Research on Distribution-Cluster-Integrated Layered Regulation Architecture and Business Control System for New Distribution Grids

With the current rapid development of information technology, but also brought the depth of intelligent development, and smart grid mode in the distribution network regulation and control of the integration of the construction of the implementation, will be a good solution to the traditional distribution network scheduling mode of the prominent manpower regulation of the way insufficient evaluation [1-3], decentralized management mode is unreasonable and prominent “blind” ask phenomenon, not only to ensure the safe and stable operation of distribution network equipment, to ensure the reliability of power supply, but also to optimize the fault handling process in the distribution network, improve the grid fault response ability, to protect people’s service to the power supply enterprises. It not only ensures the safe and stable operation of distribution network equipment and the reliability of power supply, but also optimizes the fault handling process in the distribution network, improves the responsiveness to grid faults, and guarantees people’s satisfaction with the service level of power supply enterprises [4-6].

The integrated mode of grid regulation and management changes the traditional grid regulation and management mode, and realizes the improvement of the safety coefficient of the grid system and the high efficiency operation. This management mode is mainly reflected in the fact that the grid monitoring system is transferred to the dispatching department, but the authority belongs to the grid production department [7-9]. The dispatching department can take corresponding measures in real time through the monitored information to stabilize the early warning faults of the power grid operation and enhance the effective control of the power grid. On the basis of ensuring the stability of the grid operation, for its updated technology and equipment support, so as to achieve the scientific use of grid regulation and control of the integration of operation and management mode, prompting the improvement of its management level, so that the grid management presents automation, intelligent features, reduce the cost of power units to the grid construction investment, which is the new power enterprise reform trend [10-13]. Grid regulation and control integration operation and management mode can make the staff operation efficiency, operation error probability is reduced. Through the integration of regulation and control equipment to achieve real-time monitoring of the power grid, fully grasp the status of power grid operation, found that the problem can be timely overhaul, thus improving staff efficiency and quality, reducing the failure rate of power grid operation [14-17].

In this paper, the cluster division of distribution network based on improved k-means algorithm is carried out for distribution network, and the optimal scheduling model is constructed. On this basis, the optimized regulation strategy of energy storage cluster for distribution network is proposed, and at the level of regulation within the energy storage cluster, the power allocation strategy of energy storage cluster is proposed, and the hierarchical regulation architecture of energy storage cluster is established. According to the needs of distribution network engineering practice, a cluster-integrated layered control system is established based on integrated grid operation. The study conducted an experimental analysis of the optimization scheduling effect and layered regulation effect of distribution network cluster division respectively, so as to verify the feasibility of this paper’s method in practical application.

2

Integrated hierarchical regulation and control system for new distribution networks

2.1

Optimal scheduling model for distribution networks based on cluster partitioning

The topology of the grid directly affects power transmission and dispatch management, so it is crucial to consider the electrical connections between nodes when partitioning the grid. Electrical distance, a key criterion for grid partitioning in recent years, measures the strength of electrical connections between nodes, i.e., the degree of electrical coupling between them. Cluster management in power grids is based on a model of “autonomy within clusters and cooperation between clusters”. This means that there is a strong electrical coupling between nodes within a cluster, while the coupling between different clusters is relatively weak. The definition of electrical distance allows us to accurately characterize the connection of electrical parameters between nodes and thus effectively evaluate the strength of electrical coupling between them.

In order to reflect the role of distributed power cluster division in system operation and control scenarios, power sensitivity is proposed as one of the cluster division indexes, and the tightness of the connection between each distributed power node in the distribution network can be characterized by the electrical distance indexes that are improved to active voltage sensitivity matrix and reactive voltage sensitivity matrix: (1) $D_{i j} = D_{j i} = \sqrt{{(d_{i 1} - d_{j 1})}^{2} + {(d_{i 2} - d_{j 2})}^{2} + \dots + {(d_{i n} - d_{j n})}^{2}}$ (2) $d_{i j} = S_{P i i} - S_{P i j} + S_{Q i i} - S_{Q i j}$

Where S_pi matrix element indicates the degree of influence on the voltage magnitude of node i when a change in active power occurs at node i, and similarly the definitions of S_pj, S_Qi, S_Qj. The solution process for S_pii, S_pij, S_Qii,S_Qij is as follows: (3) $[\begin{matrix} Δ P \\ Δ Q \end{matrix}] = [\begin{array}{l} H & N \\ J & L \end{array}] [\begin{matrix} Δ θ \\ Δ U \end{matrix}]$

where ΔP and ΔQ are the changes in active power and reactive power, respectively, H is the sensitivity matrix of phase angle to changes in active power, N is the sensitivity matrix of voltage magnitude to changes in active power. J is the sensitivity matrix of phase angle to reactive power changes, and L is the sensitivity matrix of voltage magnitude to reactive power changes. Δθ and ΔU represent the variation of phase angle and voltage magnitude at each node in the system, respectively.

The expansion is obtained: (4) ${\begin{array}{l} Δ P = H Δ θ + N Δ U \\ Δ Q = J Δ θ + L Δ U \end{array}$

Eliminating Δθ gives: (5) $Δ U = {(N - H J^{- 1} L)}^{- 1} Δ P + {(L - J H^{- 1} N)}^{- 1} Δ Q = S_{P} Δ P + S_{Q} Δ Q$

Where S_P is the active voltage sensitivity matrix containing SP_ii and SP_ij, and S_Q is the reactive voltage sensitivity matrix containing SQ_ii and SQ_ij. It can be seen that the effect of active power changes on voltage can be ignored in the distribution network, so the active voltage sensitivity and reactive voltage sensitivity between each distributed power node are considered simultaneously in the calculation of electrical distance.

Based on the modularity function, unlike the traditional community delineation technique, the strength of the community structure in the network can be quantified without determining the number of communities in advance, which can solve the problem of complex networks in community delineation. In the modularity function, a high value of modularity indicates a strong intra-community connection, while a low value of modularity indicates a weak inter-community connection. The improved modularity function with electrical distance as the weight characterizes the degree of electrical coupling between nodes in the distribution network, and the distributed power nodes in the distribution network are divided into clusters, and the evaluation index modularity of the cluster division results is calculated as follows: (6) $ρ = \frac{1}{2 m} \sum_{i} \sum_{j} (e_{j} - \frac{k_{i} k_{j}}{2 m}) ϕ (i, j)$

ρ is the weighted modularity of the system after dividing the DGC, which is a normalized index with values from 0 to 1. The closer ρ is to 1, it represents that the nodes at each cluster point are more closely connected, which characterizes a better division result, and the closer ρ is to 0, it represents that the nodes at each cluster point are more loosely connected, which characterizes an undesirable division result; m is the sum of the edge weights of all the nodes in the node system; e_ij is the weight of the edges between nodes i and j, which can be improved to obtain different meanings. which can be improved to obtain the modularity with different meanings, and the modularity of DGC division under electrical distance can be obtained by adopting e_ij shown in Eq. (7); k_i and k_j are the sum of the weights of the edges connected to nodes i and j, respectively; ϕ(i,j) characterizes the nodes i and j when they are in the same DGC ϕ(i,j)=1, otherwise ϕ(i,j)=0.

(7)

e_{i j} = 1 - D_{i j} / \max (D)

The net power of the cluster can be obtained by summing the net power of all the nodes contained in the cluster, and the net power of all the clusters in the system can be obtained by summing the net power normalization process to obtain the cluster net power complementary index of the system. It reflects the net power balance metric of the system after dividing the clusters under the current DG access level, and its expression is: (8) $φ_{P} = \frac{1}{N_{c}} \sum_{c = 1}^{N_{c}} [1 - \frac{1}{T_{c}} \sum_{k = 1}^{τ} P^{c} (t) / \max (P^{c} (t))]$

T_c for dividing the scenario duration period into P^c(t) for the net power of the cluster c at the t time point: (9) $P^{c} (t) = \sum_{i \in Ω} P_{i} (t)$

where P_i(t) is the net power of node i within cluster c; Ω_c is the set of nodes contained within cluster c.

The cluster net power balance index is a key indicator of how well the renewable energy output within a cluster matches the load demand. Clusters with low net power balance indicators require more inter-cluster energy regulation, which increases the difficulty of system optimization and scheduling and energy loss, while a higher degree of net power balance within the cluster indicates that the accessed renewable energy sources have already met the load demand to a certain extent, which effectively reduces the inter-cluster energy interactions and is conducive to the system optimization and scheduling to reduce the operating costs.

In view of the problem that the cluster net power balance index is limited to the optimization algorithm, a method based on the improved modularity index to characterize the degree of node net power complementarity is introduced.

The mathematical expression of the edge power of the improved node net power complementarity can be expressed as follows: (10) ${e^{'}}_{i j} = 1 - \frac{| \sum_{t} P_{i} (t) + \sum_{t} P_{j} (t) |}{\max | \sum_{t} P_{i} (t) + \sum_{t} P_{j} (t) |}$

${e^{'}}_{i j}$ is the improved node network edge power, calculated from the scalar value of the absolute value of the sum of the net power of node i and node j. Reflecting the net power complementary modularity metric between nodes ρ′ is: (11) $ρ^{'} = \frac{1}{2 m^{'}} \sum_{i j} ({e^{'}}_{i} - \frac{{k^{'}}_{i} {k^{'}}_{j}}{2 m^{'}}) ϕ (i, j)$

Where ${k^{'}}_{i}$ and ${k^{'}}_{j}$ are the sum of the weights of the edges connected to nodes i and j under the node net power complementary index, respectively; and m′ is the sum of the edge weights of all nodes under the node net power complementary index.

In order to achieve better results in optimal scheduling and economic operation for cluster division, the degree of cluster net power balance is considered, but the cluster net power balance index based on the overall evaluation of the cluster is limited to the optimization algorithm for solving, and the node net power complementary indexes are obtained through the improvement of the modularity edge weights, which can be better adapted to different DGC division algorithms. In order to meet the structural and functional requirements of DGC, the modularity function is improved by integrating the electrical distance and cluster net power characteristic indexes, and a comprehensive modularity index for evaluating the effect of DGC division is obtained ρ_DGC, which is denoted: (12) $ρ_{D G C} = \frac{1}{2 m^{″}} \sum_{i} \sum_{j} ({e^{″}}_{i j} - \frac{{k^{″}}_{i} {k^{″}}_{j}}{2 m^{″}}) ϕ (i, j)$

${k^{″}}_{i}$ and ${k^{″}}_{j}$ are the sum of the weights of the edges connected to nodes i and j, respectively, under the integrated modularity metric. m″ denotes the sum of edge weights of all nodes under the integrated modularity metric, and ${e^{″}}_{i j}$ is the weighted improved node network edge weights: (13) ${e^{″}}_{i j} = ω_{1} e_{i j} + ω_{2} {e^{'}}_{i j}$

ω₁, ω₂ satisfy the constraints of ω₁ + ω₂ = 1, which characterize the weights of the DGC electrical distance modularity index and the node net power complementarity index, respectively. Different combinations of ω₁ and ω₂ parameters can be set to flexibly obtain the DGC division result with the largest integrated modularity index.

The objective function of DGC division is finally obtained as: (14) $F_{D G C} = \max {\sum_{c = 1}^{N_{c}} ρ_{D G C} (c)}$

where F_DGC is the largest combined modularity metric for the system across different DGC division results.

In the process of dividing distributed power clusters based on specific indexes, it is especially crucial to study and select appropriate cluster division algorithms. Currently, cluster analysis algorithms and intelligent algorithms have been widely studied and applied in the field of distributed power cluster segmentation, and have demonstrated their strong representativeness and practicality. In this section, based on the conventional K-means clustering algorithm, we explore the application of improved K-means clustering algorithm in cluster partitioning. Adaptive genetic algorithms are studied for cluster division strategies, as well as a new linear programming algorithm is proposed for cluster division. Through the research and comparison of these algorithms, more categories, more efficient and more comprehensive method choices can be provided for distributed power cluster division. The cluster division process based on improved K-means clustering algorithm is shown in Fig. 1. The number of clusters K is selected to improve the strategy, using the integrated cluster internal nodes and the cluster center of the distance between the tie also known as SSE and the size of the integrated modularity of the DGC division of the different K cases as an evaluation criterion to select the best value of the number of clusters K, which maintains the structural and functionality of the DGC, but also does not over-divide.

2.2

Energy Storage Cluster Hierarchical Regulation Architecture

In the distribution network, the number of user side is large, if directly controlled individually, it will greatly increase the complexity of regulation and control, the communication cost of the system is high, the control reliability is low, and it does not have the feasibility of practical implementation. Therefore, it is necessary to carry out aggregation and participate in the regulation and control of the distribution network in a unified manner in the form of clusters.

The variation characteristics of the equivalent charge state of the generalized energy storage are mainly affected by the characteristic parameters α and β. Therefore, the generalized energy storage can be k–means clustered based on the characteristic parameters α and α, and the generalized energy storage can be divided into multiple clusters. For the clustered generalized storage clusters, the characteristic parameters γ and λ are the average values of the corresponding parameters of all the generalized storage in the cluster, and γ and λ are the sum of the corresponding parameters of all the generalized storage in the cluster. This process can be done by the load aggregator, and the generalized energy storage cluster model after aggregation can be expressed as: (15) ${\begin{cases} E_{S O C Σ}^{t + 1} = α_{a v e} E_{S O C Σ}^{t} + β_{a v e} P_{a c Σ}^{t + 1} + γ_{Σ} - λ_{Σ} T_{o u t}^{t + 1} \\ E_{S O C Σ}^{t} = \sum_{i = 1}^{n_{c}} E_{S O C i}^{t} \\ P_{x Σ}^{t + 1} = \sum_{i = 1}^{n_{c}} P_{a c i}^{t + 1} \\ α_{a v e} = \frac{\sum_{i = 1}^{n_{c}} α_{i}}{n_{c}} \\ β_{a v e} = \frac{\sum_{i = 1}^{n_{c}} β_{i}}{n_{c}} \\ γ_{Σ} = \sum_{i = 1}^{n_{c}} γ_{i} \\ λ_{Σ} = \sum_{i = 1}^{n_{c}} λ_{i} \end{cases}$

Where $E_{S O C Σ}^{t}$ is the equivalent charge state of the generalized energy storage cluster at moment t: $P_{a c Σ}^{t + 1}$ is the active power of the generalized energy storage cluster at moment t + 1: α_avc and β_avc are the characteristic parameters of the generalized energy storage cluster: n_c is the number of generalized energy storage in the cluster.

Let E_soceq = E_so∑/n_c, accordingly, the constraints to be satisfied by the generalized energy storage cluster are: (16) $0 \leq E_{s o c e q} \leq 1$ (17) $\sum_{i = 1}^{n_{c}} P_{x, \min i} \leq P_{a c Σ} \leq \sum_{i = 1}^{n_{c}} P_{a c, \max j}$ (18) $Q_{a c q Σ} = \sum_{i = 1}^{n_{c}} P_{a c i} \tan φ_{a c i}$

To maintain the initial state of charge E_SOC∑0 of the energy storage clusters unchanged (i.e., the initial set temperature T_set0 in each room is unchanged), the clusters need to be operated at a baseline power P_acb∑, viz: (19) $P_{a c b Σ}^{t} = \frac{(1 - α_{a r c}) E_{S O C Σ 0} - γ_{Σ} + λ_{Σ} T_{o u t}^{t}}{β_{a v e}}$

This paper constitutes an energy storage cluster model, which forms a hierarchical control architecture of “distribution grid-load aggregator-users”, and the hierarchical control architecture of distribution grid energy storage cluster is shown in Fig. 2. The load aggregator first obtains the parameter information of each generalized energy storage and clusters them, and then reports the characteristic parameters of the generalized energy storage cluster to the grid control center, which sends out power regulation commands to the load aggregator according to the demand for safe and economic operation of the grid. On this basis, the load aggregator, based on the E_SOC and other state information of each generalized energy storage in the cluster, issues control instructions to the terminal controllers of each generalized energy storage according to the corresponding power allocation strategy, and realizes the response to the power instructions by adjusting the frequency of the air conditioning compressor. In the regulation process, the grid control center does not need to establish communication with all user terminals, but only needs to collect relevant information of the generalized energy storage cluster and issue dispatching instructions to the cluster, thus reducing the amount of data computation, alleviating the communication burden, and protecting user privacy.

2.3

Construction of control system based on hierarchical regulation of clusters

In engineering practice, the power grid adopts the management mode of “unified dispatching and hierarchical management”. The Southern Power Grid, for example, is divided into four levels: general dispatch, central dispatch, local dispatch and county dispatch. Under this management mode, the scope of resources controlled by each level of the power grid and resource objects are different, and are deployed by independent control centers, and then integrated through the coordination of power grids at all levels. In essence, this management mode can only realize the integrated economic operation of generation, transmission, distribution and utilization resources within its control at each level. However, the power grid is physically an interconnected whole, and different levels will interact with each other. In order to realize the integrated economic optimal scheduling under the ideal state of the whole grid, it is necessary to consider the mutual coordination of hierarchical regulation and control, calculate the equivalence of the boundary conditions of the regulation and control of the grid at different hierarchical levels, such as the load nodes, tariff information, and the boundary topology, and incorporate the equivalence information into the economic operation of the grid at the upper and lower hierarchical levels.

The continuous development of the user load management metering and electricity consumption management system provides conditions for a more refined description of the power and tariff information of the border nodes, and is also conducive to the further promotion of the work of integrated economic operation. The user types of the border nodes can be further subdivided, mainly based on industry attributes and tariff levels, so as to form the load structure and tariff structure of the border nodes, and consider the correlation of different categories of loads, which will provide an important reference for the accurate prediction of the load power and its tariff information of the border nodes. That is, there are: (20) ${\bar{D}}_{t} = \sum_{j = 1}^{N_{j}} D_{j, t} 1 \leq t \leq N_{t}$ (21) ${\bar{C}}_{t}^{u x e t} \frac{\sum_{j = 1}^{N_{j}} D_{j, t} C_{j, t}^{u x u T}}{\sum_{j = 1}^{N_{j}} D_{j, t}} 1 \leq t \leq N_{t}$

Where ${\bar{D}}_{i}$ is the load demand of the boundary node; D_i,t is the load of node j in time period t; ${\bar{C}}_{t}^{u x e T}$ is the price information of the boundary node; j is the serial number of the equaled node; and N_j is the number of equaled nodes on this boundary node.

On this basis, demand side response should be further considered in the integrated economic operation of the grid to determine the response capability and response cost of the boundary node. Summing up the rigid loads of the boundary nodes that do not participate in the response, there is: (22) ${\bar{D}}_{\min}^{u n e r T} = \sum_{k = 1}^{N_{e}} D_{k, \min}^{u n e r T}$

Calculating the maximum value of the response load of the boundary node and considering the network security calibration, the maximum load of the boundary node can be obtained, viz: (23) ${\bar{D}}_{\max}^{u x e T} = \min {\sum_{k = 1}^{N_{e}} D_{k, \max}^{u x e T}, P_{\max}^{u x e T}}$

where $P_{\max}^{x e c}$ is the maximum allowable transmission capacity of the network. The load response cost is the sum of the response costs of each load in the original grid, which can be obtained by using the method of user declaration or user demand elasticity measurement.

The intelligent viewing system based on integrated hierarchical regulation and control is shown in Fig. 3, and the analysis system is deployed in the dispatching III area, and the data sharing with other dispatching informatization systems is realized through the network conditions of III area. The system includes five business function modules, which are data declaration, basic data maintenance, economic operation analysis, economic operation pre-assessment, and economic operation post-assessment. All business function modules are supported by advanced visualization technology and intelligent data analysis technology.

1) Grid economic operation analysis. Taking into full consideration of the three links of power generation, transmission and distribution, and power consumption, it analyzes the main factors affecting the economic performance of grid operation, which can be analyzed in different time sequences, such as yearly, monthly, weekly, etc., so as to provide decision-making references for the improvement of economic efficiency of the power system.

2) Grid operation economy assessment. According to the two time dimensions of pre-assessment and post-assessment, around the decision-making management of the daily power generation plan, the economics of power grid operation at different levels of the Southern Power Grid is assessed, and the economics of the cross-provincial power transmission and reception plan of the AC/DC system and the power generation plan of each province can be assessed, so as to provide decision-making reference for the power system to formulate the power generation plan before the day of scientific development, and the power generation plan and the actual power generation scheduling results are compared to identify the key elements affecting the economics of power grid operation. It also compares the power generation plan with the actual power generation dispatching results to identify the key elements affecting the economy of power grid operation.

3) Optimize and manage the economic operation of the power system from the aspects of power purchase and power sale. In terms of power purchase, research on the pre-analysis method of power purchase cost, assessment of operation ideality, and proposal of the whole process and logic of the power transaction assessment concept; in terms of power sales, analyzing the dominant industries, grasping the numerical characteristics and changing law of social electricity consumption, and forecasting and analyzing the demand for electricity.

3

Empirical analysis of distribution grid-cluster-integrated hierarchical regulation and control

3.1

Analysis of optimal scheduling for distribution network cluster division

In this paper, the following three metrics are used to measure the effectiveness of different algorithms for cluster division, i.e., modularity, average standard deviation of controllable capacity of each cluster, and average standard deviation of voltage sensitivity of each cluster. The modularity measures the advantages and disadvantages of the cluster partitioning effect, and the larger the modularity is, the more reasonable the cluster partitioning result is. The average standard deviation of the division indexes visualizes the rationality of the division results, and the smaller the average standard deviation is, the more similar the division indexes of the cluster are.

Selecting the PV output period, through the K-means traditional clustering algorithm, SLM algorithm and this paper’s improved method for comparison, the cluster division effect comparison is shown in Figure 4, where (a) ~ (c) are the module degree, voltage sensitivity average standard deviation and reactive power adjustable capacity average standard deviation of the comparison results respectively.

From the comparison of module degree, it can be seen that the module degree of SLM algorithm is larger than the traditional k-means in all time sections, while this paper’s algorithm is better than the two algorithms, with the module degree between 0.27 and 0.80, which means that the proposed algorithm of this paper delineates the results better. From the comparison of average standard deviation, it can be seen that the average standard deviation of voltage sensitivity and reactive power controllable capacity of SLM algorithm is much smaller than k-means, and the standard deviation of sensitivity of this paper’s improved method is reduced to within 13.5-15.1% on average, and the standard deviation of controllable capacity is reduced to less than 40% on average, and the lowest can be up to 13%, which means that the characteristics of the distributed power supply within the clusters of this paper’s improved algorithmic segmentation results are more similar. In summary, it can be seen that compared with other traditional clustering algorithms, this paper’s improved method of power clustering effect is better.

Comparison of global regulation effect before and after optimization is shown in Fig. 5, the highest network loss of one day distribution network before optimization is 13.9Kw/h, and the average hourly loss is 8.107Kw/h. The average hourly loss after optimization is 5.717Kw/h, and after optimization of this paper, the average hourly loss is reduced by 2.39Kw/h compared to optimization, which further verifies the application effect of the method of this paper.

3.2

Analysis of the effect of hierarchical regulation of distribution networks

In order to prove the effectiveness of the hierarchical regulation method in this paper, this paper is verified in the distribution network model. The power sources in the system are connected to the distribution network with rated generation power, and the reactive power outputs of the PV inverters on the nodes are changed to -10, -5, 0, 5, and 10 kvar, respectively.The results of the regulation under five different reactive power outputs are shown in Table 1. From the data in the table, it can be seen that the voltage regulation in this paper requires only a small amount of data and calculation, but achieves a very high accuracy, the error values are below 0.0008%.

Table 1.

Regulatory result

□Q_PV/k var	Voltage/V
□Q_PV/k var	True value	Calculated value	Error value(%)
0kvar	398.6589	—	—
5kvar	400.6335	400.6569	0.0006
10kvar	402.5114	402.6549	0.0004
-5kvar	396.5804	396.6609	0.0002
-10kvar	394.3906	394.6629	0.0007

In order to verify the effect of the hierarchical regulation strategy, two scenarios of the system appearing voltage exceeding the upper limit and voltage exceeding the lower limit are verified.

The scenario in which the system appears to have voltage over the lower limit is in the distribution network with distributed power access, and the voltage over the lower limit is likely to occur when the power supply is insufficient and the user load is large. Assuming that in the low-voltage distribution network model of this algorithm, the PV power generation power on node 9 and node 12 is 8kW, and the PV power generation power on node 19 is 10kW, and the loads of nodes 16, 17, 18, and 19 on line 5 are doubled, then there is a serious problem of the voltage crossing the lower limit mainly at node 18.

The effect of voltage hierarchical regulation (over the lower limit) is shown in Fig. 6. In the above voltage over the lower limit scenario, firstly, the node with the lowest system voltage is located in the load node on line 5 by the voltage perception algorithm, and then the amount of reactive power adjustment of the PV power supply on node 19 participating in voltage control is calculated to be 3.33kvar.

The scenario in which the system appears to have a voltage overrun limit is in a distribution network accessed by distributed power sources, where the lines are prone to overvoltage problems when the power sources are in full power. It is assumed that in the LV distribution network model of this algorithm, all three distributed power sources are running at rated power, and the load power on the line is unchanged, at which time the overvoltage problem occurs mainly at node 9 and node 19.

The effect of voltage hierarchical regulation (over the upper limit) is shown in Fig. 7, when applying the voltage hierarchical regulation strategy, in addition to node 19, which has the most serious overvoltage problem, participating in voltage regulation, the inverter of node 9 on the adjacent line also absorbs reactive power to participate in regulating the voltage, and the calculated reactive power adjustments of the two nodes are -12.80 kvar and -7.73 kvar, respectively.

The voltage distribution after applying the hierarchical regulation strategy is shown in Fig. 8, which compares the traditional in-situ Q(U) control method with the hierarchical control method in this paper. It can be seen that, for the serious voltage crossing upper limit situation in the system, the in situ control can improve the voltage problem, but the control effect is limited. In contrast, the voltage hierarchical control strategy in this paper, in addition to utilizing the reactive power adjustable capacity of this line, also coordinates the participation of other lines in voltage regulation, which can eventually adjust the voltage to the qualified range with a faster speed.

4

Conclusion

In this paper, oriented to the practical engineering needs of distribution networks, we constructed an energy storage cluster hierarchical regulation and control architecture based on power cluster delineation, and established a distribution network optimization and control system based on cluster hierarchical regulation and control. This paper mainly carries out the following research work:

1) This paper analyzes the optimal scheduling effect of this paper’s cluster delineation by comparing the two methods of K-means traditional clustering algorithm, SLM algorithm and the improved method of this paper in terms of the comparison of the modularity degree, the average standard deviation of the voltage sensitivity and the average standard deviation of the reactive power adjustable control capacity. Among them, the module degree of this paper’s algorithm is between 0.27 and 0.80, which is better than the other methods, proving that the proposed algorithm of this paper has better division results. On the other hand, the standard deviation of sensitivity of this paper’s improved method is reduced to within 13.5-15.1% on average, and the standard deviation of controllable capacity is reduced by less than 40% on average, which is also better than other methods. It further indicates that the improved algorithm in this paper performs better in power cluster segmentation compared to other traditional clustering algorithms.

2) The comparison before and after the optimization of the global regulation effect shows that the highest network loss of one day’s distribution network before optimization is 13.9Kw/h, and the average hourly loss is 8.107Kw/h. The average hourly loss after optimization of this paper is 5.717Kw/h, which is reduced by 2.39Kw/h compared with the pre-optimization period, which further verifies the application effect of this paper’s method.

3) Compared with the traditional local Q(U) control method, this paper’s hierarchical control method gives better play to the reactive power regulation capability of the distributed power supply and realizes the hierarchical control of the distribution network voltage. The voltage regulation only requires a small amount of data and calculation, but achieves a very high accuracy, with error values below 0.0008%. Its reactive power adjustment for voltage control under the scenario of voltage crossing the lower limit of the system is 3.33kvar, and the reactive power adjustment under the scenario of voltage crossing the upper limit is -12.80kvar and -7.73kvar respectively.

Funding:

State Grid Hebei Electric Power Co., Ltd. Technology Project Funding - Research and Development of Local Consumption Devices for Integrated Photovoltaic, Storage, and Charging Systems and Control Strategy Research-kj2023-032.

Idioma:: Inglés

Calendario de la edición:: 1 veces al año
Temas de la revista:: Ciencias de la vida, Ciencias de la vida, otros, Matemáticas, Matemáticas aplicadas, Matemáticas generales, Física, Física, otros

RSS Feed de revista

Research on Distribution-Cluster-Integrated Layered Regulation Architecture and Business Control System for New Distribution Grids

Yu Zhang

Shicheng Huang

Bo Yuan

Xuguang Yang

Shuai Li

Publicado en línea: 17 mar 2025

Recibido: 24 oct 2024

Aceptado: 27 ene 2025

DOI: https://doi.org/10.2478/amns-2025-0215

Palabras claveCluster division, Optimal dispatch, Hierarchical regulation, Distribution network, Control system

© 2025 Yu Zhang et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Palabras clave
Cluster division, Optimal dispatch, Hierarchical regulation, Distribution network, Control system