1. Introduction Line loss is a crucial factor affecting the economic operation of power distribution systems. With the development of the national economy and the continuous increase in electricity load, line loss has become increasingly prominent, significantly impacting the economic benefits of power supply companies. Therefore, researching methods to reduce line loss in power distribution systems has received increasing attention and importance. Power distribution networks are characterized by closed-loop design and open-loop operation. Distribution lines contain a large number of normally closed sectionalizing switches and a small number of normally open tie switches, allowing the network structure to be altered by changing the opening and closing states of these switches. Theoretically, there exists an optimal structure that minimizes line loss. The purpose of distribution network reconfiguration is to find the optimal structure that minimizes line loss while simultaneously satisfying practical operational constraints. Since distribution network reconfiguration can utilize the characteristics of the distribution network itself for network optimization without requiring additional hardware investment, it can reduce network losses while balancing loads and improving voltage quality. Therefore, it is an important means of power distribution system control and operation, and a key component of distribution management systems (DMS). Mathematically, distribution network reconfiguration is a nonlinear combinatorial optimization problem. As the system scale increases, traditional mathematical programming methods will lead to a "combinatorial explosion" problem. Currently, the main methods for solving distribution network reconfiguration problems include branch exchange algorithms, optimal flow pattern algorithms, and intelligent algorithms such as SA and GA. Branch exchange algorithms and optimal flow pattern algorithms have poor computational accuracy and cannot guarantee global optimality. SA and GA algorithms have good global optimization capabilities, but the computational load is very large. The Tabu Search (TS) algorithm is a heuristic search method that expands the neighborhood and is also a successful application of artificial intelligence in combinatorial optimization algorithms. It takes effective measures to escape local optima with a high probability, thus having strong global optimization performance. At present, the application of the TS algorithm in distribution network reconfiguration is rare. Although the literature [1] made a preliminary attempt to apply the TS algorithm in distribution network reconfiguration, it lacked effective control over the optimization process and needed to make post-judgment and processing of a large number of infeasible solutions generated during the optimization process. This paper will start from the characteristics of the distribution network itself and directly reflect the constraints in the control of the optimization, so that all solutions generated during the optimization process are structurally feasible, avoiding unnecessary calculations and improving the computational efficiency of the algorithm. [b]2. Mathematical Model of Distribution Network Reconfiguration[/b] From a mathematical perspective, distribution network reconfiguration belongs to a large-scale nonlinear combinatorial optimization problem. Distribution network reconfiguration with the goal of minimizing network loss can generally be expressed as the following minimum optimization problem: 1) Power flow equation constraints; 2) Network structure constraints, including radial and islanded networks; 3) Line capacity constraints. [b]3. TS Algorithm[/b] The TS (Tabu Search) algorithm is a highly efficient modern heuristic optimization algorithm that has received widespread attention in recent years. It was first proposed by F. Glover in the late 1970s and has been successfully applied to various fields with the development of computer technology, solving a large number of complex optimization problems. In recent years, the algorithm has been introduced into the field of power system analysis, such as the joint economic dispatch of hydropower and thermal power [2], reactive power optimization of power systems [3], and optimal planning of transmission systems [4], and has achieved certain research results. The basic idea of ​​the TS algorithm is to use a flexible "memory technology" to record the optimization process that has been carried out, so as to guide the next search direction. To avoid getting trapped in local optima, TS allows the search to move in the direction that minimizes the degradation of the objective function and restart the search. The algorithm has three fundamental elements: movement, the Tabu table, and release levels. 3.1 Movement The search process of the TS algorithm is implemented through movement; therefore, movement is the foundation of the TS algorithm. There are many ways to move, such as single-step movement, exchange movement, and multi-point movement, and the specific movement used depends on the problem being studied. During the search and optimization process, TS selects the movement that maximizes the improvement of the objective function under the constraints. If no such movement exists, it chooses the movement that minimizes the degradation of the objective function. 1) Single-step movement 2) Exchange movement Exchange movement is implemented by combining two single-step movements. For the distribution network reconfiguration problem, its physical meaning is: closing switch i while simultaneously opening switch j. 3.2 Tabu Table The Tabu table is the key to the TS algorithm and its most obvious feature distinguishing it from other algorithms. It is used to store the reverse movements of movements that have already occurred. Any movement existing in the Tabu table is prohibited from being used in the current iteration. TS effectively prevents the return of already visited local optima during the search process through this method, creating favorable conditions for obtaining the global optimal solution. There are various ways to manage the Tabu table, and this paper adopts a first-in-first-out (FIFO) queue for management. Reference [1] argues that if the new current solution is generated by moving tij, then the Tabu table needs to store the moves tjk, tki, and k for all possible values. That is, if the current solution is generated by closing switch i and opening switch j, then all moves related to opening switch i or closing switch j will be stored in the Tabu table. However, as the complexity of the system increases, the range of values ​​for k will be very large. Since the Tabu table needs to store information from multiple iterations, on the one hand, the length of the Tabu table will increase significantly, and a large number of elements need to be moved in and out each time the Tabu table is updated; on the other hand, the search efficiency of the Tabu table will also be greatly reduced. This paper argues that for the current move tij, the Tabu table only needs to store tji, because tji is sufficient to contain all the information of tjk and tki. For example, for a move tmn, if m equals j or n equals i, then tmn is considered to be within the tabu range of tji. In this way, the Tabu table records sufficient information while avoiding the aforementioned shortcomings. The maximum number of elements that the Tabu table can store is called the length of the Tabu table, which has a significant impact on the search. If the length of the Tabu table is too long, it imposes too many restrictions on moves during the search process, which may prevent moves that can produce excellent experimental solutions; conversely, if the length of the Tabu table is too short, it imposes too few restrictions on moves, which may cause the search to loop and get stuck in local optima. Therefore, the length of the Tabu table is crucial to TS, but how to determine its optimal value is still a problem to be studied. The general principle is that the length of the Tabu table increases with the size of the research problem. 3.3 Release Level Although the Tabu table is an effective means of avoiding local optima, it may also prevent further optimization of the solution, which is obviously detrimental to the optimization process. The "release level" is used to solve this problem. For a valuable move, even if it is in the Tabu table, it can be released from the Tabu table as long as the "release level" is reached. The release level adopted in this paper is: when a move in the Tabu table acts on the current solution and can produce the best solution so far, the move is considered to have reached the "release level". 3.4 Processing of the TS algorithm in the distribution network reconfiguration problem. Like other algorithms, the key to solving the distribution network reconfiguration problem with the TS algorithm is how to combine the algorithm with the problem under study to improve the computational efficiency and accuracy of the algorithm. TS is a stochastic search algorithm. If the characteristics of the distribution network reconfiguration problem itself are not considered, a large number of infeasible solutions will be generated during the optimization process, which will greatly affect the computational efficiency. For example, the generated solutions may not satisfy the radial structure or network islands may appear. Therefore, it is necessary to control the optimization process based on the characteristics of the distribution network reconfiguration problem to avoid the generation of infeasible solutions. This paper takes the following measures: 1) The initial solution is taken from the original structure of the distribution network; 2) Only exchange moves are used, because single-step moves will inevitably produce isolated nodes; 3) When performing exchange moves, after closing one switch, another switch can only be opened within the formed loop. Through the above three measures, the generation of the initial solution and each experimental solution strictly follows the structural constraints of the distribution network. Therefore, any solution generated during the optimization process is structurally feasible, thus avoiding the judgment and processing of a large number of infeasible solutions and saving computation time. [b]4. Solution Steps[/b] The main steps of applying the TS algorithm to solve the distribution network reconfiguration problem are as follows: 1) Read in the original data. This includes network parameters, Tabu table depth, maximum number of iterations Kmax, and the number of experimental solutions Smax generated in each iteration; 2) Generate the initial solution R0, which is the original structure of the network in this paper. Set the current solution RC=R0 and the optimal solution Ropt=R0. 3) Generate experimental solutions. Apply the exchange movement tij to the current solution to generate an experimental solution. i is randomly determined by the set of open switches, and j is randomly selected from the loop formed by the closed switch i. Calculate the corresponding power flow and objective function values. If there is a line capacity or node voltage exceeding the limit, regenerate the experimental solution. Repeat this step until the number of experimental solutions reaches the required number Smax; 4) Update the current solution. In the trial solution, select the solution R* with the optimal objective value. If the move that generated this solution is not in the Tabu table, or if it is in the Tabu table but has reached the release level, then update the current solution Rc with it. If the move that generated this solution is in the Tabu table but has not reached the release level, then select the second-best solution and repeat this process. 5) Update the Tabu table. Store the reverse moves that have been realized into the Tabu table. 6) Update the optimal solution. If the objective value of the new current solution is less than the objective value of the optimal solution, then update the optimal solution with the new current solution. 7) If the number of iterations has not reached Kmax, go to step 3); otherwise, end. [b]5. Example[/b] The example used in this paper comes from reference [5]. The distribution system has 33 nodes, 32 branches, and 5 tie lines, with a rated voltage of 12.66 kV. The system structure is shown in Figure 1. In the calculation, it is assumed that each branch is equipped with a switch, and the parameters related to TS are: Kmax=20, Smax=10, and the Tabu table depth is 6. The calculation results are shown in Table 1. For easy comparison, the calculation results of genetic algorithm [6] and ant colony optimization algorithm [7] are also given in the table. [b]6. Conclusion[/b] As a new modern heuristic optimization algorithm, TS has been proven to be an effective method for solving complex combinatorial optimization problems. This paper introduces the basic principle of the TS algorithm and, based on the characteristics of the distribution system itself, applies the TS algorithm to solve the distribution network reconfiguration problem. By selecting and controlling the movement of TS, the problem of generating a large number of infeasible solutions in the optimization process is effectively solved, and the computational efficiency is improved. Calculations based on actual examples show that the TS algorithm is very suitable for solving the distribution network reconfiguration problem. **References** 1. Wang Xiuli, Wu Hongxiao. Distribution network reconfiguration aimed at improving system reliability [J]. China Electric Power, 2001, 34(9): 40-43 2. Bai X, Shahidehpour S. Hydro-thermal scheduling by tabu search and decomposition method [J]. IEEE PWRS, 1996, 11(2): 968-974 3. Gan D, Qu Z, Cai H. Large-scale var optimization and planning by tabu search [J]. Electric Power System Research, 1996, 39(3): 195-204 4. Wen Fushuan, Han Zhenxiang. Optimal planning of transmission system based on Tabu search method [J]. Power System Technology, 1997, 21(5): 2-7 5. SK Goswami, Sk Basu. A new algorithm for the reconfiguration of distribution feeders or loss minimization [J]. IEEE Trans. on Power Delivery, 1992, 7(3): 1484-1491 6. Liang Yong, Zhang Yan, Hou Zhijian. Application of genetic algorithm in distribution network reconfiguration [J]. Journal of Electric Power System and Automation, 1998, 10(4): 29-34 7. Chen Genjun, Wang Lei, Tang Guoqing. Distribution network reconfiguration algorithm based on ant colony optimization [J]. Journal of Electric Power System and Automation, 2001, 13(2): 48-53 8. Lu Hongyu, Hu Linxian. Real-time optimization switching strategy of distribution network capacitor based on genetic algorithm and TS algorithm [J]. Power System Technology, 2000, 24(11): 56-59 Journal of Electric Power System and Automation