Yang Liang1,a,*, Lei Wei2,b, Zhengning Pang1,c, Guoyu Zhu1,d, Fuquan Zhao1,e
1NARI Technology Development Limited Company, Nanjing 211106, Jiangsu Province, China
2State Grid Jiangsu Electric Power Co., Ltd., Nanjing 211106, Jiangsu Province, China
aEmail: liangyang@sgepri.sgcc.com.cn
bEmail: weilei@js.sgcc.com.cn
cEmail: pangzhengning@sgepri.sgcc.com.cn
dEmail: 992623734@qq.com
eEmail: zhaofuquan@sgepri.sgcc.com.cn
*Corresponding Author: Yang Liang
Email: liangyang@sgepri.sgcc.com.cn
Abstract: To address the challenges of diversified data processing and dynamic load requirements in the adaptive optimization of power grid topology structure in the construction of intelligent power grid, this article studied a graph database-based algorithm for the adaptive construction and optimization of main distribution integrated power grid topology. A unified power grid data model was developed and the graph technology was utilized to process mixed heterogeneous data, achieving efficient integration and updating of power grid data. On this basis, the genetic algorithm (GA) was utilized to evaluate the fitness of the power grid configuration scheme, gradually evolving the optimal power grid topology structure. The real number encoding was used to represent the characteristics of each scheme, and finally, the optimal solution was output through multiple iterations to optimize the power grid configuration parameters. Through simulation experiments, this article simulated the operation of the power grid under different load conditions and fault scenarios. The results showed that after GA optimization, the efficiency of the power grid was significantly improved, with an increase of more than 10% in power grid efficiency compared to distributed models, centralized models, and linear programming models. In addition, it also had more reasonable load distribution, faster fault recovery speed, and higher system stability.
Keywords: Power Grid Topology Adaptive Optimization; Graph Database; Genetic Algorithm; Graph Technology; Smart Grid
- Introduction
As a key infrastructure in modern society, the stability and efficiency of the power grid system have a significant impact on economic development and residents’ quality of life. With the rapid development of smart grid technology, traditional power grid systems are facing increasing challenges, such as dynamic load changes, integration of renewable energy, and complex grid topology. These challenges require the power grid system to have higher intelligence and adaptive capabilities to ensure its stable and reliable operation. Traditional power grid data models usually describe the structure and state of the power grid in a static manner, but they are inadequate in dealing with dynamic changes and real-time data in the power grid. Graph database technology provides new possibilities for modeling and optimizing power grid systems due to its ability to efficiently process complex relational data [1-3]. However, how to combine graph database technology with the dynamic optimization requirements of the power grid system remains an urgent problem to be solved. The current research mainly focuses on static power grid models and traditional optimization algorithms, with relatively little research on real-time data integration and automatic updates, which limits the further development of intelligent power grid systems.
This article aims to propose a graphical technology based on a unified power grid data model to solve the main problems in existing power grid data management and optimization. The core objective of the research is to construct a unified data model that automatically maps heterogeneous data in the power grid system to a graph database, thereby achieving data integration and automatic updates. This method can not only overcome the shortcomings of traditional data models in handling dynamic changes and real-time data, but also enhance the intelligence level of the power grid system. By applying graph computing technology and genetic algorithm (GA), this article aims to achieve dynamic adaptive optimization of power grid topology, and improve load balancing, fault recovery, and overall operational efficiency of the power grid system.
The study introduces the design and implementation of graph technology based on the unified power grid data model, including data mapping mechanism, construction of graph database, application of graph computing technology, and specific implementation of genetic algorithm. In the simulation experiment section, the process and results of data preparation, algorithm implementation, and simulation testing are demonstrated to verify the effectiveness and practicality of the proposed method. In the summary section, the research results are summarized; the limitations of the study are pointed out; future research directions are proposed.
The main contribution of this article is the proposal and implementation of a graphical technology based on the main distribution integrated power grid data model, which can effectively handle heterogeneous data in the power grid system and support real-time data integration and automatic updates. By combining graph databases and GA, the dynamic adaptation problem in power grid topology optimization has been innovatively solved. In addition, it also involves a GA optimization framework applicable to the power grid system, which improves the intelligence level and operational efficiency of the power grid system. This method not only fills the gap in existing research, but also provides new solutions for the optimization and intelligent management of future power grid systems.
- Related Work
The traditional power grid data model mainly focuses on the description of the static power grid structure, such as the topology, equipment parameters, and load distribution of the power grid. However, with the intelligence and dynamic operation of the power grid, traditional data models are gradually exposing their shortcomings. For example, traditional models exhibit certain limitations when dealing with real-time data, heterogeneous data, and dynamic changes in the power grid. Researchers have proposed various improvement schemes [4-6], including dynamic models based on time-series data and comprehensive data models for smart grids [7-9].
Zhang W et al. [10] studied the short-term prediction of offshore wind power generation and proposed a hybrid model that combined Discrete Wavelet Transform (DWT), Seasonal Autoregressive Integral Moving Average Model (SARIMA), and Long Short-Term Memory (LSTM) based on deep learning, successfully handling complex patterns in wind power time series. An D et al. [11] proposed a detection method based on deep reinforcement learning for dynamic state estimation attacks in smart grids. Their research revealed the vulnerability of smart grid systems to network attacks and improved the accuracy of attack detection through LSTM networks. Scholars such as Li Y [12] focused on false data injection attacks in smart grids and proposed a detection method based on secure federated deep learning, which combined Transformer and Paillier cryptographic systems. Although these models can better reflect the operational status and changes of the power grid, there are still challenges in handling cross domain data and data integration.
A graph database is a database system that stores and queries data in a graph structure, characterized by its ability to efficiently process complex relational data. In the power grid system, graph databases can be used to model the topology of the power grid, including substations, generators, transmission lines, etc. [13-15]. Researchers have achieved good results in modeling and analyzing the power grid using graph databases. Meng F et al. [16] proposed a method for constructing a fault knowledge graph of power equipment based on the BERT-BiLSTM-CRF(Bi-directional Encoder Representation from Transformers-Bi-directional Long Short-Term Memory-Conditional Random Field) model. This study extracted power equipment entities and their semantic relationships from Chinese technical literature, constructed a high-precision power equipment fault knowledge graph, and stored the extracted knowledge in the Neo4j database for visualization. To study the topology recognition of low-voltage distribution networks, Zhang H et al. [17] proposed a low-voltage distribution network topology recognition method based on latent tree model. This method utilized end-user data from smart meters and employed Bayesian networks and expectation maximization algorithms to successfully identify potential nodes and their topology in the distribution network. By analyzing the data flow of various measuring equipment in the smart grid and utilizing the performance of graph signals in different domains (such as vertex domain, graph frequency domain, and vertex frequency domain), the network and physical pressure in the system were detected and located. Hasnat M A and other scholars [18] proposed a smart grid security and reliability analysis method based on Graph Signal Processing (GSP) framework. However, existing research mainly focuses on the construction of static power grid models, and further exploration is needed on how to combine graph databases with dynamic optimization problems.
Many scholars have proposed using graph computing techniques to handle complex data analysis and optimization problems. Graph computing involves representing data as graph structures and utilizing graph algorithms to efficiently compute and optimize the data [19-21]. Some optimization problems in the power grid system, such as load balancing, fault recovery, and topology optimization, can be effectively solved through graph computing techniques. To address the issue of Alternating Current Optimal Power Flow (AC-OPF) in real-time communication, Liu S et al. [22] proposed a graph neural network method that combined topological information. By combining graph neural network (GNN) with physical constraints, the power flow calculation of the power grid was optimized. This method not only reduced the complexity of the model, but also improved its adaptability under different topological structures. Falconer T et al. [23] explored the problem of power grid topology utilization in machine learning assisted optimal power flow calculation, systematically comparing the performance of Fully Connected Neural Networks (FCNNs), Convolutional Neural Networks (CNNs), and GNNs under fixed and changing power grid topologies. Research found that although the advantages of GNN were not obvious under fixed topology, it exhibited significant advantages when the topology changed. To study the transient dynamics of online prediction of power systems, Zhao T et al. [24] proposed a structured graph learning framework that improved the accuracy of power grid transient dynamics prediction by combining GNN with spatial-temporal information transfer mechanism. Wu H et al. [25] proposed a Grid Topological Generative Adversarial Network (Gridtopo-GAN) based on generative adversarial networks to address the identification problem of different topology configurations in large-scale systems. This method effectively addressed the identification problem of different topology configurations in large-scale systems by preserving the node embedding architecture. However, how to combine these algorithms with graph databases to achieve real-time data integration and automatic updates is still a problem that needs to be solved.
- Construction of Adaptive Models for the Main Distribution Integrated Power Grid Topology Based on Graph Database
3.1 Unified Power Grid Data Model
In order to achieve standardization and unified management of complex heterogeneous data in the power grid system, this article designs a unified power grid topology data model. Figure 1 shows the framework diagram of a unified power grid topology data model for main distribution integration.
Figure 1. Unified power grid topology data model framework
As illustrated in Figure 1, the model includes basic structural information of the power grid, real-time operating data, equipment status data, etc., covering key elements in the power grid, including substations, generators, transmission lines, node connection relationships, etc. Each element has relevant attribute information, such as capacity, status, geographic location, etc. The model adopts the object-oriented design concept and represents the topology structure in the power grid through the relationships between classes.
In addition, based on a unified data model, this article applies an automated data mapping mechanism [26] that seamlessly maps real data from business systems to graph databases. This mechanism adopts a bidirectional mapping strategy: on the one hand, it extracts and transforms multi-source data from the business system in real-time and maps it to the graph database; on the other hand, when the data in the graph database changes, it automatically synchronizes back to the business system.
3.2 Graph Technology and Mixed Heterogeneous Data Processing
The various equipment, lines, nodes, and other information in the power grid are converted into graph structured data, and the powerful functions of graph databases are utilized to achieve efficient management of complex relationships in the power grid system [27-28]. Graph technology intuitively represents the topology structure of the power grid and supports complex query and analysis operations [29-30]. Figure 2 shows the flowchart of graph technology and mixed heterogeneous data processing.
Figure 2. Flowchart of graph technology and mixed heterogeneous data processing
As shown in Figure 2, a mixed heterogeneous data processing method based on graph technology converts these heterogeneous data into unified graph structured data. Through data cleaning and transformation techniques, noise and redundant information in heterogeneous data are eliminated; secondly, through the efficient indexing and querying functions of the graph database, fast access and processing of these data can be achieved; finally, combined with graph computing technology, the operational status of the power grid system is analyzed and optimized in depth.
3.3 GA Principle
This article uses GA to optimize the topology of the power grid, with the main goal of improving the load balancing ability of the power grid, shortening the fault recovery time, and achieving dynamic topology optimization of the power grid. The basic process of GA includes individual initialization, fitness calculation, selection, crossover, mutation, and update operations, and ultimately obtains optimization results through multiple iterations [31-33]. Applying GA to power grid topology optimization problems, the optimal power grid topology structure is gradually evolved by evaluating the fitness of individuals (power grid configuration schemes). Taking into account multiple indicators such as load balancing, fault recovery capability, and system stability of the power grid system, this article designs a comprehensive fitness function , specifically:
(1)
Among them, , , and are the load balance function, fault recovery capability function, and system stability function, respectively, while , , and are the corresponding weights. , and are shown in Formula (2-3):
(2)
(3)
Among them, N is the total number of nodes; is the load of the i-th node; is the average value of the load; is the fault recovery time.
The system stability function is a comprehensive indicator that combines oscillation amplitude, system frequency response, and system transient stability. The formula is:
(4)
Among them, represents the amplitude of oscillation; is the system response time function; t is time; is the value of system response time; represents the system frequency response; T is the time window; is the frequency response of the system; is the mean frequency response; represents the transient stability of the system; represents the transient response of the system; is the mean of transient response. , , and represent the weights of oscillation amplitude, system frequency response, and system transient stability, respectively.
3.4 Topology Adaptive Modeling and Optimization of Main Distribution Integrated Power Grid
With the support of graph databases, this article constructs a dynamically adjustable main distribution integrated power grid topology adaptive model, as illustrated in Figure 3.
Figure 3. Topology adaptive model and optimization of main distribution integrated power grid
As illustrated in Figure 3, the topology adaptive model of the power grid can autonomously adjust the topology structure of the power grid based on real-time data and environmental changes to optimize the operational efficiency and stability of the power grid. Specifically, graph databases are utilized to transform information such as nodes and lines in the power grid into graph structures. The attributes of nodes and edges are set to reflect the actual load, capacity, and connection relationships in the power grid, and mapped to entities in the graph database. These entities are adaptively adjusted through GA optimization algorithms to automatically select the optimal topology structure and achieve dynamic optimization of the power grid, using real number encoding to represent the characteristics of each power grid configuration scheme. The optimization objectives include load balancing, fault recovery speed, and system stability. In addition, the model verifies and challenges the constructed topology adaptive model of the power grid by monitoring the operating status of the power grid in real-time, in order to cope with unexpected events such as load changes and equipment failures.
3.5 Selection of Optimization Objectives and Model Indicators for Power Grid
The optimization objectives of the power grid include three aspects: load balancing, fault recovery speed, and system stability. The goal of load balancing is to avoid local overload by allocating loads reasonably in the power grid; the goal of fault recovery is to focus on the ability of the power grid to quickly resume operation after equipment failure occurs; the goal of system stability is to ensure that the power grid maintains a stable operating state under dynamic conditions.
This article selects the load difference coefficient , fault recovery time , system stability function , and comprehensive fitness function as the evaluation criteria for optimizing the power grid. Among them, the load difference coefficient is:
(5)
In addition, three indicators, namely the power grid reliability index , power grid efficiency index , and power grid response flexibility , are chosen as the criteria for evaluating the power grid topology adaptive model. Among them, the reliability index of the power grid is:
(6)
is the average failure rate of the power grid, and is the average recovery rate of power grid failures.
is the ratio of the total power supply to the total input power , and the formula is:
(7)
The formula for power grid response flexibility is:
(8)
is the load response function, and is the average value of load response.
- Simulation Experiments
4.1 Data Preparation
This simulation experiment selects multiple key element datasets as the simulation basis, including substation configuration parameters, transmission line status information, real-time load, and fault history data. Table 1 shows the specific content of some data:
Table 1. Part of data content
| Substation Configuration Parameters | ||||||
| Substation ID | Geographical Location | Transformer Capacity (MVA) | Operating Status | |||
| BS001 | Chaoyang District, Beijing | 150 | Normal | |||
| BS002 | Pudong New Area, Shanghai | 200 | Normal | |||
| BS003 | Tianhe District, Guangzhou | 100 | Under Maintenance | |||
| … | … | … | … | |||
| Transmission Line Status Information | ||||||
| Line ID | Starting Substation ID | Ending Substation ID | Length (km) | Impedance (Ω) | Power Transmission Capacity (MW) | Operating Status |
| TL001 | BS001 | BS002 | 50 | 0.5 | 100 | Normal |
| TL002 | BS002 | BS003 | 60 | 0.6 | 120 | Normal |
| TL003 | BS003 | BS004 | 70 | 0.7 | 140 | Under Maintenance |
| … | … | … | … | … | … | … |
| Real-time Load Data | ||||||
| Node ID | Time | Load (MW) | ||||
| Node01 | 2024/8/19 8:00 | 50 | ||||
| Node02 | 2024/8/19 8:00 | 60 | ||||
| Node03 | 2024/8/19 8:00 | 70 | ||||
| … | … | … | ||||
| Fault History Records | ||||||
| Fault ID | Fault Recovery Time | Fault Location | Fault Type | |||
| F001 | 1h30min | BS003 | Short Circuit | |||
| F002 | 1h45min | TL005 | Overload | |||
| F003 | 2h | BS007 | Short Circuit | |||
| F004 | 2h | TL009 | Equipment Failure | |||
| … | … | … | … | |||
As displayed in Table 1, the configuration parameters of a substation are an important component of the power grid topology, including the geographical location of the substation, transformer capacity, equipment type, operating status, and other information. The status information of transmission lines includes the length, impedance, power transmission capacity, and historical operation records of the transmission lines. Real-time load data reflects the load situation of each node at different time periods. The fault history records data including the location, type, and recovery time of the fault occurrence.
The collected data is subjected to data cleaning and preprocessing. For potential missing data in substation configuration parameters or transmission line status information, interpolation or historical data backfilling is used to complete the data, ensuring its integrity. Abnormal data points are detected and eliminated through statistical analysis methods and machine learning algorithms. In terms of data integration, mapping, and layering, data sources from different systems are uniformly integrated to form a structurally complete and time aligned power grid dataset. After data integration is completed, the data is mapped to the graph database. During the mapping process, an automated data mapping mechanism is used to layer and process data based on the physical structure and functional requirements of the power grid, dividing it into a basic layer, an operational layer, and a historical layer.
To ensure the efficiency and accuracy of the simulation experiment, a high-performance computing platform is selected for the experimental environment configuration, equipped with multi-core CPUs, high memory capacity, and GPU accelerators. In terms of software environment, Python-based data processing toolkits (such as Pandas, NumPy, etc.) are used and seamlessly integrated with graph databases (Neo4j) to manage and query complex power grid topology information. In addition, the experiment is configured with a GA implementation framework that supports parallel computing, enabling the algorithm to complete multiple iterations of optimization within a reasonable time.
4.2 GA Implementation
In the adaptive optimization of power grid topology, GA, as a global optimization method, can gradually optimize the topology structure of the power grid by simulating natural selection and genetic mechanisms. The power grid configuration scheme is encoded as an individual, using binary or real number encoding to represent the characteristics of each scheme. In addition, the randomly generated initial population contains multiple power grid configuration schemes, and the diversity of the population helps to enhance the optimization effect of GA. GA reproduces superior individuals through selection, crossover, and mutation operations. Crossover operations generate new offspring individuals, while mutation operations increase population diversity and avoid falling into local optima. After multiple iterations, these operations gradually evolve into a more optimal power grid topology structure. The iterative process stops when the fitness function value no longer significantly improves or reaches the preset number of iterations, and finally outputs the optimal solution. Figure 4 displays the changes of fitness function values with the iteration process.
Figure 4. Changes of fitness values with iteration times
Figure 4 (a): Changes in fitness values within 20 iterations
Figure 4 (b): Changes in fitness values within 50 iterations
Figure 4 illustrates the changes of fitness values with the number of iterations. Figure 4 (a) shows the changes in fitness values over the course of 20 iterations. It can be seen that in the first few iterations, the fitness value shows an upward trend. However, in the later iterations, the fitness value fluctuates and remains on an upward trend, indicating that the algorithm has found a relatively good solution but has not reached the optimal state. Figure 4 (b) depicts the changes in fitness values over 50 iterations. It can be seen that although the fitness value fluctuates, it is still on an upward trend overall. In the later stage of iteration, the change in fitness value is gentle. As the number of iterations increases, the fitness function value gradually stabilizes, indicating that the algorithm has reached its optimal state.
4.3 Simulation Testing
By loading experimental data into the graph database, this article simulates the operation of the power grid under different load conditions and fault scenarios. Table 2 and Figure 5 show the specific operation of the power grid in different complex scenarios.
Table 2. Overview of different simulation environments
| Scenario | Load Condition | Fault Type | Dynamic Changes |
| Scenario 1 | High | Short Circuit | Oscillatory |
| Scenario 2 | Medium | Overload | Stable |
| Scenario 3 | Low | Equipment Failure | Unstable |
| Scenario 4 | High | Short Circuit | Oscillatory |
| Scenario 5 | Medium | Overload | Stable |
Figure 5. Performance of the topology adaptive model of the main distribution integrated power grid in different scenarios
Figure 5 (a): Load coefficient performance of the model under different scenarios
Figure 5 (b): Performance of the model’s fault recovery time under different scenarios
Figure 5 (c): System stability index performance of the model under different scenarios
Figure 5 (d): Performance of model fitness values in different contexts
From Table 2, it can be seen that the experiment simulates different load conditions and fault types, including high load, medium load, and low load, as well as situations such as short circuit, overload, and equipment failure. In addition, dynamic factors such as oscillation and stability differences are also taken into account. Figure 5 shows the changes in load factor, fault recovery time, system stability, and fitness values under different scenarios. Figure 5 (a) shows the changes of load factor in different scenarios. The load factor is an indicator that measures the capacity of the power grid to withstand, and the higher the value, the greater the load that the power grid can bear. From Figure 5 (a), it can be seen that the load coefficients for scenarios 3 and 5 are relatively high, while the load coefficients for scenarios 1 and 4 are relatively low. Figure 5 (b) shows the fault recovery time for each scenario. The fault recovery time refers to the time required for the power grid to resume normal operation after a fault occurs. The shorter fault recovery time in scenarios 1 and 4 indicates that they can recover faster when the power grid encounters a fault. The fault recovery time for scenario 3 is relatively long, as the model faces greater difficulty in recovering in situations of dynamic instability and equipment failure. Figure 5 (c) shows the changes in system stability index under different scenarios. The system stability index is used to measure the ability of the power grid to recover and resist interference after receiving disturbances. From Figure 5 (c), it can be learned that the system stability index of scenario 1 is relatively high, indicating that the model can quickly recover to a stable state in the face of high load conditions, oscillations, and paragraphs. The system stability index of scenario 2 is relatively low, indicating that the model is susceptible to interference in overload situations. Figure 5 (d) shows the changes in fitness values for each scenario. Adaptability is a comprehensive data. It can be clearly learned from Figure 5 (d) that the fitness values of the model in all scenarios are 0.8 or above, which is at a good level, indicating that GA has achieved good results in optimizing the power grid topology structure. Based on Table 2 and Figure 5, it can be concluded that GA has a good effect on optimizing the topology structure of the power grid, with good performance in load distribution, fault recovery, and system stability.
Figure 6 shows the power grid configuration before and after GA optimization.
Figure 6. Comparison of various data before and after GA optimization
Figure 6 (a): Power grid configuration before and after GA optimization
Figure 6 (b): Performance indicators of the model before and after GA optimization
From Figure 6 (a), the power grid configuration before and after GA optimization can be seen, demonstrating the improvement effect after applying GA algorithm optimization. The blue line in Figure 6 (a) represents the basic state of the power grid without any optimization operation, including the initial configuration parameters of key facilities such as substations and transmission lines. The red line represents the power grid configuration after GA optimization. It can be observed that after GA optimization, the transformer capacity increases from 150MVA to 180MVA; the route length increases from 50 kilometers to 55 kilometers; the system load increases from 70MV to 75MV. The increase in capacity directly enhances the load carrying capacity of the power grid, and the increase in route length reflects the adjustment of the topology structure of the power grid during the optimization process. The growth of routes has improved power transmission paths, reduced route losses, and covered a wider range of power supply areas, enabling electricity to be distributed more evenly to load centers. The increase in system load indicates that the optimized power grid system can handle larger power demands, thereby withstanding higher load capacity without sacrificing stability. Figure 6 (b) shows the correlation between the reliability index, efficiency index, and response flexibility of the model before and after GA optimization. It can be seen that the reliability index increases from around 0.7 to around 0.9 before and after GA optimization, and the efficiency index also shows a significant improvement, but the magnitude is not as large as the reliability index. Among them, the change in response flexibility is the largest, increasing from 0.66 to nearly 0.9. From the overall perspective of Figure 6 (b), GA optimization significantly improves the reliability index, efficiency index, and response flexibility of the model, indicating that the optimized main distribution integrated power grid topology adaptive model has stronger adaptability and flexibility in the face of uncertainty and dynamic changes. From the comprehensive analysis of Figures 6 (a) and (b), it can be learned that the optimization of GA is crucial for enhancing the power supply capacity of the power grid, reducing operational risks, improving overall system stability, and enhancing the ability to respond to emergencies.
Figure 7 shows the performance of different algorithms in optimizing the topology adaptive model of the main distribution integrated power grid.
Figure 7. Optimization performance of five different algorithms in the power grid topology adaptive model
Figure 7 (a): Optimization efficiency of five different algorithms
Figure 7 (b): Solution stability of five different algorithms
Figure 7 (c): Convergence speed of five different algorithms
Figure 7 selects Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC), Simulated Annealing (SA), and Genetic Algorithm-Ant Colony Optimization (GA-ACO). Figure 7 (a) shows the optimization efficiency performance of five different algorithms, which are ranked from high to low as GA, GA-ACO Hybrid, PSO, ABC, and SA, respectively. Figure 7 (b) shows the stability performance of five different algorithms, in descending order: GA, PSO, GA-ACO Hybrid, ABC, and SA. Figure 7 (c) shows the convergence speed performance of five algorithms, from highest to lowest, namely GA, SA, PSO, ABC, and GA-ACO Hybrid. Through comparative analysis, it can be learned that GA performs excellently in optimization efficiency, solution stability, and convergence speed, making it particularly suitable for adaptive optimization tasks in power grid topology. Next is the GA-ACO hybrid algorithm, which performs poorly in convergence speed, but its performance in optimization efficiency and stability is close to GA.
In addition, Table 3 covers the performance of different models in terms of reliability, efficiency, and response flexibility.
Table 3. Evaluation of different models
| Model | Power grid reliability index | Power grid efficiency index | Power grid response flexibility |
| Topology adaptive model of main distribution integrated power grid based on graph database | 0.95 | 0.93 | 0.89 |
| Rule-based expert system model | 0.85 | 0.88 | 0.7 |
| Mixed integer linear programming model | 0.88 | 0.86 | 0.78 |
| Centralized control model | 0.8 | 0.82 | 0.72 |
| Distributed control model | 0.75 | 0.79 | 0.65 |
| Linear programming model | 0.82 | 0.84 | 0.73 |
From the data in Table 3, it can be learned that in terms of power grid reliability, the topology adaptive model of the main distribution integrated power grid based on graph database performs the best at 0.95, followed by the mixed integer linear programming model and the rule-based expert system model. The centralized control model and distributed control model perform relatively weakly, while the linear programming model performs moderately. In terms of power grid efficiency index, the topology adaptive model of the main distribution integrated power grid based on graph database still performs the best in power grid efficiency, at 0.93, indicating that it can have higher efficiency and lower losses in resource allocation and energy distribution. Next are rule-based expert system models and mixed integer linear programming models, with scores of 0.88 and 0.86, respectively. The centralized control model and distributed control model perform poorly. In terms of flexibility in power grid response, the topology adaptive model of the main distribution integrated power grid based on graph database also performs the best, at 0.89, which reflects that the topology adaptive model of the main distribution integrated power grid based on graph database can adapt to demand changes and emergencies more quickly. The performance of the mixed integer linear programming model is slightly inferior, at 0.78, with the distributed control model performing the worst. Based on Table 3, the topology adaptive model of the main distribution integrated power grid based on graph database performs well in all indicators. In comparison, the other five models have their own advantages and disadvantages in performance, but the overall performance is still not as good as that of the topology adaptive model of the main distribution integrated power grid. Especially in dealing with the complexity and dynamic response of the power grid, the other five models show significant shortcomings.
Table 4 displays the comparison of the main distribution integrated power grid topology adaptive model based on graph database with other models in terms of power grid reliability, efficiency, and response flexibility.
Table 4. Comparison data with other models
| Comparative Model | Power grid reliability index | Power grid efficiency index | Power grid response flexibility |
| Rule-based expert system model | 0.12 | 0.06 | 0.27 |
| Mixed integer linear programming model | 0.08 | 0.08 | 0.14 |
| Centralized control model | 0.19 | 0.13 | 0.24 |
| Distributed control model | 0.27 | 0.18 | 0.37 |
| Linear programming model | 0.16 | 0.11 | 0.22 |
Table 4 shows the comparison between the topology adaptive model of the main distribution integrated power grid based on graph database and other models in terms of power grid reliability, efficiency, and response flexibility. Positive values indicate growth, while negative values indicate decrease. From Table 4, it can be seen that the topology adaptive model of the main distribution integrated power grid based on graph database has increased power grid reliability by 27%, power grid efficiency by 18%, and response flexibility by 37% compared to the distributed model. In addition, it can be observed that all values in the comparison data are positive, indicating that the topology adaptive model of the main distribution integrated power grid based on the graph database is superior to other models in all aspects.
4.4 Experimental Discussion
In the simulation experiment, a detailed analysis is conducted on the power grid operation under different simulation scenarios, the power grid configuration before and after GA optimization, and the performance of different algorithms.
From the results in Table 2 and Figure 5, it can be learned that different load lines and fault types have a significant impact on the operational performance of the power grid. However, the fitness values of the GA optimized model remain at a high level in all scenarios, indicating that GA has significant advantages in optimizing the topology structure of the power grid. Figure 6 shows the changes in power grid configuration and performance indicators before and after GA optimization. The optimized power grid configuration shows an increase in transformer capacity, line length, and system load, indicating that the optimized power grid has stronger load carrying capacity and a wider power supply range. Figure 6 (b) further illustrates the changes in the reliability index, efficiency index, and response flexibility of the power grid after GA optimization. Among them, the most significant improvement in response flexibility indicates that the GA optimized power grid can flexibly respond to dynamic changes and emergencies. In addition, the performance of GA and four other optimization algorithms in the power grid topology adaptive model is compared through Figure 7. From the results, GA performs well in optimization efficiency, solution stability, and convergence speed. Finally, the topology adaptive model of the main distribution integrated power grid based on graph database is compared with other power grid topology models to further verify the effectiveness of GA optimization.
In summary, the experimental results indicate that GA has significant advantages in optimizing the topology structure of the power grid, which can significantly improve the reliability, efficiency, and response flexibility of the power grid. It also has more reasonable load allocation, faster fault recovery speed, and higher system stability.
- Conclusions
This article discussed and analyzed the topology adaptive modeling and optimization algorithm of the main distribution integrated power grid based on graph database. Firstly, a unified power grid data model was constructed, and a scheme for processing mixed heterogeneous data using graph technology was proposed. The adaptive modeling and optimization of power grid topology were achieved through GA. In the simulation experiment section, the performance of the power grid under different load conditions and fault conditions was tested, further verifying the effectiveness of GA optimization and the superiority of the graph database model. Finally, by comparing with other power grid models and excellent algorithms, it was demonstrated that the GA optimized graph database-based topology adaptive model had significant advantages in optimization efficiency, solution stability, and convergence speed. However, there are still shortcomings in this article, and there is room for further discussion on the experimental results. There is less exploration on the combination of GA and other hybrid algorithms. In future research, algorithm mixing, implementation optimization, and broader application scenarios can be explored to further enhance the intelligence and adaptability of the power grid system, providing more comprehensive technical support for the efficient and safe operation of the power grid.
Funding
This paper is sponsored by a key project of National Key R&D Program of China (2022YFB2404200).
References
[1] Srivastava S, Singh A K. Fraud detection in the distributed graph database. Cluster Computing, 2023, 26(1): 515-537.
[2] Park S, Cheng T. Framework for constructing multimodal transport networks and routing using a graph database: A case study in London. Transactions in GIS, 2023, 27(5): 1391-1417.
[3] Matter H, Buning C, Stefanescu D D, et al. Using Graph Databases to Investigate Trends in Structure–Activity Relationship Networks. Journal of chemical information and modeling, 2020, 60(12): 6120-6134.
[4] Li D, Gong Y. The design of power grid data management system based on blockchain technology and construction of system security evaluation model. Energy Reports, 2022, 8: 466-479.
[5] Ramakrishna R, Scaglione A. Grid-graph signal processing (grid-GSP): A graph signal processing framework for the power grid. IEEE Transactions on Signal Processing, 2021, 69: 2725-2739.
[6] Zhou M, Yan J, Zhou X. Real-time online analysis of power grid. CSEE Journal of Power and Energy Systems, 2020, 6(1): 236-238.
[7] Jiang Z, Lv H, Li Y, et al. A novel application architecture of digital twin in smart grid. Journal of Ambient Intelligence and Humanized Computing, 2022, 13(8): 3819-3835.
[8] Du L, Gao R, Suganthan P N, et al. Bayesian optimization based dynamic ensemble for time series forecasting. Information Sciences, 2022, 591: 155-175.
[9] Jafari M, Kavousi-Fard A, Chen T, et al. A review on digital twin technology in smart grid, transportation system and smart city: Challenges and future. IEEE Access, 2023, 11: 17471-17484.
[10] Zhang W, Lin Z, Liu X. Short-term offshore wind power forecasting-A hybrid model based on Discrete Wavelet Transform (DWT), Seasonal Autoregressive Integrated Moving Average (SARIMA), and deep-learning-based Long Short-Term Memory (LSTM). Renewable Energy, 2022, 185: 611-628.
[11] An D, Zhang F, Yang Q, et al. Data integrity attack in dynamic state estimation of smart grid: Attack model and countermeasures. IEEE Transactions on Automation Science and Engineering, 2022, 19(3): 1631-1644.
[12] Li Y, Wei X, Li Y, et al. Detection of false data injection attacks in smart grid: A secure federated deep learning approach. IEEE Transactions on Smart Grid, 2022, 13(6): 4862-4872.
[13] Besta M, Gerstenberger R, Peter E, et al. Demystifying graph databases: Analysis and taxonomy of data organization, system designs, and graph queries. ACM Computing Surveys, 2023, 56(2): 1-40.
[14] Wang J, Wang X, Ma C, et al. A survey on the development status and application prospects of knowledge graph in smart grids. IET Generation, Transmission & Distribution, 2021, 15(3): 383-407.
[15] Wang Z, Xia M, Lu M, et al. Parameter identification in power transmission systems based on graph convolution network. IEEE Transactions on Power Delivery, 2021, 37(4): 3155-3163.
[16] Meng F, Yang S, Wang J, et al. Creating knowledge graph of electric power equipment faults based on BERT–BiLSTM–CRF model. Journal of Electrical Engineering & Technology, 2022, 17(4): 2507-2516.
[17] Zhang H, Zhao J, Wang X, et al. Low-voltage distribution grid topology identification with latent tree model. IEEE Transactions on Smart Grid, 2022, 13(3): 2158-2169.
[18] Hasnat M A, Rahnamay-Naeini M. A graph signal processing framework for detecting and locating cyber and physical stresses in smart grids. IEEE Transactions on Smart Grid, 2022, 13(5): 3688-3699.
[19] Hossain R R, Huang Q, Huang R. Graph convolutional network-based topology embedded deep reinforcement learning for voltage stability control. IEEE transactions on power systems, 2021, 36(5): 4848-4851.
[20] Liao W, Bak-Jensen B, Pillai J R, et al. A review of graph neural networks and their applications in power systems. Journal of Modern Power Systems and Clean Energy, 2021, 10(2): 345-360.
[21] Zhao T, Wang J. Learning sequential distribution system restoration via graph-reinforcement learning. IEEE Transactions on Power Systems, 2021, 37(2): 1601-1611.
[22] Liu S, Wu C, Zhu H. Topology-aware graph neural networks for learning feasible and adaptive AC-OPF solutions. IEEE Transactions on Power Systems, 2022, 38(6): 5660-5670.
[23] Falconer T, Mones L. Leveraging power grid topology in machine learning assisted optimal power flow. IEEE Transactions on Power Systems, 2022, 38(3): 2234-2246.
[24] Zhao T, Yue M, Wang J. Structure-informed graph learning of networked dependencies for online prediction of power system transient dynamics. IEEE Transactions on Power Systems, 2022, 37(6): 4885-4895.
[25] Wu H, Xu Z, Zhao J, et al. Gridtopo-GAN for distribution system topology identification. IEEE Transactions on Industrial Informatics, 2022, 19(4): 5356-5366.
[26] Zhen Z, Liu J, Zhang Z, et al. Deep learning based surface irradiance map** model for solar PV power forecasting using sky image. IEEE Transactions on Industry Applications, 2020, 56(4): 3385-3396.
[27] Presekal A, Stefanov A, Rajkumar V S, et al. Attack graph model for cyber-physical power systems using hybrid deep learning. IEEE Transactions on Smart Grid, 2023, 14(5): 4007-4020.
[28] Simeunovic J, Schubnel B, Alet P J, et al. Spatio-temporal graph neural networks for multi-site PV power forecasting. IEEE Transactions on Sustainable Energy, 2021, 13(2): 1210-1220.
[29] Farivar G G, Manalastas W, Tafti H D, et al. Grid-connected energy storage systems: State-of-the-art and emerging technologies. Proceedings of the IEEE, 2022, 111(4): 397-420.
[30] Hashimoto J, Ustun T S, Suzuki M, et al. Advanced grid integration test platform for increased distributed renewable energy penetration in smart grids. Ieee Access, 2021, 9: 34040-34053.
[31] Mayer M J, Grof G. Techno-economic optimization of grid-connected, ground-mounted photovoltaic power plants by genetic algorithm based on a comprehensive mathematical model. Solar Energy, 2020, 202: 210-226.
[32] Abadlia I, Hassaine L, Beddar A, et al. Adaptive fuzzy control with an optimization by using genetic algorithms for grid connected a hybrid photovoltaic–hydrogen generation system. International journal of hydrogen energy, 2020, 45(43): 22589-22599.
[33] Yusran Y, Rahman Y A, Gunadin I C, et al. Mesh grid power quality enhancement with synchronous distributed generation: optimal allocation planning using breeder genetic algorithm. Przegląd Elektrotechniczny, 2020, 96(1): 82-86.