, is the total number of iterations of the algorithm, is an adjustment factor used to increase the time when the algorithm falls into a local optimum If the value of Then the strategy of seeking benefits and avoiding harm is implemented, otherwise the local search guided by Q-learning is performed. for In this stage, if the current solution is in the continuous If there is no performance improvement in the iteration, the current solution is considered to have fallen into a local optimum, and the algorithm will automatically increase , thereby increasing the frequency of using the strategy of seeking benefits and avoiding harm to jump out of the local optimal state;

[0020] In the later stage, Q-learning-guided local search is used to optimize the solution.

[0021] Furthermore, in step 3, the strategy of seeking benefits and avoiding harm includes adjusting the opening of each balancing valve, determining the current optimal solution and the worst solution from the population, and updating the current solution by reducing the similarity with the worst solution while increasing the closeness with the optimal solution. The updating method is ,in, and for A random number uniformly distributed between is the newly generated solution Balance valve opening, The current solution The opening of a balancing valve, and is the best solution and the worst solution in the population. The opening of the balancing valve is updated and a new solution is generated after the update is completed. , if the newly generated solution The energy consumption is less than the current solution , then use the new solution Replace the current solution.

[0022] Furthermore, the elements used by Q-learning to guide local search include state, action, reward and Q value, among which the state is the opening configuration of the balancing valve in the current heating system; the action is the three neighborhood operators to be selected; the reward is the energy-saving effect based on the selected action, that is, the degree of improvement in system energy efficiency after taking the action; the Q value is used to evaluate the heating energy-saving effect after taking the action under the current balancing valve opening configuration, and guide decision-making to achieve the optimal valve adjustment strategy.

[0023] Furthermore, the local search guided by Q-learning includes three neighborhood operators, namely, segmentation operator, mutation operator and crossover operator, among which,

[0024] The specific operation of the segment operator is to set the segment length and the starting subscript for the current solution, and start from the starting subscript to adjust the opening of each balancing valve in the segment according to the adjustment formula. Make adjustments, including After adjustment Balance valve opening, The current solution Balance valve opening, for A random number between The adjustment step is used to control the adjustment range of the balancing valve. After the balancing valve in the segment is adjusted, a new solution is generated. ,like If the heating energy consumption is lower, the current solution in the population is replaced;

[0025] The specific operation of the mutation operator is to determine the number b of balancing valves whose openings are to be mutated for the current solution, and randomly select the corresponding number of balancing valves, where the range of the number b of balancing valves is greater than 0 and not greater than 1 / 2 of the total number of balancing valves. Adjust b balancing valves, Used to generate Random balancing valve opening between After the mutation The opening of a balancing valve, after the adjustment of b balancing valves, a new solution is generated. ,like The heating energy consumption is lower. Replace the current solution in the population;

[0026] The specific operation of the crossover operator is to randomly select another solution in the population as the reference solution to cross with the current solution, and determine the number of balancing valves c to be exchanged, and the number of balancing valves c to be exchanged does not exceed 1 / 5 of the total number of balancing valves. Then, randomly select a crossover subscript, and exchange the balancing valve opening at the crossover subscript of the current solution with the balancing valve opening at the crossover subscript of the reference solution. After the exchange of the c balancing valves to be exchanged is completed, a new solution is generated. ,like The heating energy consumption is lower. Replace the current solution in the population.

[0027] Furthermore, in step 4, the maximum number of iterations of the algorithm is 1000 iterations.

[0028] The present invention provides a heating system balancing valve optimization method based on the improved Jaya algorithm. Compared with the prior art, the present invention solves the balancing valve optimization problem in the secondary pipe network environment of the heating system and has the following positive effects:

[0029] (1) The present invention designs three scrambling rules to enhance the diversity of solutions and achieve a comprehensive exploration of the solution space;

[0030] (2) The present invention overcomes the shortcoming of the traditional Jaya algorithm that it is easy to fall into the local optimum. The improved Jaya algorithm is more likely to jump out of the local optimum and obtain a higher quality solution;

[0031] (3) The adaptive search balance strategy designed in the present invention optimizes the exploration depth and breadth of the algorithm by dynamically adjusting the execution frequency of the solution update method, thereby improving the quality of the solution and the convergence speed of the algorithm;

[0032] (4) In the present invention, Q-learning is introduced to guide local search, so that the algorithm can optimize the exploration of the solution space based on past experience and effectively avoid falling into the local optimum;

[0033] (5) The segmented operator designed for this problem effectively reduces the computational complexity of the algorithm and significantly improves the computational efficiency. In summary, the present invention can well solve the problem of optimizing the opening of the balancing valve in the secondary pipe network of the heating system. This method can not only adjust the opening of the balancing valve to match the actual heating demand, but also significantly reduce the energy consumption caused by hydraulic imbalance, which has the positive effect of improving the energy utilization efficiency and operation reliability of the entire heating system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flow chart for realizing the present invention;

[0035] Figure 2 A schematic diagram of secondary pipe network heating of a heating system according to the method of the present invention;

[0036] Figure 3 This is a diagram for verifying the effectiveness of the Q-learning technology of the method of the present invention;

[0037] Figure 4 This is a diagram for verifying the effectiveness of the segmented operator technology of the method of the present invention;

[0038] Figure 5 This is a diagram for verifying the effectiveness of the adaptive search balance strategy technology of the method of the present invention;

[0039] Figure 6 The average and 95% confidence graph of the present invention and the comparison algorithm;

[0040] Figure 7 It is the minimum energy consumption value convergence diagram of the present invention. DETAILED DESCRIPTION

[0041] like Figure 1 As shown, the heating system balancing valve optimization method based on the improved Jaya algorithm provided by the present invention is mainly implemented through the following steps.

[0042] Step 1: Initialize algorithm parameters. Set the population size PS and the maximum number of iterations during the algorithm evolution process. And the threshold of the number of times each individual in the population is allowed to evolve , where the number of solutions contained in the population corresponds to the number of solutions for the configuration of balancing valves in the secondary pipe network heating system; the threshold It represents the upper limit of the number of times during the iteration that the balancing valve configuration solution fails to reduce the heating energy consumption.

[0043] Step 2: Initialize the population, encode it using the representation method based on the balancing valve opening sequence, adopt a mixed random and linear interval strategy to generate an initial solution, and use three scrambling rules to generate the remaining PS-1 solutions, with a total of PS solutions in the population.

[0044] Specifically, the encoding steps based on the representation method of the balance valve opening sequence are as follows: each solution consists of a one-dimensional vector Represents a one-dimensional vector The subscript corresponds to the terminal user's ID, and the subscript corresponds to The elements in represent the balance valve opening of the end user, so one solution is used to represent the balance valve opening configuration of all end users. Therefore, a solution can be expressed as , Indicates that the ID is The balancing valve opening for heating the end user.

[0045] In the process of generating solutions, the random strategy is to randomly select The value between is taken as the balance valve opening. Indicates the minimum opening of the balancing valve. Indicates the maximum opening of the balancing valve; the linear interval strategy is expressed by the formula To generate the balance valve opening, where is the newly generated solution The opening of a balancing valve, Indicates the number of balancing valves supplying heat to end users. Center front The balance valve opening is generated by random strategy. The balance valve opening is generated by alternating random strategy and linear interval strategy, so as to generate the first solution of the population by mixing random strategy and linear interval strategy. The scrambling rules include exchange scrambling, reverse scrambling and insertion scrambling, among which,

[0046] The operation process of exchange scrambling is to randomly select the openings of two balancing valves in the solution and exchange them;

[0047] The reverse scrambling operation process is to randomly select an index point and reverse the balancing valve opening sequence from the index point to the end of the solution;

[0048] The operation process of inserting scrambling is to randomly insert a new balance valve opening value into any position in the solution and remove the last balance valve opening value in the solution.

[0049] Step 3: Using the adaptive search balance strategy, dynamically choose to update the solution by executing the benefit-seeking and harm-avoiding strategy and the local search guided by Q-learning.

[0050] Specifically, the adaptive search balance strategy divides the algorithm's iterative process into three stages, including:

[0051] In the early stage, the solution is updated through the strategy of seeking benefits and avoiding harm;

[0052] In the mid-term, the current solution is adjusted by combining the profit-seeking and harm-avoiding strategy and the local search guided by Q-learning. The profit-seeking and harm-avoiding strategy and the local search guided by Q-learning are selected according to the strategy ratio. Mixed use, By formula Generate, where is the current iteration number of the algorithm, is the total number of iterations of the algorithm, is an adjustment factor used to increase the time when the algorithm falls into a local optimum If the value of Then the strategy of seeking benefits and avoiding harm is implemented, otherwise the local search guided by Q-learning is performed. for In this stage, if the current solution is in the continuous If there is no performance improvement in the iteration, the current solution is considered to have fallen into a local optimum, and the algorithm will automatically increase , thereby increasing the frequency of using the strategy of seeking benefits and avoiding harm to jump out of the local optimal state;

[0053] In the later stage, Q-learning-guided local search is used to optimize the solution.

[0054] The strategy of seeking benefits and avoiding harm used in the early and middle stages of the algorithm iteration includes adjusting the opening of each balancing valve, determining the current optimal solution and the worst solution from the population, and updating the current solution by reducing the similarity with the worst solution while increasing the closeness to the optimal solution. The update method ,in, and for A random number uniformly distributed between is the newly generated solution Balance valve opening, The current solution The opening of a balancing valve, and is the best solution and the worst solution in the population. The opening of the balancing valve is updated and a new solution is generated after the update is completed. , if the newly generated solution The energy consumption is less than the current solution , then use the new solution Replace the current solution.

[0055] The elements used in Q-learning guided local search include state, action, reward and Q value, among which the state is the opening configuration of the balancing valve in the current heating system; the action is the three neighborhood operators to be selected; the reward is the energy-saving effect based on the selected action, that is, the degree of improvement in system energy efficiency after taking the action; the Q value is used to evaluate the heating energy-saving effect after taking the action under the current balancing valve opening configuration, and guide the decision to achieve the optimal valve adjustment strategy.

[0056] In addition, the local search guided by Q-learning includes three neighborhood operators, namely, segmentation operator, mutation operator and crossover operator, among which,

[0057] Segment operator: First, for the current solution, set the segment length and the starting subscript, and then start from the starting subscript, adjust the opening of each balancing valve in the segment according to the adjustment formula Make adjustments, including After adjustment Balance valve opening, The current solution Balance valve opening, for A random number between The adjustment step is used to control the adjustment range of the balancing valve. After the balancing valve in the segment is adjusted, a new solution is generated. ,like If the heating energy consumption is lower, the current solution in the population is replaced.

[0058] Mutation operator: For the current solution, first determine the number of balancing valves b whose openings need to be varied, and randomly select the corresponding number of balancing valves, where the range of balancing valve b is greater than 0 and not greater than 1 / 2 of the total number of balancing valves that need to be adjusted, and then use the formula Adjust b balancing valves, Used to generate Random balancing valve opening between After the mutation The opening of a balancing valve, after the adjustment of b balancing valves, a new solution is generated. ,like The heating energy consumption is lower. Replace the current solution in the population.

[0059] Crossover operator: First, randomly select another solution in the population as the reference solution to cross with the current solution, and determine the number of balancing valves to be exchanged c, and the number of balancing valves to be exchanged c should not exceed 1 / 5 of the total number of balancing valves. Then, randomly select a crossover subscript, and exchange the balancing valve opening at the crossover subscript of the current solution with the balancing valve opening at the crossover subscript of the reference solution. After the exchange of the c balancing valves to be exchanged is completed, a new solution is generated. ,like The heating energy consumption is lower. Replace the current solution in the population.

[0060] Step 4: Update the optimal solution in the current population and determine whether the maximum number of iterations has been reached. If so, the evolution ends and the optimal solution and the corresponding minimum energy consumption are output. If not, go to step 3 to continue. Specifically, in the specific example of the present invention, the maximum number of iterations of the algorithm is 1000.

[0061] The optimization method for the balance valve opening of the urban heating system based on the improved Jaya algorithm based on Q-learning proposed in this invention proposes a new solution expression method, uses an adaptive search balance strategy, dynamically adjusts the execution frequency of the solution update method in different iteration periods of the algorithm, improves the algorithm search capability, and avoids falling into the local optimal solution too early. The strategy of seeking benefits and avoiding harm improves the quality of the solution by approaching the optimal solution and away from the worst solution. In the Q-learning stage, the local search capability of the algorithm is improved by selecting three neighborhood operators, and the balance valve opening configuration with low heating energy consumption is developed as much as possible.

[0062] The present invention is further described below through a specific embodiment.

[0063] In a specific embodiment of the present invention, the population size PS=30 (there are 30 solutions in the population); (If the current solution does not evolve after 20 iterations, it is determined that the solution has fallen into a local optimum); Maximum number of iterations of the algorithm Take this as an example for detailed explanation and experimental verification.

[0064] like Figure 2 As shown in the figure, in the secondary pipe network of a heating system, a heat exchange station can supply heat to multiple end users. Each end user is affected by various environmental factors, resulting in different heat consumption. The heat exchange station uses a circulating pump to send high-temperature hot water through the heating pipeline to each end user. Here, the hot water releases heat through the radiator and floor heating equipment to heat the indoor air, thereby gradually raising the room temperature, and then the hot water becomes cold water. These cold waters enter the return pipe through the balancing valve and finally converge to the heat exchange station to complete a heating cycle. The purpose of maintaining or raising the room temperature is achieved through a continuous heating cycle. In the above process, because the adjustment of the balancing valve will affect the amount of heat supply and the power consumption of the circulating pump, the purpose of adjusting the heating energy consumption can be achieved by adjusting the balancing valve.

[0065] According to the heating situation of the above heating system, the information sets such as the terminal user number, indoor calculated temperature, outdoor temperature, building volume, building heat transfer coefficient and water supply temperature are collected. Simulation experiments are carried out and the experimental results are collected. The test examples used in the simulation experiment are 50 standard examples, among which the number of balancing valves supplying heating to the terminal users is For these scales, the number of heat exchange stations provided is 1, the number of circulating pumps is 1, and the opening range of the balancing valve is [1,100].

[0066] The Q-learning-based Jaya algorithm (QJaya for short) proposed in the present invention is compared with the artificial jellyfish search algorithm (AJS), the flower pollination algorithm (FPA), the symbiotic hybrid search algorithm (SOHS) and the population-based simulated annealing algorithm (PSA) to verify the effectiveness of the present invention. These comparative algorithms are high-performance optimization algorithms related to the energy-saving optimization problem of heating systems proposed in recent years. In order to reduce the experimental error and make the calculation results more effective and general, each example is repeated 20 times to generate statistical results. The present invention uses relative percentage increase (RPI) as a performance evaluation index. The calculation formula of RPI is RPI=(FF best ) / F best *100%, where F is the energy consumption calculated by the given algorithm for the current situation, F best is the minimum energy consumption found among all existing solutions under the same circumstances. Obviously, the smaller the RPI value, the smaller the heating energy consumption caused by the balancing valve opening configuration found by the algorithm, and the better the performance. The statistical results generated by the five algorithms when solving 50 instances of different task sizes are shown in Table 1 below:

[0067] Table 1 Statistical results of the five algorithms when solving 50 instances of different task sizes

[0068]

[0069] The table covers various instances with task scales ranging from 100 to 500. Each instance of each task scale has performance data of 5 algorithms, namely FPA, AJS, PSA, SOHS and QJaya. Each row of data represents the average solution effect of the corresponding task scale instance, that is, the RPI value, which is used to show the performance of each algorithm when processing the same task scale instance. It can be seen from the table that in instances of any task scale, the RPI value obtained by the QJaya algorithm of the present invention has a significant advantage, and QJaya is also significantly better than other comparison algorithms in terms of average RPI value, followed by PSA algorithm and FPA algorithm, and the worst performance is AJS and SOHS algorithm. In order to further verify and prove the performance advantage of the QJaya algorithm proposed in the present invention over other algorithms, the algorithm will continue to be analyzed from the aspects of the quality of the solution obtained by the algorithm, the convergence speed of the algorithm, and the scalability and flexibility of the algorithm.

[0070] Figure 3 This is an analysis of the effectiveness of the Q-learning technology introduced in QJaya. Figure 4 This is an analysis of the effectiveness of the segmentation operator technology designed in QJaya. Figure 5This is an analysis of the effectiveness of the adaptive search balance strategy technology proposed in QJaya. Figure 6 is the average and 95% confidence graph of QJaya and the comparison algorithm, Figure 7 It is the minimum energy consumption value convergence diagram of the present invention. Specifically, Figure 3 It shows the RPI comparison results of QJaya and QL-N (Q-Jaya without Q-leaning applied in QL-N) on all task scale instances. Figure 3 It can be seen that the RPI value of QJaya on any task scale instance is smaller than the RPI value obtained by QL-N, that is, the local search guided by Q-learning introduced in the QJaya algorithm is effective; Figure 4 It can be seen that the RPI value obtained by the QJaya algorithm is significantly smaller than the RPI value obtained by PARS-N (PARS-N means QJaya without applying the segmentation operator), and the energy saving effect of the solution obtained on the instances with 100 and 200 tasks is more obvious, which shows that the segmentation operator proposed in the present invention is effective and more significant on the instances with 100 and 200 tasks. Figure 5 It can be seen that in the instances with 100 and 200 tasks, QJaya has a greater advantage over SBS-N (SBS-N represents QJaya without the adaptive search balancing strategy). In the instances with 300-400 tasks, although the gap is narrowing, the effect achieved by the QJaya algorithm is still ahead of SBS-N, which proves that the adaptive search balancing strategy proposed in the present invention is also effective. Figure 6 The average and 95% confidence graph of QJaya and the comparison algorithm. From the 95% confidence interval, it can be seen that there is no overlap between the algorithms, proving that there are significant differences in performance between the algorithms. From the average, it can be seen that the proposed QJaya has the best performance in this indicator. Figure 7 It can be seen from the convergence diagram of the minimum energy consumption value that QJaya achieved the best result in the minimum heating energy consumption, which is significantly better than the results obtained by other algorithms.

[0071] In summary, through the above specific embodiments, it is verified that the QJaya algorithm proposed in the present invention significantly optimizes the opening configuration of the balancing valve in the heating system, reduces the heating energy consumption, thereby effectively reducing carbon emissions and improving energy efficiency. Through graphical analysis, it can be seen that the present invention exhibits significant performance advantages in instances of all task scales, especially in the lowest heating energy consumption index. Through multi-factor variance analysis, it is further confirmed that the QJaya algorithm of the present invention has significant differences in performance with other algorithms, proving its broad application prospects in the market. Therefore, the present invention provides an effective method for solving the optimization problem of the balancing valve in the urban heating system. Compared with the prior art, the present invention has made substantial progress in improving heating efficiency and reducing costs.

Claims

1. A heating system balancing valve optimization method based on improved Jaya algorithm, characterized in that: The following steps are included: Step 1: Initialize the algorithm parameters, set the population size PS, and the maximum number of iterations during the algorithm evolution process And the threshold of the number of times each individual in the population is allowed to evolve , where the number of solutions contained in the population corresponds to the number of solutions for the configuration of balancing valves in the secondary pipe network heating system; the threshold It represents the upper limit of the number of times that the balancing valve configuration solution fails to reduce the heating energy consumption during the iteration process; Step 2, initialize the population, encode it by using the representation method based on the balancing valve opening sequence, mix the random strategy and the linear interval strategy to generate the first solution of the population, and use the scrambling rule to generate the remaining PS-1 solutions of the population, with a total of PS solutions in the population; Step 3: Using the adaptive search balance strategy, dynamically select and execute the strategy of seeking benefits and avoiding harm, and the local search guided by Q-learning to update the solution; Adopting adaptive search balance strategy to dynamically update the solution, the specific process includes: In the early stage, the solution is updated through the strategy of seeking benefits and avoiding harm; In the mid-term, the current solution is adjusted by combining the profit-seeking and harm-avoiding strategy and the local search guided by Q-learning. The profit-seeking and harm-avoiding strategy and the local search guided by Q-learning are selected according to the strategy ratio. Mixed use, By formula Generate, where is the current iteration number of the algorithm, is the total number of iterations of the algorithm, is an adjustment factor used to increase the time when the algorithm falls into a local optimum If the value of Then the strategy of seeking benefits and avoiding harm is implemented, otherwise the local search guided by Q-learning is performed. for In this stage, if the current solution is in the continuous If there is no performance improvement in the iteration, the current solution is considered to have fallen into a local optimum, and the algorithm will automatically increase , thereby increasing the frequency of using the strategy of seeking benefits and avoiding harm to jump out of the local optimal state; In the later stage, Q-learning-guided local search is used to optimize the solution; The strategy of seeking benefits and avoiding harm includes adjusting the opening of each balancing valve, determining the current optimal solution and the worst solution from the population, and updating the current solution by reducing the similarity with the worst solution while increasing the closeness with the optimal solution. The update method is ,in, and for A random number uniformly distributed between is the newly generated solution Balance valve opening, The current solution The opening of a balancing valve, and is the best solution and the worst solution in the population. The opening of the balancing valve is updated and a new solution is generated after the update is completed. , if the newly generated solution The energy consumption is less than the current solution , then use the new solution Replace the current solution; The elements used by Q-learning to guide local search include state, action, reward and Q value, where the state is the opening configuration of the balancing valve in the current heating system; the action is the three neighborhood operators to be selected; the reward is the energy-saving effect based on the selected action, that is, the degree of improvement in system energy efficiency after taking the action; the Q value is used to evaluate the heating energy-saving effect after taking the action under the current balancing valve opening configuration, and guide the decision to achieve the optimal valve adjustment strategy; The local search guided by Q-learning includes three neighborhood operators, namely, segmentation operator, mutation operator and crossover operator, among which, The specific operation of the segment operator is to set the segment length and the starting subscript for the current solution, and start from the starting subscript to adjust the opening of each balancing valve in the segment according to the adjustment formula. Make adjustments, including After adjustment Balance valve opening, The current solution Balance valve opening, for A random number between The adjustment step is used to control the adjustment range of the balancing valve. After the balancing valve in the segment is adjusted, a new solution is generated. ,like If the heating energy consumption is lower, the current solution in the population is replaced; The specific operation of the mutation operator is to determine the number b of balancing valves whose openings are to be mutated for the current solution, and randomly select the corresponding number of balancing valves, where the range of the number b of balancing valves is greater than 0 and not greater than 1 / 2 of the total number of balancing valves. Adjust b balancing valves, Used to generate Random balancing valve opening between After the mutation The opening of a balancing valve, after the adjustment of b balancing valves, a new solution is generated. ,like The heating energy consumption is lower. Replace the current solution in the population; The specific operation of the crossover operator is to randomly select another solution in the population as the reference solution to cross with the current solution, and then determine the number of balancing valves c to be exchanged, and the number of balancing valves c to be exchanged does not exceed 1 / 5 of the total number of balancing valves. Then, randomly select a crossover subscript, and exchange the balancing valve opening at the crossover subscript of the current solution with the balancing valve opening at the crossover subscript of the reference solution. After the exchange of the c balancing valves to be exchanged is completed, a new solution is generated. ,like The heating energy consumption is lower. Replace the current solution in the population; Step 4: Update the optimal solution in the current population and determine whether the maximum number of iterations has been reached. If so, the evolution ends and the optimal solution and the minimum heating energy consumption corresponding to the optimal solution are output. Otherwise, execute step 3.

2. The heating system balancing valve optimization method based on the improved Jaya algorithm according to claim 1 is characterized in that: In step 2, the encoding steps are as follows based on the representation method of the balance valve opening sequence. Each solution consists of a one-dimensional vector Represents a one-dimensional vector The subscript corresponds to the terminal user's ID, and the subscript corresponds to The elements in represent the balance valve opening of the end user, so one solution is used to represent the balance valve opening configuration of all end users. Therefore, a solution is expressed as , Indicates that the ID is The balancing valve opening for heating the end user.

3. The heating system balancing valve optimization method based on the improved Jaya algorithm according to claim 2 is characterized in that: In step 2, the random strategy is to randomly choose The value between is taken as the balance valve opening. Indicates the minimum opening of the balancing valve. Indicates the maximum opening of the balancing valve; The linear interval strategy is expressed by the formula To generate the balance valve opening, where is the newly generated solution The opening of a balancing valve, Indicates the number of balancing valves supplying heat to end users. Center front The balance valve opening is generated by random strategy. The opening of the balancing valve is generated by alternating between the random strategy and the linear interval strategy, thereby mixing the random strategy and the linear interval strategy to generate the first solution of the population.

4. The method for optimizing the balancing valve of a heating system based on the improved Jaya algorithm according to claim 3 is characterized in that In step 2, the scrambling rules include exchange scrambling, reverse scrambling, and insertion scrambling, where: The operation process of exchange scrambling is to randomly select the openings of two balancing valves in the solution and exchange them; The reverse scrambling operation process is to randomly select an index point and reverse the balancing valve opening sequence from the index point to the end of the solution; The operation process of inserting scrambling is to randomly insert a new balance valve opening value into any position in the solution and remove the last balance valve opening value in the solution.

5. The method for optimizing a heating system balancing valve based on an improved Jaya algorithm according to claim 1 is characterized in that: In step 4, the maximum number of iterations is 1000.

Images