Reconstruction of Low Voltage Distributed Power Supply in Distribution Network Based on Improved Particle Swarm and Fish Swarm Fusion Algorithm

       Thank you for visiting Nature.com. The browser version you are using has limited CSS support. For best results, we recommend using a newer browser (or disabling compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we will display the site without styles and JavaScript.
       With the development of distributed power sources in the distribution network, distribution network reconstruction algorithms have attracted the attention of experts and scholars. Its purpose is to reduce the power loss in the process of power transmission, thereby reducing the loss of the power grid in the process of power transmission. . And weaken the electrothermal effect in the process of power transmission, so as to ensure the safety of the surrounding residents. Due to the wire impedance effect, a large amount of electrical energy in the circuit is lost due to electrical heating, which can easily cause local overheating and cause a fire. This not only causes electrical energy loss, but also endangers the safety of the surrounding residents. To solve this problem, an improved particle swarm fish algorithm is used to conduct a distribution network reconstruction experiment based on Elecgrid’s own dataset. First, low-voltage distributed power supplies are connected in parallel to the circuit to reduce the internal resistance and electrical heating. Then, through the circuit in the control system, the dual split relay adjusts the inductance and capacitance of the wire, thereby reducing the reactance length. In addition, the particle swarm particles are mutated to jump out of the local optimum, and the elite fish method is used to expand the search space. Finally, the proposed fusion algorithm is applied to the dataset generated by Elecgrid and compared with other three algorithms. The fusion algorithm is used as a standard test system for this comparison. Under the working voltage of 0.74 V, the active power loss of the hybrid algorithm is 63 kW, and the power losses of the other three algorithms are 74 kW, 97 kW and 109 kW, respectively. The loss of the hybrid algorithm is the smallest among the four algorithms. The experiment was repeated 6 times, and the linear fitting degrees of the four algorithms were 0.9804, 0.9527, 0.9612 and 0.9503, respectively. The experimental results show that the application of this algorithm can effectively reduce the active power loss in the reconstruction process of the distribution network, thereby reducing the energy consumption. The main points of this study are threefold: firstly, the algorithm is used to reduce the resistance in the network transmission path so as to more accurately analyze the efficiency of power transmission; secondly, the new algorithm enriches the methods for ensuring the security of power supply; finally, the fusion algorithm is used to reduce the fire caused by local overheating of lines;
       Nowadays, household appliances occupy a large part of residents’ lives. Efficient transmission of electricity is not only related to economic savings but also to the personal safety of residents. In the power grid, the distribution network is the bridge between power plants and residents, and usually works in the form of long-distance power transmission. The longer the wire, the greater the resistance and reactance, resulting in the loss of electrical energy in the form of electrical heat during power transmission1. During transportation, multiple components installed along the way can affect the charge distribution2. Even if a branch switch is used for disconnection, the subsequent increase in the maintenance cost of the switch will increase the effect. In recent years, the fusion algorithm of the artificial fish swarm (AF) algorithm and the particle swarm optimization (PSO) algorithm has received wide attention from experts due to its global search capability and potential for studying the reconstruction of the distribution network3. However, PSO-AF is only suitable for medium and long-distance distribution networks. When the distance from the power grid is less than 10 km or more than 25 km, this model has a problem of slow data transmission. When operating over long distances, it is easy to fall into local optimality, which will lead to premature termination of the algorithm4. To solve this problem, this study creatively optimized the PSO-AF algorithm based on the mutant bird (b) and elite fish (e) strategies. The novelty of this paper is the generative fusion algorithm (bPSO-eAF). b can modify particles when the algorithm falls into the local optimum. e is more active, has a wider search range, and can effectively target long-distance power transmission. When multiple components affect the charge distribution, the bPSO-eAF algorithm can reduce the inductance and capacitance of the power system and improve the capacity and stability of the power transmission system. The study is mainly divided into four parts. The first part mainly analyzes and summarizes the current applications and effects of PSO and AF. The second part introduces the factors affecting the power transmission efficiency and constructs a reconstruction model of bPSO-eAF. The third part analyzes and compares the performance of the optimization model and the traditional model. In the last part, simulation experiments are conducted on the Elecgrid dataset and the shortcomings in the study are proposed. There are three main advantages in this study: First, the algorithm is used to reduce the resistance on the transmission path, which can more accurately analyze the efficiency of power transmission. Second, the new algorithm extends the power security method. The fusion algorithm is used to reduce the number of fires caused by local overheating. The aim of the study is to solve the problem of large losses in long-distance power transmission of bPSO-EAF, so as to reduce the risk of fires caused by local overheating of transmission lines, thereby reducing the loss of fossil fuel energy.
       Reconstruction of low-voltage distributed power supplies in the power distribution network is of great significance to ensuring the safety of residents and regulating energy consumption. Liao Zhijun et al. 5 developed a parameter fitting method based on the hybrid particle swarm and fish swarm algorithms by taking the experimental variance value given by the variance function as the optimization target and converting it into a minimization problem by setting the objective function. Their hybrid algorithm has strong convergence and can obtain satisfactory approximation values. Comparing the fitting results of VARFIT with the optimization algorithm, the fitting results of the optimization algorithm are 3.39 lower than that of VARFIT. The experimental results show that the method has strong optimization capability and high optimization accuracy, and can effectively realize automatic parameter tuning. Based on the performance of wireless sensor networks, Yang Zhijun et al. 6 analyzed and summarized their advantages and disadvantages and proposed a clustering method based on the K-means algorithm. To maximize network coverage while ensuring quality of service, a wireless sensor network coverage optimization method based on an improved artificial fish swarm algorithm is proposed and the performance of the proposed algorithm is analyzed through controlled experiments. The experimental results show that the proposed method has certain advantages and can serve as a benchmark model for related research. Jain et al. 7 proposed a robust hybrid annealing optimization algorithm, simulated annealing, which improved the resilience of wild networks against adversaries without changing the temperature distribution. This algorithm eliminated the unpredictable fluctuations of the algorithm. It also accelerates the convergence rate. To simulate the generation of scale-free networks, they used real networks to verify the results of synthetic scale-free networks. The experimental results show that this model is superior to other models in many aspects.
       Gunawan et al.8 proposed four new independent optimization algorithms to solve single-objective problems in the field of renewable energy. They compared the proposed metaheuristic optimization with alternative methods including particle swarm optimization, moth flame optimization, and gray wolf optimization using the IEEE 30-node benchmark system. To address the challenges of state-of-the-art energy models, they propose tests under different operating conditions to detect the presence or location of renewable energy sources. The results showed that their proposed algorithm could solve the problem more efficiently compared to other algorithms. Compared with the PSO algorithm, their algorithm improves the bias by 27%.9 proposed a generalized space transform evolution method and combined it with an improved particle swarm optimization algorithm. Their generalized space is based on contrastive learning, which not only improves the utilization of the current space but also extends its exploration. The improved algorithm uses spatial transformation search for generational leaps, and experiments with well-known unconstrained benchmark functions show empirical evidence of the generalized spatial contribution. They also compared particle swarms with quantum behavior with some typical scaling methods, and the results showed that their optimization algorithm was more promising than other algorithms. Yoganand et al. 10 used two algorithms, genetic algorithm and particle swarm algorithm respectively, to solve the service composition problem. Due to the large number of services with similar functions, various characteristics have become a key issue in service management, and the service quality is composed of various factors such as service cost and reliability. Selecting a suitable IoT-based supply chain to meet user needs is the key problem they solved, and they used particle swarm optimization to solve the supply chain problem and further evaluate it. The comparison results show that the genetic algorithm can improve the supply chain efficiency and outperform the particle swarm algorithm. Zhang 11 believes that distributed technologies have high efficiency in road construction, especially highways. He found the advantages of using distributed energy in tunnels. In order to realize the effective application of this technology, a tunnel power supply highway was selected and the effect of its application was discussed. The experimental results show that this technology has advantages and application possibilities in highway tunnels, and can also serve as a model for the transformation of distributed power distribution networks on highways. Daus et al.12 considered the penetration effect of photovoltaic technology when connecting to the grid, improved the efficiency of the optimized distributed power generation when selecting parameters, and based on their distributed power generation system, believed that the solar energy intensity would decrease with the consumption. Finally, an experiment was conducted on distributed power supply, and the experimental results show that the method is effective. Naguib et al.13 proposed a distribution network reconstruction method that can reduce the loss of distributed distribution networks and improve the working environment of the distribution system. Based on renewable energy, they discussed its intermittent load curve to minimize the annual energy loss. to the goal. The key point of this method is the firefly algorithm, and experiments have been conducted on a nodal power distribution system. The results of numerous examples show that this method is effective in reducing losses in the power grid.
       Due to the complexity and variability of distribution networks, recent research efforts have some shortcomings such as algorithm uncertainty. These methods mainly focus on theory and modeling and have not yet been applied and verified on a large scale. In addition, their methods require additional data support and real-world case studies for further refinement and improvement. Articles by various scholars indicate an international interest in distribution network reconstruction algorithms. But fusion algorithms have not received much attention so far. For the first time, Mutant Bird and Elite Fish were combined with PSO-AF to create a fusion algorithm (bPSO-eAF) that takes into account the influence of components such as low-voltage power distribution and relays.
       With the development of electric power industry, low-voltage relay protection is becoming more and more important, and research on distribution network transformation has emerged. However, objective factors such as large heat loss and low transmission rate in distribution networks make the task more difficult. model transformation and further increase the complexity of relay protection. This paper combines particle swarm optimization (PSO) algorithm and artificial fish swarm (AF) algorithm. It first introduces a model based on these two, and then explains how to combine them to achieve relay protection of 35 kV and below. The swarm-fish swarm algorithm has good parallel processing capability and can conduct a global search on the entire distributed power source. Compared with some complex algorithms, the swarm-fish algorithm is simpler, easier to implement, and has adaptive capability. The search step size can be adjusted according to the distribution network conditions, which makes it easy to adapt to the needs of optimization and reduce network losses.
       The workflow of the PSO algorithm simulates the similar behavior of a flock of birds during feeding. The search area of ​​the PSO algorithm is set to the forest in the region, and the particles represent the flocks of birds that feed in it. Then the position of the bird flock in search of food is called the global optimum, and the bird flock that finds food is defined as the optimal solution, causing the rest of the bird flock to gradually approach this position. In the iterative process of the PSO algorithm, the particle position update method is based on the formula (1)14.
       In the formula (1), \(d\) is the size of the forest where the group of birds is located, and the number of birds is \(i\). \(t\) represents the number of velocity updates, and the position of the bird at this time is \(x_{i,d}^{t}\). \(v_{i,d}^{t + 1}\) represents the particle velocity update method, as shown in the formula (2) 15.
       In equation (2), the speed of the bird at this time is written as \(v_{i,d}^{t}\), and the position of the food is written as \(p_{i,d}^ {t}\). The learning environment of the bird flock is written as \(c_{1},c_{2}\), and its value is 2. \(r_{1},r_{2}\) takes the value at \(\left[ {0,1} \right]\), and its value should satisfy randomness. The flow chart of the PSO is shown in Figure 1.
       The complete PSO process includes particle initialization, particle adaptation, and position updating. The most important of these is the tuning of various parameters. Too many PSO particles can easily fall into local optimality, and too few iterations can easily result in no PSO being found. The optimal solution is to use the PSO particle velocity update method to reconstruct the mesh, as shown in Equation (3) 16.
       When \(v_{i,d}^{t} = x\), as \(v_{i,d}^{t}\) increases, \(Sigmond\left( {v_{i, d}^{ t}} \right)\) gradually converges to \(1\). As a mature algorithm, the PSO algorithm has a relatively small parameter tuning range and does not affect the final result. It only needs to be tuned step by step through information matching. During the running of the PSO algorithm, there is a sensitive parameter \(\omega\), which is closely related to the update rate of the PSO particles. The larger the value of \(\omega\), the larger the global search range. It is beneficial for the PSO algorithm for the particles to jump out of the local optimum; the smaller the value of \(\omega\), the shorter the running time of the algorithm. However, an incorrect value of \(\omega\) will cause the oscillation range of the PSO curve to increase in the later period. Therefore, Equation (4) is true when \(\omega\) is 17.
       In the formula (4), the maximum iteration number of the PSO algorithm is written as \(T\), where the number of iterations is written as \(t\), and the maximum and minimum values ​​of \(\omega\) are written as \(\omega_ {\max } ,\omega_{\min }\). Therefore, the objective function of this model can reduce the loss of the transmission line by changing the topology of the power grid. Due to the anti-interference ability of the power grid, the faulty distributed power supplies can still maintain stable power supply. In addition, the objective function of the distribution network reconstruction model has limitations, that is, the obtained reconstruction scheme satisfies various operation requirements of the power grid. In this study, the actual configuration of the power grid equipment is considered, the power grid topology is maintained, and the arrangement between the nodes is satisfied. When the power load of each node achieves balanced distribution, the power distribution equipment operates within the rated power range. . The survival strategy of AF is manifested in searching for food. This means that fish will be attracted to areas with rich food sources, which also improves the searching ability of AF, as shown in equation (5)18.
       In equation (5), the survival difficulty of a fish school is written as \(h\), the current position of the energy food is written as \(E\), the iteration period of the population is written as \(T\), \(\lambda\) represents the regional time. The energy lost by fish schools as a result of production activities. The competition mechanism of AF is reflected in the competition between individual fish to occupy the position with the highest energy. Interspecific competition in AF not only promotes competitive mating of the strongest individuals, but also enhances the local searching ability of fish that have not reached the optimal position, as shown in equation (6) 19 .
       In the above equation (6), the energy level at the global optimal position of the fish school is written as \(E_{\max }\), and the scaling factor between them is represented by \(\varepsilon\). The particle motion characteristics are consistent with randomness, so it is difficult to find a loop in the two-dimensional full mesh topology. To avoid this problem, a network identification framework is established. By observing the loop state of the smallest branch in the network, the network loss weight of this branch is calculated. Then, restore the loop operation, as shown in Figure 2.
       In Figure 2, there are a total of 33 barriers, 32 segment switches and 5 contact switches. The representation of the closed loop by a circular arrow helps to prevent local overvoltage at the loop nodes. The distribution network shown in Figure 2 can be solved step by step in the same way as the admittance matrix and impedance matrix in the busbar calculation. However, this method has certain limitations when the number of network nodes increases and the impedance range increases. To reduce this effect, an improved equation proposed by Niura is shown in Equation (7) 4.
       In equation (7), \(\Delta P\) represents the active power of the grid, and the reactive power is . \(\Delta Q\), \(H, N, J, L\) represent the voltages of four critical points of the power grid. The potential difference between the start point and the end point is denoted by \(\Delta U\). The phase voltage is denoted by \(\delta\), and the value of \(\cos \delta\) is \(\left[ {0,0,5\pi } \right]\). The Oxla equation establishes the relationship between potential and electric power without considering the influence of resistance and reactance. To overcome this limitation, this study proposes a modified Oxla equation 20 in equation (8) below.
       In the above equation (8), \(B^{\prime}\) represents the error of the unexpected work performed by the flow. The Oxla equation converges quickly in calculating the power flow, but has certain requirements for the primary value. Unreasonable primary values ​​can easily fall into local optimality.
       When applying the PSO-AF model to transform low-voltage distributed power grids, a common problem is that it falls into local optimality and narrow search objectives. In order to solve the problem of premature termination of the PSO algorithm, binary coding is added to PSO. It is used to improve the performance of the algorithm, as shown in equation (9) below21.
       In the above formula (9), \(r_{i,d}\) represents the binary encoding constant, and the range of values ​​is \(\left[{0,1} \right]\. All the switching data sets in the loop scheme are denoted by \(K\). Using Equation (9) to update the particle position, even if there is a breakpoint in the working scheme, it can be avoided by looping at this position, thereby jumping out of the local optimum. The implementation of binary PSO increases its workload. In order to ensure the efficiency of the improved algorithm, the breakpoint distribution of discrete branches is designed based on the circuit tail, as shown in Equation (10) 22.
       In equation (10), the probability of loop tripping is written as \(\beta\), the total number of bridge loops is written as \(Y\), \(l\) represents the bridge loop of the current circuit, and the number of turns of the circuit spiral is written as \(Z\), the current spiral turn is written as \(k\). The dynamic PSO model constructed using equation (10) relates the movement of losses in the network of the composite system. It is assumed that the changes in the weighting factors within the regional time are 2, 1, and 2, respectively. Then the change in network losses during this period can be described by equation (11) 23.
       In equation (11), the range of network loss variations over a period of time is written as \(Fara\), in which the effect of time is represented by \(\Delta t\). \(f_{k}\) represents the net loss at the current time. The maximum, median, and minimum net loss at this time are written as \(f_{\max }, f_{mid}, f_{\min }. \) respectively. The range of \(k\) is \(k \in \left[ {0,\max } \right]\). The improved PSO in equation (11) can fully represent the net loss in the region at any time, so that the model can accurately account for the lifetime error. The lifetime error refers to the circuit selection process in the distribution network, and it has a negative impact on distributed power sources, usually due to unexpected events or already occurred. The traditional distribution network consists of voltage arrays, relays, closed networks, and consumer groups. On this basis, an experimental circuit was constructed, shown in Figure 3.
       Figure 3 shows the circuit diagram used in this experiment. Since this distributed power supply consists of voltage modules connected in parallel, the electric heating effect can be maximized. Using the relay in the double-split mode also leads to a decrease in the reactance length. In order to expand the search range of the AF, multi-criteria optimization of fish schools in the AF is carried out, and then the optimal solution is found in a changing environment. For stable power supply from distributed power supplies, the determination formula is shown in equation (12) 24.
       In the formula (12), the total number of branches of the distribution network is written as \(n\), the current branch is denoted by \(i\), and \(V_{i}\) represents the actual operating potential. The difference, the rated voltage here is expressed as \(V_{iN}\). The optimized AF using the equation (12) expands the fish group to a certain extent, but also takes into account the elite fish in the fish group. The search range of the elite fish is larger, and the gene should be better inherited. The congestion degree requirements of the total radius and the calculation method are shown in the equation (13)25.
       In Equation (13), the experimental extraction point is written as \(i\), the crowding degree of the point is written as \(s_{i}\), the total number of constraints is \(m\), and the number of targets is \(j\), the function of \(j\) with respect to \(i\) is written as \(\chi_{j}^{i}\). According to Equation (13), the crowding degree of elite fish in the neighborhood can be calculated to determine the optimal survival position of the fish. Then, select the best individuals among the offspring, merge them, and repeat this operation until a predetermined number of times is reached. Finally, the PSO optimized for binary coding is combined with AF taking into account the elite fish, and the merging algorithm bPSO-eAF is obtained. The process is shown in Figure 4.
       The workflow of the bPSO-eAF algorithm, shown in Figure 4, can be divided into four stages. During the whole process of the algorithm, the optimal solution can be selected from the distribution network data, and the solution sets that do not meet the requirements can be selected. be mixed with the main set. When reconfiguring the network, study the traffic limitation of certain links through switches. These tools can be adjusted according to the real-time network traffic situation to ensure that the network link constraints are maintained. For the network link parameters, select the quality strategy and adjust the distribution parameters according to the flow of the distribution network. Using the above method, the link constraints are maintained in the reconstructed network, while the link parameters of the selected network are selected. In this study, it is assumed that there are obstacles in the operation of the POP2 process in this experiment, which can easily cause the model to fall into local optimality during operation, thereby affecting the experimental results of the study. The researchers believe that in order to reduce this loss, the accuracy of the experiment is more important, so after thorough consideration, the POP2 process was removed from this experiment. When the distribution network is operating, the transformer power is output as losses, and formula (14)26 is established to calculate its losses.
       In formula (14), the output power of the transformer as a single electrical appliance is expressed by \(\iota_{0z}\). The power loss at nominal resistance is denoted by \(\iota_{kz}\). The coefficient when the circuit is not supplied is denoted by \(K_{Q}\). \(0,k\) represents the operating time and the current duty cycle, respectively. Under ideal operating conditions involving multiple transformers, their failure time (e.g., short circuit) is calculated using equation (15) 27 .
       In equation (15), the action time is written as \(T\), the probability of failure at this time is represented by \(u_{1}, u_{2}\), and the time required to restore the working state is written as \(r_ {1},r_{2}\), so the failure time can be expressed by \(u\). In the low-voltage circuit with a supply voltage of 35 kV, study the overload check of each electrical appliance, determine the loss in the circuit, and determine the relay protection method. After that, first study the debugging risks of each equipment in the circuit, determine the location of the equipment, and ensure the timely replacement of the faulty equipment. Then, set the power supply to the normal line voltage and debug the parameters of the device accordingly. Then estimate the maximum current in the circuit based on the power supply busbar. The operability of the device is judged by its overload current. Finally, in order to prevent reverse current from flowing in the circuit, a diode is added to circuit 28. The pseudocode generated by this model is shown in Table 1.
       In this chapter, we construct a bPSO-eAF model based on a low-voltage distributed power supply and verify the performance of the bPSO-eAF algorithm on real distribution networks through iteration, accuracy verification, and other methods. Finally, the bPSO-eAF model is applied to conduct simulation experiments on the Elecgrid dataset.
       In this study, the Elecgrid dataset in the IEEE standard test system is selected and several objective functions are set. In the process of creating this dataset, the researchers first collected data related to the power grid, including load, energy consumption, etc., and then cleaned the data to remove erroneous data. Then, integrate the data from different sources, and finally standardize the data to ensure the consistency of the data format and units. Through the above steps, the Elecgrid dataset was created, which became the dataset for this study. Due to the small number of data types, the dataset is divided into training set and testing set in the ratio of 4:6. The hardware and software used in the experiment are shown in Table 2.
       In a distribution network containing distributed energy sources, the objective to be optimized usually consists of several objectives. If the requirements of simultaneously reviewing all objectives and fulfilling important objective functions are met, the optimal compromise solution under the state can be found in the dominant solution. Based on the optimized objective function, dominant and non-dominant experiments were conducted to calibrate the reactive power loss and network loss of each user when the circuit is in the standby state. The calibration results are shown in Figure 5. Figure 5 shows the experiments between single-shot and multiple-shot objective functions on a three-feeder diagram. S5 represents the amount of change in loss when the objective function reaches the optimal state. S6 represents the network loss of the single-shot objective when the maximum load is 10. S1, S2, S3 and S4 respectively represent the switches of the three feedback curves. By adjusting the on and off switching of the switches, the three feedback curves can be controlled, thereby adjusting the standby state of each branch.
       In Figure 5, the reactive power loss and no-load network loss are 162 kW and 245 kW, respectively. This is because the distribution network reconstruction model is mutually constraining and will occur due to the specific conditions in reality. distribution network reconstruction to achieve efficient access to distributed energy resources. The calibrated scheme requires further processing of the collected dataset before conducting experiments. The experiment uses bPSO-eAF for iterative optimization and compares convolutional neural network (CNN), random forest (RF), backpropagation neural network (BPNN), fusion algorithms, etc. Among them, CNN can automatically learn the features of the original data and is widely used in various reconstruction technologies. It has translation invariance and local sensitivity. When solving problems, it can extract the characteristics of the power system through a series of layers and reconstruct them. RF is an ensemble learning method that achieves classification and regression tasks by merging multiple decision trees. This method can reduce the phenomenon of overfitting by combining multiple decision trees, thereby improving the accuracy and generalization ability of the model. When RF processes datasets with a large number of features, it needs to perform feature selection or dimensionality reduction. Since there are a large number of missing values ​​and outliers in the distribution network, this method is highly specialized. The BPNN method is a gradient descent method that works well in both classification and regression problems. It consists of multiple neurons and contains forward and backward propagation paths that can effectively filter outliers in the data. This process passes the parameters to the update network to minimize the loss function. All three methods can learn complex nonlinear relationships and work well with a large number of features and samples, which are of great comparative importance in the reconstruction of the distribution network. The results of the accuracy of the training set and the error rate are shown in Figure 6.
       In Figure 6, as the number of iterations increases, the accuracy of the bPSO-eAF algorithm stabilizes at 155 iterations. At this time, the accuracy is higher than that of the other three algorithms, and the error rate is also at the lowest level. After 155 iterations, the accuracy and error rate of bPSO-eAF are not obvious because the number of iterations increases, and the impact on the experimental results is very small and can be ignored. Considering the cost saving and the efficiency of the algorithm, the number of iterations was set to 155. At this point, the training of the bPSO-eAF algorithm is completed, considering the changes in load and power supply stability caused by the increase of reactive power loss, as shown in Figure 7.
       In Figure 7, as the reactive power loss in the circuit increases, the circuit load and power supply stability change unevenly, so we can only find the point where the three values ​​​​can bring the circuit to the best working state. . When there is no circuit. When the power loss value is exactly 162 kW, the power supply stability value is 0.083, which is enough to meet the needs of the experiment. The circuit load is 0.72, which can withstand the work of a fixed power supply. -nominal resistor.
       To verify the bPSO-eAF algorithm model, the parameters are set in a loop containing up to 6 objective functions, and the activity of four algorithms that can solve the set simultaneously is examined when the number of infeasible solution sets is determined. . Use different graphical representations for each algorithm, set the solution whose activity value is greater than the infeasible solution set as the dominant solution set, and use the number of objective functions to represent it. When the number of infeasible solution sets is determined, four algorithms can be used at once. Cancel the activity. The experimental results are shown in Figure 8.
       As shown in Figure 8, when the scheme is composed of 5 objective functions, the activity of the dominant solution set bPSO-eAF is 0.94, close to 1, which is better than that of the other three algorithms. When the objective function is composed of 2, 3, 4 and 6, the activity of bPSO-eAF is lower but still higher than that of CNN, RF and BPNN. Before and after the scheme reconstruction, the voltage of the scheme changes, so the potential difference between the four algorithms in the reconstruction process is analyzed, and the results are shown in Figure 9. The reconstruction of the distribution network can plan and design the distribution network according to the existing load to determine a reasonable network structure; it can upgrade the obsolete equipment, including the replacement of old transformers and protection devices, and install new intelligent equipment, thereby improving the reliability and operation efficiency. The equipment efficiency can be reasonably implemented, including the installation of smart meters and remote monitoring systems, to achieve real-time monitoring of the distribution network.
       In Figure 9, the bPSO-eAF voltage value at node 4 before reconstruction is the highest. At this time, the voltage values ​​of bPSO-eAF, CNN, RF and BPNN are 0.83, 0.76, 0.64 and 0.77 V, respectively. After reconstruction, the bPSO-eAF voltage value at node 4 is the lowest. At this point, the voltage values ​​of the four algorithms are 0.29, 0.32, 0.31 and 0.39 V, respectively. The optimization effect of this operation on bPSO-eAF is stronger than that of the other three algorithms, and the reconstructed bPSO-eAF algorithm has the best effect. Therefore, in order to analyze the power loss of the protruding solution installed in the active mode, the change in potential difference at both ends is measured by electrical equipment, the image is shown in Figure 10.
       As the voltage increases, all the algorithms in Figure 10 perform more additional work. The Awesome algorithm in Figure 10 indicates that it performs better in the distribution network reconstruction experiment, and the low algorithm indicates that it performs better in the distribution network reconstruction experiment. The experimental results were not ideal. Since the device at node 4 works at 0.74 V, this study only analyzes the active power loss of four algorithms at 0.74 V. As can be seen from Figure 10, the bPSO-eAF algorithm has the lowest active loss value, which is 63 kW. As can be seen from Figure 10, the bPSO-eAF algorithm has the lowest loss value, which is 63 kW. The active power loss values ​​of CNN, BPNN, and RF are 74, 97, and 109 kW, respectively. However, the results obtained in one experiment could not satisfy the requirements of the generality of the algorithm, so 6 experiments were conducted under the same conditions and the results of each experiment were compared with the expected values. The drawn straight line approximation graph is shown in Figure 11. With the rapid development of new energy, distributed energy sources such as solar photovoltaic systems should be taken into account in the reconstruction of the distribution network. Among them, a large part is due to the adjustment of the distribution network structure, which can ensure smooth access to sustainable energy. This paper studies how to strengthen the safety and reliability management of distribution networks by improving the equipment automation systems, which can enhance the equipment redundancy and thereby improve the fault detection capability. The data show that the reconstruction of the distribution network is a complex and comprehensive project that requires comprehensive consideration of many factors such as technology, economy, environment and society. Study and adopt reasonable planning and design to improve the reliability of electricity supply. This approach can promote energy transition and sustainable development under the backdrop of increasing load requirements. At the same time, the experiment on the study of the calculation delay of the algorithm was studied. The results of the experiment on the study of the calculation delay showed that the parameters of the calculation delay of the algorithm work well.
       Figure 11 shows the comparison of the predicted active power loss and the actual active power loss of the four algorithms at Node 4. As can be seen from Figure 11, the linear fitting degree (\(R^{2}\)) of bPSO-eAF algorithm is 0.9804, while \(R^{2}\) of CNN, RF and BPNN are 0.9527, 0.9612 and 0.9503. This indicates that none of the four models is underfitted. To summarize, it can be concluded that the bPSO-eAF algorithm model has high accuracy and low loss, and is suitable for the reconstruction of distribution network under low-voltage power supply condition.
       The electron motion characteristics are consistent with randomness, and the two-dimensional topology of the entire power system is difficult to reveal. This paper studies the adjustment of the network identification structure in the distribution network and improves the voltage distribution in the process of voltage transmission by monitoring the ring state of the smallest branch of the power grid. The aim of this study is to improve the particle swarm-fish fusion algorithm and apply it to the reconstruction of the 35 kV distributed distribution network. It proposes and uses the bPSO-eAF algorithm to optimize the structure and operation mode of the distribution network. and improve the reliability of the system. The results are obtained and discussed through experiments and simulations.
       This study successfully reconstructed the distribution network. The reconstructed distribution network has higher performance and can quickly respond to power demand and faults, which is very similar to the results obtained by Yang et al. 29 . In the study of system reliability, the power outage probability of the system was the lowest among the four algorithms due to the reasonable adjustment of the line connection method, which significantly improved the fault tolerance of the system. Hu et al. 30 also obtained similar results. For the economic study of the system, the energy consumption and operating cost of the system are lower than those of the CNN, RF and BPNN systems. The study also improves the efficiency of power supply, allowing the restructured distribution network to enhance the resilience of the system. In addition, the study also explored the optimization of power distribution strategies, which effectively reduces the dependence on traditional energy sources, thereby reducing carbon dioxide emissions.
       In summary, the bPSO-eAF algorithm proposed in this study has made revolutionary progress in the research on the reconstruction of distributed power grids below 35 kV. The study and optimization of the structure and operation mode of the distribution network can improve the reliability, economy and stability of the system, providing useful information for future power supply. However, the related technical management problems encountered in the study still need to be solved in practice. In addition, there are still some technical and management problems in accessing and managing distributed energy resources, forecasting and dispatching energy loads. These issues will be gradually solved in future research.
       With the rapid development of distribution networks in low-voltage distributed energy, the research on long-distance power transmission to reduce power loss is becoming increasingly important. This paper is based on the optimized PSO-AF generation fusion algorithm (bPSO-eAF) of mutant birds and elite fish, and studies the efficiency of power supply and relay. Using the proprietary Elecgrid dataset to conduct distribution network reconstruction experiments, it is compared with CNN, RF and BPNN. By testing different experimental parameters, it is found that the stability value of power supply is 0.083, the circuit load is 0.72, and the number of iterations is set to 155. When the circuit consists of 5 objective functions, bPSO-eAF can cure the overload with an activity of 0.94, which is the highest activity among the four algorithms. The voltage values ​​of the four algorithms after reconstruction are 0.29, 0.32, 0.31 and 0.39 V, respectively. It can be seen that the bPSO-eAF algorithm gets the best effect after reconstruction. The bPSO-eAF algorithm has the smallest active loss value of 63 kW under the use of 0.74 V. The active loss values ​​of CNN, BPNN and RF are 74, 97 and 109 kW respectively; Among the 6 experimental results of a large number of tests, bPSO-eAF has the smallest loss. The experimental results show that the bPSO-eAF algorithm can significantly reduce the active power loss in the reconstruction of the distribution network, and it is found that this method is suitable for the reconstruction of long-distance distribution network under low-voltage power supply. The actual result of this study is to effectively reduce the resistance of long lines, thereby reducing the active power loss in power transmission. However, this experiment is carried out in the low voltage environment below 35 kV, ignoring the influence of high voltage on bPSO-eAF. When the voltage at both ends of the circuit is set to high voltage, the bPSO-eAF model isolates it. It becomes an isolated point in the reconstruction of the distribution network, which introduces a large error in the experimental results. In addition, the protective measure of this circuit is the rubber protective sheath. At voltages above 35 kV, the insulation layer is destroyed, which endangers the personal safety of the experimenter. This will be the direction of the next stage of research, and protective measures will be introduced in the future.
       The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.
       Fambri, G., et al. Techno-economic analysis of power plants in natural gas and electricity distribution systems with high penetration of renewable energy sources. Applied Energy 312, 118743–118743 (2022).
       Ahmad F. et al. Taking into account energy losses, land costs and electric vehicle ownership, EV fast charging stations are established in the distribution network. Eco Efficiency of Energy Recycling 44(1), 1693–1709 (2022).
       Kirthana, G., Anandan, P., and Nachimuthu, N. A robust hybrid artificial fish swarm simulating annealing optimization algorithm for protecting free-scale networks from malicious attacks. Computers Mater. 66 (1), 903–917 (2021).
       Ahmadi, P. and Rastegar, H. Optimal allocation of interconnected combined heat and power systems based on minimization of electrical and thermal transfer losses. IET General Transm. 16 (13), 2701–2715 (2022).
       Babanejad, M., Noudeh, S., Abdelaziz, Y., and Kotb, H. Reactive power-based capacitor allocation in distribution networks using mathematical remote sensing optimization algorithms considering operating costs and load conditions. Alex Eng. J. 61(12), 10511–10526 (2022).
       Dashtaki A.A. et al. Optimal control algorithm for distribution network microgrids taking into account the uncertainty of renewable energy systems. International Journal of Energy and Power Systems. 145, 108633–108633 (2023).


Post time: Dec-16-2024