High permeability rate distributed photovoltaic feeder area closed loop regulation method and device

Abstract

The application provides a high-permeability distributed photovoltaic feeder area closed-loop regulation method and device, relates to the distributed photovoltaic control technical field of electric power systems, introduces a feeder regulation mechanism in the upper layer, introduces an area aggregation mechanism and an area regulation mechanism in the lower layer, three form a whole-process regulation closed loop to obtain a double-layer collaborative support framework of the feeder-area high-permeability distributed photovoltaic with whole-process closed-loop control, establishes an upper feeder regulation model, a lower area regulation model and a lower area aggregation model based on the framework, obtains a multi-stage distributed photovoltaic multi-objective optimization regulation model and solves to obtain the optimal regulation scheme of each stage. The application can realize the closed-loop accurate control of the medium and low voltage level, fully excavate the active / reactive power active support potential of the distributed photovoltaic, and fully solve the problems of the traditional method, such as the lack of timeliness and the lack of feedback in execution.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of distributed photovoltaic control of power systems, and particularly relates to a high-penetration distributed photovoltaic feeder area closed-loop regulation method and device. BACKGROUND

[0002] With the rapid development of distributed photovoltaic power generation technology and the continuous growth of installed capacity, photovoltaic development has shown a development pattern of centralized and distributed development. With the gradual increase of photovoltaic penetration rate of medium and low voltage distribution network, the strong randomness and volatility of photovoltaic output make the possibility of changing the power flow direction of distribution feeder surge, and overvoltage problems also occur in some extremely high-penetration distribution networks, which seriously affect the safe and stable operation of the distribution network. Traditional control methods mainly adjust the tap position of the on-load voltage regulating transformer in each substation or control the switching group number of the compensation capacitor bank, but this adjustment method has serious lag and lacks economic efficiency. Therefore, it is urgent to study the adjustment method of distributed photovoltaic resources supporting medium and low voltage distribution network to ensure the safe and stable operation of the power grid.

[0003] At present, the support means of distributed photovoltaic resources for the power grid can be summarized as local optimization, centralized optimization and distributed optimization. Among them, the local optimization strategy has the advantages of high reliability and no additional communication demand, but the local optimization effect is limited due to the lack of global information. The centralized optimization strategy can realize the global optimization of the distribution network and coordinate the use of various resources in the distribution network, but it lacks efficient aggregation and control means and has not modeled low-voltage photovoltaic objects in depth. The distributed optimization strategy can achieve a certain degree of balance optimization and autonomy between regions, retain higher expandability and reduce communication burden. Overall, the three methods have not effectively solved the problems of the lack of coordinated operation mode of photovoltaic in multiple voltage levels in physical sense and the difficulty in fine characterization of dynamic characteristics of low-voltage distributed photovoltaic clusters, which restricts the deep mining of active support capacity of high-penetration photovoltaic. In view of this problem, the existing technology proposes a double-layer control strategy for medium and low voltage distributed photovoltaic, which adopts a day-ahead-day-in double-layer control architecture for distributed photovoltaic clusters. However, the existing technology decouples the regulation scale of medium and low voltage distribution network, only considers the algorithm logic of day-ahead optimization of medium voltage distribution network and real-time regulation of low voltage distribution network, that is, the medium voltage distribution network only interacts with the low voltage distribution network at the day-ahead layer, and fails to realize double-layer real-time control. Under the condition of rapid fluctuation of source and load, there are problems of insufficient control timeliness and lack of feedback in execution. SUMMARY

[0004] To solve one of the technical problems in the background technology, the purpose of the present application is to provide a high-penetration distributed photovoltaic feeder area closed-loop regulation method and device, which realizes closed-loop accurate control of medium and low voltage levels and fully excavates the active support potential of distributed photovoltaic active / reactive power.

[0005] To achieve the above object, the present application provides the following technical solutions.

[0006] The present application provides, in the first aspect, a high-permeability distributed photovoltaic feeder area closed-loop regulation method, which is used for regulation of a medium-low voltage distribution network, an upper layer of the medium-low voltage distribution network being a medium-voltage feeder level system, and a lower layer being a plurality of low-voltage areas containing high-permeability distributed photovoltaic at different positions of a feeder, the method comprising the following steps: introducing a feeder regulation mechanism in the upper layer, introducing an area aggregation mechanism and an area regulation mechanism in the lower layer, the feeder regulation mechanism, the area aggregation mechanism and the area regulation mechanism forming a whole-process regulation closed loop, so as to obtain a double-layer collaborative support framework of the feeder-area high-permeability distributed photovoltaic with whole-process closed-loop control; based on the double-layer collaborative support framework, an upper feeder regulation model, a lower area regulation model and a lower area aggregation model are established for the feeder regulation mechanism, the area aggregation mechanism and the area regulation mechanism respectively, and a multi-stage distributed photovoltaic multi-objective optimization regulation model is obtained after combination.

[0007] The present application provides, in the first aspect, a preferred scheme, in the step of introducing a feeder regulation mechanism in the upper layer, introducing an area aggregation mechanism and an area regulation mechanism in the lower layer, the feeder regulation mechanism, the area aggregation mechanism and the area regulation mechanism forming a whole-process regulation closed loop, so as to obtain a double-layer collaborative support framework of the feeder-area high-permeability distributed photovoltaic with whole-process closed-loop control, the whole-process regulation closed loop specifically being: the area aggregation mechanism obtains area power flow information from the area regulation mechanism and generates area control information containing area operation domain boundaries according to the area power flow information, and transmits the area control information to the feeder regulation mechanism and the area regulation mechanism; the feeder regulation mechanism generates feeder instruction information according to the area control information and transmits the feeder instruction information to the area regulation mechanism; the area regulation mechanism updates the area power flow information according to the instruction information and in combination with the area control information, and uploads the area power flow information to the area aggregation mechanism.

[0008] The present application provides, in the first aspect, a preferred scheme, the plurality of low-voltage areas at least including three low-voltage areas with adjustable distributed photovoltaic and photovoltaic permeability decreasing in turn, and one low-voltage area without distributed photovoltaic; the upper layer and the lower layer exchange power through a transformer gateway.

[0009] The application provides a preferred scheme in the first aspect, and the lower-layer transformer area aggregation model is established by adopting a lower-layer transformer area dynamic domain aggregation model in the step of establishing the upper-layer feeder regulation model by the feeder regulation mechanism, the lower-layer transformer area regulation model by the transformer area aggregation mechanism, and the lower-layer transformer area aggregation model by the transformer area regulation mechanism, and the multi-stage distributed photovoltaic multi-objective optimization regulation model is obtained after combination.

[0010] The application provides a preferred scheme in the first aspect, and the step of establishing an upper feeder regulation model for a feeder regulation mechanism, a lower transformer area regulation model for a transformer area aggregation mechanism, and a lower transformer area aggregation model for a transformer area regulation mechanism based on a double-layer collaborative support framework, and combining the models to obtain a multi-stage distributed photovoltaic multi-objective optimization regulation model, wherein the step of establishing the upper feeder regulation model for the feeder regulation mechanism, the lower transformer area regulation model for the transformer area aggregation mechanism, and the lower transformer area aggregation model for the transformer area regulation mechanism based on the double-layer collaborative support framework, and combining the models to obtain the multi-stage distributed photovoltaic multi-objective optimization regulation model, the upper feeder regulation model adopts an upper feeder multi-mode optimization regulation model, and the establishing step comprises the following steps: minimizing the sum of the absolute value of active power and the absolute value of reactive power at the feeder outlet as the target to establish a feeder self-balancing degree evaluation function; minimizing the sum of the voltage deviation of each node from the reference voltage as the target to establish a feeder voltage deviation degree evaluation function; setting feeder voltage out-of-limit penalty values according to the out-of-limit conditions of the voltage of each node in the feeder to establish a feeder voltage penalty function; setting feeder transformer area heavy load penalty values according to the load rate of each transformer area to establish a feeder transformer area heavy load penalty function; minimizing the distance between the load rate of each transformer area and the average value of the load rate of all transformer areas as the target to establish a feeder transformer area load rate balance degree evaluation function; introducing weight coefficients based on the feeder self-balancing degree evaluation function, the feeder voltage deviation degree evaluation function, the feeder voltage penalty function, the feeder transformer area heavy load penalty function, and the feeder transformer area load rate balance degree evaluation function, and combining the selected feeder main optimization target and the feeder operation mode to obtain a second target function; setting a second constraint condition for the second target function to obtain an upper feeder multi-mode optimization regulation model; and the second constraint condition comprises node voltage, line current, network power flow, and transformer area adjustable power constraint conditions.

[0011] The application provides a preferred scheme in the first aspect, and the upper feeder regulation model is established for a feeder regulation mechanism, the lower transformer area regulation model is established for a transformer area aggregation mechanism, and the lower transformer area regulation model is established for a transformer area regulation mechanism based on the double-layer cooperative support framework, so that the multi-stage distributed photovoltaic multi-objective optimization regulation model is obtained after combination, in the step, the lower transformer area regulation model adopts a lower transformer area low-voltage photovoltaic regulation model, and the establishing step comprises the following steps: based on active power and reactive power at a transformer area outlet and preset active power instructions and reactive power instructions, a deviation of current transformer area output power from a set value is acquired, an instruction approximation degree evaluation function is established with the minimum approximation degree as a target; a transformer area voltage out-of-limit penalty value is set according to an out-of-limit condition of voltage of each node in the transformer area, and a transformer area voltage penalty function two is established; a third objective function is obtained based on the instruction approximation degree evaluation function and the transformer area voltage penalty function two and by introducing a weight coefficient; a third constraint condition is set for the third objective function, and the lower transformer area low-voltage photovoltaic regulation model is obtained; and the third constraint condition comprises node voltage, line current, network power flow and distributed photovoltaic adjustable power constraint conditions.

[0012] The application provides a preferred scheme in the first aspect, and the multi-stage distributed photovoltaic multi-objective optimization regulation model is solved, and the optimal regulation scheme of each stage is obtained in the step, wherein the improved NSGA-III algorithm is used to solve the multi-stage distributed photovoltaic multi-objective optimization regulation model, and the improved NSGA-III algorithm is obtained by introducing an adaptive Latin hypercube sampling initialization mechanism and a reference point updating strategy based on historical information to improve the NSGA-III algorithm.

[0013] The application provides a preferred scheme in the first aspect, and the multi-stage distributed photovoltaic multi-objective optimization regulation model is solved, and the optimal regulation scheme of each stage is obtained in the step, wherein the optimization algorithm is used for solving and optimization, and when iteration is completed, the transformer area operation domain boundary, the feeder control instruction and the distributed photovoltaic output control instruction are output in stages, and are used as the optimal regulation scheme of each stage respectively.

[0014] Further, in the step of solving the multi-stage distributed photovoltaic multi-objective optimization regulation model and obtaining the optimal regulation scheme of each stage, a feeder control instruction checking step is arranged: whether the feeder control instruction and the transformer area operation domain boundary are located in the boundary is analyzed, if not, the parameter information of the model is updated, and the feeder control instruction is iterated again; if yes, the feeder control instruction output is generated.

[0015] The application provides a high-permeability distributed photovoltaic feeder area closed-loop regulation device in a second aspect, which is used for regulation of a medium-low voltage distribution network, an upper layer of the medium-low voltage distribution network is a medium voltage feeder level system, and a lower layer is a plurality of low voltage areas containing high-permeability distributed photovoltaic at different positions of a feeder, and the device comprises: a support frame modeling module, which is used for introducing a feeder regulation mechanism in the upper layer, introducing an area aggregation mechanism and an area regulation mechanism in the lower layer, and forming a whole-process regulation closed loop of the feeder regulation mechanism, the area aggregation mechanism and the area regulation mechanism, so as to obtain a double-layer collaborative support frame of the feeder-area high-permeability distributed photovoltaic of whole-process closed-loop control; a regulation model construction module, which is used for establishing an upper feeder regulation model for the feeder regulation mechanism, a lower area regulation model for the area aggregation mechanism, and a lower area aggregation model for the area regulation mechanism based on the double-layer collaborative support frame, and obtaining a multi-stage distributed photovoltaic multi-objective optimization regulation model after combination; and a regulation scheme output module, which is used for solving the multi-stage distributed photovoltaic multi-objective optimization regulation model and obtaining an optimal regulation scheme of each stage.

[0016] Compared with the prior art, the application has the following advantages:

[0017] Compared with the traditional day-ahead-day-in double-layer control architecture for distributed photovoltaic clusters, the application provides a high-permeability distributed photovoltaic feeder area closed-loop regulation method and device, a double-layer collaborative support frame of the feeder-area high-permeability distributed photovoltaic of whole-process closed-loop control, that is, a "lower-upper-lower" regulation framework, further extends the regulation level to the low voltage side, enhances the closed-loop interaction logic between the medium-low voltage levels, and specifically realizes the control mode of area resource aggregation through the area aggregation mechanism, feeder centralized regulation through the feeder regulation mechanism, and area decomposition execution through the area regulation mechanism, realizes the real-time and efficient collaborative optimization regulation method of "centralized calculation-distributed execution" through the low-voltage distributed photovoltaic cluster aggregation operation domain and the instruction distribution mechanism, and fully solves the problems of insufficient timeliness and lack of feedback in the execution of the traditional method. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, brief introductions will be given to the drawings needed to be used in the embodiments or prior art descriptions. Obviously, the drawings in the following description are only embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of the provided drawings.

[0019] Figure 1 is a flowchart of the regulation method of the application;

[0020] Figure 2It is a schematic diagram of a feeder-substation high permeability distributed photovoltaic double-layer collaborative support frame of the application;

[0021] Figure 3 It is an interactive principle diagram of a full-process regulation strategy based on dynamic operation domains of the application;

[0022] Figure 4 It is an algorithm flowchart of a multi-stage distributed photovoltaic multi-objective optimization regulation model;

[0023] Figure 5 It is a flowchart structure diagram of the improved NSGA-III algorithm;

[0024] Figure 6 It is a distribution network topology diagram;

[0025] Figures 7 to 11 It is a typical day lower controllable substation node 8, 15, 24, 31, 36 active operation domain boundary and output-time variation curve diagram;

[0026] Figures 12 to 16 It is a typical day lower controllable substation node 8, 15, 24, 31, 36 reactive operation domain boundary and output-time variation curve diagram;

[0027] Figure 17 It is a typical day lower controllable substation node 8, 15, 24, 31, 36 feeder voltage offset-time variation curve diagram before and after optimization;

[0028] Figure 18 It is a typical day normal mode / management mode feeder output active / reactive power and lower controllable substation node 8, 15, 24, 31, 36 substation variable load rate-time variation curve diagram before and after optimization;

[0029] Figure 19 It is a typical day node 8 controllable substation photovoltaic active output / substation outlet target output-time variation curve diagram;

[0030] Figure 20 It is a typical day node 8 controllable substation photovoltaic reactive output / substation outlet target output-time variation curve diagram. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0032] Please refer to Figure 1In an alternative embodiment, a high penetration rate distributed photovoltaic feeder area closed-loop regulation method is provided for the regulation of a medium and low voltage distribution network. The upper layer of the medium and low voltage distribution network is a medium voltage feeder level system, and the lower layer is a plurality of low voltage areas containing high penetration rate distributed photovoltaics connected to different positions of the feeder. In this paper, the photovoltaic penetration rate of the photovoltaic high penetration rate distribution network is 50% or more of the load, which is considered as high penetration rate. In this paper, 50% is the dividing point, and more than 50% is considered as higher penetration rate, and more than 100% of the load is considered as extremely high penetration rate.

[0033] The method is achieved by the following steps:

[0034] S1. Introduce a feeder regulation mechanism in the upper layer, and introduce an area aggregation mechanism and an area regulation mechanism in the lower layer. The feeder regulation mechanism, the area aggregation mechanism and the area regulation mechanism form a whole-process regulation closed loop, so as to obtain a double-layer collaborative support framework of the feeder-area high penetration rate distributed photovoltaic with whole-process closed-loop control. This step realizes the construction of the "bottom-up-bottom" whole-process feeder-area closed-loop control framework, specifically, the following two sub-steps are included:

[0035] S11. Construct a double-layer collaborative support framework of the feeder-area high penetration rate distributed photovoltaic.

[0036] In a preferred embodiment, the plurality of low voltage areas includes at least three low voltage areas with adjustable distributed photovoltaics and decreasing photovoltaic penetration rates, and one low voltage area without distributed photovoltaics. The upper layer and the lower layer exchange power through a transformer gateway.

[0037] In order to fully consider the influence of the internal topology relationship of the area on the adjustable distributed photovoltaic resources in the lower layer area, a double-layer collaborative support framework considering the topology structure of different voltage levels (including 10kV and 220V voltage levels) is designed to realize the energy and information exchange between the feeder and the area, as shown in Figure 2 .

[0038] Figure 2 The upper layer is a 10kV feeder level system considering the topology structure (considering the detailed topology structure of the low voltage area and the medium voltage distribution network), and the lower layer is a plurality of 380V voltage level typical areas a to d connected to different positions of the feeder, with decreasing photovoltaic penetration rates. Among them, area d represents an area with only user load and does not participate in regulation; the areas with adjustable distributed photovoltaics in the lower layer, such as areas a, b and c, realize distributed photovoltaic aggregation under the condition of considering their internal topology, that is, the dynamic operation domain aggregation mechanism of the lower layer area transmits adjustable photovoltaic control information to the upper layer system, and receives control instructions from the upper layer for distribution. The upper and lower distribution networks can exchange power through a transformer gateway.

[0039] S12, construct a "bottom-up-bottom" whole-process regulation mechanism considering dynamic operation domain:

[0040] Active regulation of key areas with high or even extremely high photovoltaic penetration in feeders is of great significance for achieving voltage control or internal power self-consumption of feeder-level systems. Based on the previously established dual-layer collaborative support framework of medium and low voltage distribution network containing high penetration of distributed photovoltaic (feeder- substation high penetration of distributed photovoltaic dual-layer collaborative support framework), a feeder-substation whole-process closed-loop regulation mechanism is designed, and its operation logic is shown in Figure 3

[0041] In an optional embodiment, the whole-process regulation closed loop is as follows: the substation aggregation mechanism obtains the substation power flow information from the substation regulation mechanism and generates substation control information containing the substation operation domain boundary, which is transmitted to the feeder regulation mechanism and the substation regulation mechanism; the feeder regulation mechanism generates feeder instruction information according to the substation control information and issues it to the substation regulation mechanism; the substation regulation mechanism updates the substation power flow information according to the instruction information and combines the substation control information, and distributes the substation distributed photovoltaic output, and then uploads it to the substation aggregation mechanism. In combination with Figure 3 , specifically: the feeder-substation whole-process regulation mechanism constructed in this embodiment, which has the sequential characteristics of "bottom-up-bottom" (bottom-layer substation dynamic operation domain aggregation mechanism-top-layer feeder multi-mode optimization regulation mechanism-bottom-layer substation low-voltage photovoltaic regulation mechanism), includes: the bottom-layer substation dynamic operation domain aggregation mechanism generates the substation operation domain boundary as control information according to the substation power flow information (substation power flow information), and transmits the control information to the top-layer feeder multi-mode optimization regulation mechanism and the bottom-layer substation low-voltage photovoltaic regulation mechanism; the top-layer regulation generates a set of feeder instructions based on the substation control information, and issues the instruction information to the bottom-layer substation low-voltage photovoltaic regulation mechanism; the bottom-layer low-voltage photovoltaic regulation distributes the distributed photovoltaic output in the substation with the substation outlet power as the target, updates the substation power flow information after obtaining the new distributed photovoltaic output, and returns it to the bottom-layer substation dynamic operation domain aggregation mechanism for subsequent circulation. In this embodiment, the operation domain boundary is a more accurate algorithm operation constraint than the traditional control interval.

[0042] Compared with the traditional day-ahead-day-ahead dual-layer control architecture for distributed photovoltaic clusters, the "bottom-up-bottom" regulation framework (feeder-substation high penetration of distributed photovoltaic dual-layer collaborative support framework) further expands the regulation level to the low-voltage side, enhances the closed-loop interaction logic between the medium and low voltage levels, and realizes the real-time and efficient collaborative optimization regulation method of "centralized calculation-distributed execution" through the aggregation of the low-voltage distributed photovoltaic cluster operation domain (bottom-layer substation dynamic operation domain aggregation mechanism) and the instruction distribution mechanism (bottom-layer substation low-voltage photovoltaic regulation mechanism).

[0043] ​S2. Based on the double-layer cooperative support framework, a feeder regulation mechanism, a feeder aggregation mechanism and a feeder regulation mechanism are respectively established to establish an upper feeder regulation model, a lower feeder regulation model and a lower feeder aggregation model, and a multi-stage distributed photovoltaic multi-objective optimization regulation model is obtained after combination. Specifically, in an optional embodiment, the establishment process of each model is as follows:

[0044] (I) The lower feeder aggregation model adopts a lower feeder dynamic domain aggregation model, also known as a lower feeder operating domain search multi-objective model, and the establishment steps are as follows:

[0045] S21a. The sum of the upper and lower adjustable ranges of the active power at the feeder outlet and the upper and lower adjustable ranges of the reactive power at the feeder outlet is maximized to construct an adjustable interval upper limit evaluation function and an adjustable interval lower limit evaluation function ;

[0046] S21b. According to the over-limit situation of the voltage of each node in the feeder, a feeder voltage over-limit penalty value is set to establish a feeder voltage penalty function one ;

[0047] S21c. Based on , and , a weight coefficient is introduced and a first objective function is obtained in combination with the interval search direction ;

[0048] S21d. The first constraint condition is set for the objective function to obtain the lower feeder dynamic domain aggregation model. In a preferred embodiment, the first constraint condition has: node voltage, line current, network power flow and distributed photovoltaic adjustable power constraint condition.

[0049] More specifically, the model has strong universality and high calculation efficiency, and can adapt to high penetration rate photovoltaic feeder power flow rapid change scenarios, with the farthest distance between the upper and lower limits of the operating domain at the feeder outlet and no voltage over-limit situation in the feeder as the target, and the feeder network topology and distributed photovoltaic adjustable power as the constraint condition.

[0050] The lower feeder operating domain search multi-objective model has strong universality and high calculation efficiency, and can adapt to high penetration rate photovoltaic feeder power flow rapid change scenarios, with the farthest distance between the upper and lower limits of the operating domain at the feeder outlet and no voltage over-limit situation in the feeder as the target, and the feeder network topology and distributed photovoltaic adjustable power as the constraint condition.

[0051] To evaluate the advantages and disadvantages of each particle in the population, the objective function of the lower feeder operating domain search multi-objective model (also known as the lower feeder dynamic operating domain aggregation model) is as shown in equation (1):

[0052] (1);

[0053] wherein, is the objective function of the multi-objective model for searching the operation domain of the lower layer area; 、 is the weight value for evaluating the importance of the sub-objective in the total objective function; represents the selected interval search direction, if 1, the upper limit of the search interval, otherwise if -1, the lower limit of the search interval; 、 、 is the sub-objective function in the total objective function, respectively representing the upper limit adjustable interval evaluation function, the lower limit adjustable interval evaluation function, and the voltage penalty function. The optimization result of is the operation domain boundary of the area, which is used as control information. The so-called dynamic operation domain refers to real-time updating of the operation domain boundary for the selected example to achieve accurate aggregation of the adjustable power of the area.

[0054] For the upper limit adjustable interval evaluation function and the lower limit adjustable interval evaluation function , the sum of the upper and lower adjustable ranges of the active power at the outlet of the area and the upper and lower adjustable ranges of the reactive power at the outlet of the area is maximized as the objective function and are respectively used to evaluate the upper and lower limits of the operation domain at the outlet of the area corresponding to the particle (single solution in the optimization algorithm). In order to evaluate the active and reactive boundaries of the operation domain in the same scale, the active and reactive power values at the outlet of the area need to be normalized, and then the upper and lower adjustable ranges of the active and reactive power can be compared under the unified standard. After obtaining the normalized upper and lower limit values, their distances from the corresponding reference values are compared, and the greater the distance, the better the performance of the target, and the calculation formula is shown in formula (2) and formula (3):

[0055] (2);

[0056] (3);

[0057] wherein, 、 are the upper and lower limits of the adjustable range of the active power at the outlet of the area; 、 are the upper and lower limits of the adjustable range of the reactive power at the outlet of the area; 、 are the reference active and reactive power at the outlet of the area.

[0058] The objective function This is the voltage penalty function for the transformer area, also known as the transformer area voltage penalty function one. During the iterative optimization process of particles in space, the power flow situation within the transformer area will inevitably change, and voltage over-limit may occur within the transformer area, violating the requirements for safe and stable operation of the transformer area operating domain. Therefore, a voltage penalty function needs to be introduced to monitor each node within the transformer area. The lower-level transformer area operating domain generation mechanism also introduces a variable voltage penalty range, and the updated voltage penalty calculation formula is shown in equation (4):

[0059] (4);

[0060] in, This represents the total number of nodes within the transformer area. Number the currently traversed node; The voltage penalty parameter is set, which is also known as the voltage over-limit penalty value for the transformer area; For nodes The voltage value; , The set upper and lower voltage penalty values ​​can be achieved by changing the upper and lower voltage limits to adapt to voltage requirements under different conditions.

[0061] The multi-objective model for searching the lower-level transformer area operation domain needs to consider constraints such as node voltage, line current, network power flow, and adjustable power of distributed photovoltaic systems.

[0062] (1) Node voltage

[0063] (5);

[0064] in, This represents the node voltage at node i at time t, in V; Represents the minimum allowable node voltage value at node i, in V; This represents the highest allowed node voltage value at node i, in V.

[0065] (2) Line current

[0066] (6);

[0067] in, express Constant connection node , The current flowing through the line; Indicates the connection node , The maximum allowable current carrying capacity of the line.

[0068] (3) Internet trends

[0069] (7);

[0070] wherein: , Pi, qipresent the active and reactive part of the load power at node i; , Pi, qipresent the active and reactive part of the photovoltaic power at node i; , Gi, Bi represent the conductance and susceptance of the line between node i and node j; present the phase angle difference of the node voltage at node i and node j; N represents the node number.

[0071] (4) Distributed photovoltaic adjustable power

[0072] The distributed photovoltaic adjustable power needs to consider the active power limit, the reactive power limit, the total output power limit and other hard constraint limits of the photovoltaic inverter, and the formulas are shown in equations (8) to (10):

[0073] (8);

[0074] (9);

[0075] (10);

[0076] wherein, , Pi, qipresent the active and reactive part of the photovoltaic power at time t. , Pi, qipresent the active and reactive part of the photovoltaic power at time t. present the apparent power upper limit of the distributed photovoltaic equipment; , Pi, qipresent the active and reactive part of the photovoltaic power at time t.

[0077] (II) Upper feeder regulation model, adopt upper feeder multi-mode optimization regulation model, also called: upper feeder multi-objective optimization model. The establishment steps are as follows:

[0078] S22a. Establish a feeder self-balancing degree evaluation function with the minimum sum of the absolute value of active power and the absolute value of reactive power at the outlet of the feeder as the target ;

[0079] S22b. Establish a feeder voltage offset degree evaluation function with the minimum sum of the deviation of each node voltage from the reference voltage as the target ;

[0080] S22c. According to the out-of-limit situation of each node voltage in the feeder, set the feeder voltage out-of-limit penalty value in steps, and establish a feeder voltage penalty function​ ;

[0081] S22d. Set the feeder transformer load rate according to the transformer load rate of each area, set the feeder transformer heavy load penalty value, and establish the feeder transformer heavy load penalty function ;

[0082] S22e. With the minimum distance between the transformer load rate of each area and the average value of the transformer load rate of all areas as the target, the feeder transformer load rate balance evaluation function is established ;

[0083] S22f. Based on , , , and , a weight coefficient is introduced and combined with the selected feeder main optimization target and feeder operation mode to obtain the second objective function ;

[0084] S22g. Set the second constraint condition for the objective function , and obtain the upper feeder multi-mode optimization control model. In an preferred embodiment, the second constraint condition is: node voltage, line current, network power flow and area adjustable power constraint condition.

[0085] More specifically, the upper feeder multi-objective optimization model sets two types of four groups of sub-objective function combinations corresponding to the normal mode and the overload control mode of the feeder operation: the feeder voltage without over-limit, the feeder active / reactive power self-consumption or the minimum deviation of the feeder total voltage is taken as the target in the normal operation mode of the feeder; the overload control mode further includes the feeder transformer without overload phenomenon and the optimal sub-target of the balance of the feeder transformer load rate based on the above targets. Based on the above targets, the model takes the feeder network topology and distributed photovoltaic adjustable power as the constraint condition, and constructs a multi-objective optimization model suitable for various feeder control requirements and efficient calculation.

[0086] Considering the two control modes of the upper feeder and the different optimization rules in each control mode, the objective function of the feeder level multi-objective model is designed in four combinations, as shown in formula (11):

[0087] (11);

[0088] Wherein, represents the objective function of the upper feeder multi-objective optimization model; , , , , is the weight value of evaluating each sub-objective function; Value represents the main optimization target of selected feeder, value 0 means that the feeder selects the power self-consumption as the target, value 1 means that the feeder selects the voltage control (voltage deviation degree control) as the main target; Value represents the feeder operation mode, if 0, the feeder runs in normal mode, if 1, the feeder runs in the mode of substation heavy overload governance; is a feeder self-balancing degree evaluation function, is a feeder voltage deviation degree evaluation function, is a feeder voltage penalty function; is a feeder substation heavy load penalty function, is a feeder substation variable load rate balance degree evaluation function.

[0089] Feeder self-balancing degree evaluation function Based on the active and reactive power at the outlet of the feeder, the current self-balancing degree of the feeder is evaluated, the lower the sum of the absolute values of the active and reactive power at the outlet is, the better the index is, and the calculation formula is shown in formula (12):

[0090] (12);

[0091] Wherein, , are the active and reactive power values transmitted to the upper grid at the outlet of the feeder respectively.

[0092] Feeder voltage deviation degree evaluation function The voltage of each node in the feeder (taking the voltage amplitude) is used as the input quantity, the lower the sum of the deviation of each node voltage from the reference voltage is, the better the index is, and the calculation formula is shown in formula (13):

[0093] (13);

[0094] Wherein, represents the total number of nodes in the feeder; is the current traversal node number; represents the voltage value of node ; represents the reference voltage value.

[0095] Feeder voltage penalty function Based on the voltage of each node in the feeder, the voltage out-of-limit condition is evaluated, the node voltage is represented by the per-unit value, and the voltage out-of-limit penalty value is set as a step function under the condition of considering the voltage out-of-limit risk, and the calculation formula is shown in formula (14):

[0096] (14);

[0097] Wherein, 、 、 represents a voltage penalty parameter representing different voltage ranges, that is, the feeder voltage out-of-limit penalty value.

[0098] Feeder substation transformer overload penalty function Taking the load rate of each substation transformer as an input quantity, in some preferred embodiments, the load rate of the substation transformer reaching 80% to 100% is set as a heavy load condition of the substation transformer, and a penalty value is added if the heavy load condition occurs, and the calculation formula is shown as formula (15):

[0099] (15);

[0100] wherein, represents the total number of substation transformers of the feeder, represents the current iteration substation transformer number; represents the load rate of the substation transformer ; and represents a penalty value, that is, the feeder substation transformer overload penalty value.

[0101] Feeder substation transformer load rate balance degree evaluation function Also taking the load rate of each substation transformer as an input quantity, the average value thereof is calculated, and the distance between each transformer load rate and the average value is compared. The smaller the value is, the more excellent the index is, and the calculation formula is shown as formula (16) to formula (17):

[0102] (16);

[0103] (17);

[0104] wherein, represents the average value of the load rate of the substation transformer of the feeder.

[0105] The form of the node voltage, line current, and network power flow constraint of the upper-layer feeder multi-objective optimization model is the same as that of the lower-layer substation operation domain search multi-objective model, that is, formula (5), (6), and (7). The adjustable power constraint of the substation needs to consider the economy index at the outlet of the substation and the adjustable range of active and reactive power provided by the lower-layer substation, and the calculation formula is shown as formula (18) to formula (20):

[0106] (18);

[0107] (19);

[0108] (20);

[0109] wherein, , , represent active, reactive, and apparent power at the transformer substation; represent power factor at the transformer substation; , are upper and lower limits of the adjustable range of active power at the transformer substation outlet; , are upper and lower limits of the adjustable range of reactive power at the transformer substation outlet.

[0110] (Three) Lower-level transformer substation control model, adopts lower-level transformer substation low-voltage photovoltaic control model, also known as: lower-level transformer substation task allocation multi-objective model, the establishment steps are as follows:

[0111] S23a. Based on the active power and reactive power at the transformer substation outlet and the preset active power instruction and reactive power instruction, the deviation of the current transformer substation output power from the set value is obtained, and an instruction approximation degree evaluation function is established with the minimum approximation degree as the target ;

[0112] S23b. According to the over-limit situation of the voltage of each node in the transformer substation, set the transformer substation voltage over-limit penalty value, and establish transformer substation voltage penalty function two ;

[0113] S23c. Based on and and introducing a weight coefficient to obtain a third objective function ;

[0114] S234. Set the third constraint condition for the objective function , and obtain the lower-level transformer substation low-voltage photovoltaic control model. In one preferred embodiment, the third constraint condition has: node voltage, line current, network flow and distributed photovoltaic adjustable power constraint condition.

[0115] Specifically, the lower-level transformer substation task allocation multi-objective model takes the minimum deviation degree of the outlet power from the instruction and the no over-limit situation of the transformer substation voltage as the objective function, and the transformer substation network topology and the distributed photovoltaic adjustable power as the constraint condition, to construct the lower-level transformer substation low-voltage photovoltaic control multi-objective optimization model.

[0116] The objective function of the lower-level transformer substation task allocation multi-objective model is composed of the instruction approximation degree evaluation function and the transformer substation voltage penalty function , and its calculation formula is shown in formula (21):

[0117] (21);

[0118] Wherein, ​​a target function representing the lower-level substation task allocation multi-objective model; , represent weight values of sub-objective functions, respectively; is an instruction approximation degree evaluation function, is a substation voltage penalty function, that is, the substation voltage penalty function two.

[0119] Instruction approximation degree evaluation function Based on the active and reactive power at the substation outlet and the set instruction active and reactive power, the approximation degree is calculated to minimize the deviation between the current substation output power and the set value by respectively subtracting the two powers and taking the absolute value, and the calculation formula is shown in formula (22):

[0120] (22);

[0121] wherein, , represent active and reactive power at the substation outlet, respectively, and are adjustable variables of the current optimization step, that is, the output control instruction set; , represent active and reactive power regulation instructions sent by the upper feeder to the lower substation, respectively, and are reference set power given by the last optimization.

[0122] Substation voltage penalty function Based on the voltage of each node in the substation, it is evaluated whether there is voltage out-of-limit condition in the substation, and the calculation formula is shown in formula (23):

[0123] (23);

[0124] wherein, is the total number of nodes in the substation; is the current iteration node number; is a set voltage penalty parameter, that is, the substation voltage out-of-limit penalty value; is the voltage value of node .

[0125] The action object and control level of the lower-level substation task allocation multi-objective model are the same as those of the lower-level substation dynamic operation domain aggregation model, and the constraints are also consistent, that is, formula (5)-(10), which will not be repeated here.

[0126] According to the above obtained , , , the multi-stage distributed photovoltaic multi-objective optimization control model considering feeder-substation constraints is obtained after combination. Multi-stage refers to the solving and control stages of each model corresponding to the "lower-upper-lower" mechanisms.

[0127] S3. Solving the multi-stage distributed photovoltaic multi-objective optimization regulation model to obtain the optimal regulation scheme of each stage.

[0128] As shown in Figure 4 , the algorithm flow of the multi-stage distributed photovoltaic multi-objective optimization regulation model is given, which is as follows:

[0129] Stage one, the algorithm flow of the lower-layer substation dynamic operation domain aggregation model is as follows: taking the substation power flow information as the input quantity, setting the upper bound optimization target function, such as the target function of the lower-layer substation operation domain search multi-objective model , judging the optimization direction, and then establishing the corresponding optimization model according to the optimization direction and under the specific constraint conditions (5)-(10). In , the search direction of the selected interval is represented, if it is 1, the upper bound of the interval is searched, otherwise if it is -1, the lower bound of the interval is searched, and the upper and lower bounds correspond to different optimization models under the above constraints. Finally, the upper bound of the operation domain and the lower bound of the operation domain are outputted when the iteration is completed, and the substation operation domain boundary is obtained by combining them, which is the output quantity, and is the optimal regulation scheme of stage one.

[0130] Stage two, the algorithm flow of the upper-layer feeder multi-mode optimization regulation model is as follows: taking the substation operation domain boundary, the substation operation state, and the feeder power flow information as the input quantity, first judging whether the feeder is continuous, and if the substation is overloaded, it is divided into the feeder normal operation mode and the feeder substation overload treatment mode, then setting the self-consumption / voltage control target function (the first two stages of ) and the self-consumption / voltage control target function + substation variable overload treatment target function (the last two stages of ) for the two modes respectively, wherein value represents the selected main optimization target of the feeder, and the value of 0 represents that the feeder selects the feeder power self-consumption as the target, and the value of 1 represents that the feeder selects the voltage control as the main target; value represents the feeder operation mode, if it is 0, the feeder operates in the normal mode, if it is 1, the feeder operates in the substation overload treatment mode. Then, the corresponding optimization model is established under the specific constraint conditions (5), (6), (7), (18), (19), (20). Finally, the feeder control instruction is outputted when the iteration is completed, which is the output quantity, and is the optimal regulation scheme of stage two.

[0131] In a preferred embodiment, a feeder control instruction verification step is also given in the second stage process: analyze whether the feeder control instruction and the substation operating domain boundary are located within the boundary, if not, update the parameter information of the model, re-iterate the feeder control instruction; if yes, generate the feeder control instruction output.

[0132] The algorithm flow of the third stage, the lower substation low-voltage photovoltaic regulation model, is: taking the feeder control instruction as the input quantity, setting the optimization objective function, that is, , then establishing the corresponding optimization model under the specific constraint conditions (5)-(10). Finally, the optimization algorithm is used to solve and optimize, and when the iteration is completed, the photovoltaic aggregate control instruction (that is, , the optimal solution, is output, which is used to control the distributed photovoltaic output, also known as the distributed photovoltaic output control instruction), and the substation power flow information is updated according to the instruction as the output quantity, which is the optimal regulation scheme of the third stage.

[0133] In a preferred embodiment, the improved NSGA-III algorithm is used to solve the multi-stage distributed photovoltaic multi-objective optimization regulation model, and the improved NSGA-III algorithm is obtained by introducing the adaptive Latin hypercube sampling initialization mechanism and the reference point updating strategy based on historical information.

[0134] Specifically, the algorithm flow of each part of the sub-module (mechanism) in the feeder-substation closed-loop regulation strategy involves the step of using an optimization algorithm for solving. At the same time, the sub-models of each stage of the multi-stage distributed photovoltaic multi-objective optimization model considering the feeder-substation constraints all have the characteristics of high dimension and nonlinearity, which puts forward the requirements of high real-time and strong accuracy for the solving method. The third generation non-dominated genetic (NSGA-III) algorithm uses a reference point-based environmental selection method to replace the traditional non-dominated genetic algorithm based on the crowding distance, has better high-dimensional target processing capability and stronger convergence performance, and is suitable for solving high-dimensional nonlinear problems, but its calculation efficiency is restricted by the selection of initialization parameters and the setting of reference points, which is difficult to meet the requirements of online real-time regulation of the whole process. Therefore, this paper improves the NSGA-III algorithm by introducing the adaptive Latin hypercube sampling (LHS) initialization mechanism and the reference point updating strategy based on historical information, and improves its calculation efficiency and optimization performance. The algorithm structure before and after improvement is shown in Figure 5 .

[0135] S3 is implemented through the following sub-steps:

[0136] S31. Adaptive LHS initialization, also known as adaptive Latin hypercube sampling initialization. The adaptive Latin hypercube sampling initialization mechanism includes the following steps: variable range normalization, hierarchical partitioning, adaptive probability generation, sampling point generation, dimension combination, denormalization, and generation of the initial population.

[0137] Specifically, the existing NSGA-III algorithm uses a random initialization strategy, which may lead to initial population clustering, resulting in decreased optimization performance and even failure to find the global optimum within a finite number of iterations. In contrast to the random initialization strategy, adaptive LHS initialization combines an adaptive initialization mechanism with an LHS-based initialization mechanism. Building upon the multidimensional equally distributed decision space of the LHS initialization mechanism, it introduces an adaptive interval probability generation mechanism based on historical initialization data (i.e., ...). Figure 5 The adaptive probability generation in the algorithm improves population diversity and coverage, and also increases the convergence rate of the algorithm. The steps of adaptive LHS initialization are as follows: Figure 5 The block diagram of the adaptive LHS initialization algorithm is shown.

[0138] The variable range normalization step will transform the original Unifying the decision variable space of unequal dimensions into The [0,1] dimension space; based on this, if the hierarchical partitioning step is based on the initial population size... ,Will Each dimension in 3D space is divided into A series of equal-length intervals, forming a... A length of The normalized decision space is formed by the combination of intervals.

[0139] The adaptive probability generation step first creates a buffer with a capacity to store the initial population and its convergence count for 10 iterations, outputting the default probability during the first 10 iterations. After the first 10 iterations, the selected probability of the interval corresponding to the population in the buffer is modified according to equations (24) and (25).

[0140] (twenty four);

[0141] (25);

[0142] in, This indicates the number of times interval k is selected during the initialization of the population stored in the buffer; This represents the total number of intervals in the dimension corresponding to interval k; This represents the selection probability after the interval k is updated.

[0143] The selection probability of the remaining intervals is modified according to equation (26).

[0144] (26);

[0145] wherein, denotes the updated selected probability of the rest of the intervals except the selected interval.

[0146] When the buffer is full, the convergence number of the population in the buffer is compared with the convergence number of the current population at the end of each iteration calculation. If the convergence number of the current population is lower than the convergence number of the population in the buffer, the buffer is updated (replaced by the current population), otherwise the original population is maintained.

[0147] The sampling point generation step selects a sampling set from the interval set corresponding to each dimension according to the selected probability of the interval given by the adaptive probability generation step, and randomly samples in the set to generate sampling numbers with values between 0 1. The dimension combination step combines the sampling array to generate a normalized population value vector, and after denormalization, the final initialization population is obtained.

[0148] S32, reference point update based on historical information. The reference point update strategy based on historical information includes the following steps: creating a historical information storage structure, sliding window trend analysis, reference point utility evaluation, reference point update, and generating a new reference point set.

[0149] Specifically, compared with the strategy of fixing the reference point in the iteration process of the existing NSGA-III algorithm, the reference point (the reference point of the optimal solution update in the algorithm) update strategy based on historical information can fully utilize the multi-generation information in the iteration process, predict the advancing trend of the Pareto front, and improve the convergence performance of the model. The algorithm structure is as shown in the reference point update strategy module based on historical information in Figure 5 .

[0150] Figure 5 In the reference point update strategy module based on historical information in , the historical information storage structure step creates a sliding window with a span of K iterations according to the set parameters for storing the reference point information and action information of each generation. The sliding window trend analysis step first performs weighted processing on the populations of each generation in the window, and then generates a trend vector by comprehensively considering the target weight setting.

[0151]

[0152] (27);

[0153] wherein, denotes the current iteration round the evaluation vector corresponds to the reference point a component of the evaluation vector, denotes the last iteration round denotes the reference point corresponding to the evaluation vector at the last iteration round a component of the evaluation vector; denotes a dynamic update factor of the utility of the reference point; denotes the current iteration round denotes the amount of canonical improvement of the solution associated to the reference point at the current iteration round denotes the current iteration round denotes the amount of canonical improvement of the solution of all reference points at the current iteration round

[0154] The reference point update step is based on the evaluation vector , the existing reference points are divided into high-utility reference points and low-utility reference points: the high-utility reference points are retained and updated in the direction of the trend vector; the low-utility reference points are randomly retained or replaced with a certain probability. Specifically, if the value of the evaluation vector value associated with the solution of the reference point is lower than 0, it is considered as low-utility.

[0155] A high-penetration distributed photovoltaic feeder area closed-loop regulation method is given according to the above embodiment, and a high-penetration distributed photovoltaic feeder area closed-loop regulation device is provided for regulation of a medium-low voltage power distribution network. The upper layer of the medium-low voltage power distribution network is a medium voltage feeder level system, and the lower layer is a plurality of low voltage areas containing high-penetration distributed photovoltaic at different positions of the feeder. The system comprises: a support framework modeling module, which is used to introduce a feeder regulation mechanism in the upper layer, and introduce an area aggregation mechanism and an area regulation mechanism in the lower layer. The feeder regulation mechanism, the area aggregation mechanism and the area regulation mechanism form a whole-process regulation closed loop to obtain a two-layer collaborative support framework of the feeder-area high-penetration distributed photovoltaic with whole-process closed-loop control. A regulation model construction module is used to establish an upper feeder regulation model for the feeder regulation mechanism, a lower area regulation model for the area aggregation mechanism, and a lower area aggregation model for the area regulation mechanism based on the two-layer collaborative support framework, and a multi-stage distributed photovoltaic multi-objective optimization regulation model is obtained after combination. A regulation scheme output module is used to solve the multi-stage distributed photovoltaic multi-objective optimization regulation model to obtain the optimal regulation scheme of each stage.

[0156] The effectiveness of the high-penetration distributed photovoltaic feeder area closed-loop regulation method, device and improved NSGA-III algorithm provided by the above embodiment of the application will be verified based on a feeder-area two-layer power distribution network model built according to a reference actual topology. The feeder-area two-layer power distribution network model takes a certain actual feeder and its mounted area in Zhejiang as a reference. The voltage level of the upper feeder model is 10kV. After equivalence, the feeder contains one feeder outlet transformer, 36 nodes and four branches, and its topology model is as follows: Figure 6The results are shown in the figures. By the high-permeability distributed photovoltaic feeder area closed-loop regulation method and device provided by the above-embodiment of the application, the following results are obtained by taking the nodes 8, 15, 24, 31 and 36 with the distributed photovoltaic area as examples: Figures 7 to 11 The figures are the active operation domain boundary and output-time variation curves of the lower controllable area nodes 8, 15, 24, 31 and 36 in a typical day; Figures 12 to 16 The figures are the reactive operation domain boundary and output-time variation curves of the lower controllable area nodes 8, 15, 24, 31 and 36 in a typical day; Figure 17 The figure is the feeder voltage offset-time variation curve of the lower controllable area nodes 8, 15, 24, 31 and 36 in a typical day before and after optimization; Figure 18 The figure is the feeder output active / reactive power and the lower controllable area node 8, 15, 24, 31 and 36 area variable load rate-time variation curve in a typical day before and after optimization in normal mode / governance mode; Figure 19 The figure is the node 8 controllable area photovoltaic active output / area outlet target output-time variation curve in a typical day; Figure 20 The figure is the node 8 controllable area photovoltaic reactive output / area outlet target output-time variation curve in a typical day.

[0157] Next, the control index comparison results of the regulation method of the embodiment of the application and other control methods are given:

[0158] Table 1: Comparison table of the running results of the algorithm of the application and the existing algorithm in the voltage control mode of the feeder-area model

[0159]

[0160] Table 2: Comparison table of the running results of the algorithm of the application and the existing algorithm in the self-consumption mode of the feeder-area model

[0161]

[0162] In Table 1, the voltage control mode is compared with the running results of the algorithm and the existing algorithm by the voltage overrun rate and the total voltage deviation of the feeder. The average total voltage deviation of the algorithm is 0.1386, which is still greatly improved compared with the optimal average total voltage deviation 0.2144 obtained by the existing algorithm, indicating that the algorithm has stronger optimization ability than the existing algorithm. In Table 2, the voltage overrun rate of the algorithm running in the feeder self-consumption mode is 0%, which is more excellent than the existing single algorithm; the average output active power and reactive power at the outlet of the feeder of the algorithm are-0.061 MW and 0.005 MVar, which are significantly better than the single algorithm, and are more excellent than the results-0.103 MW, 0.014 MVar and-0.073 MW, 0.009 MVar obtained by the existing advanced algorithm. The calculation time of the algorithm in the two modes is 37.6s and 31.8s respectively, which is shorter than the calculation time of most existing algorithms, indicating that the algorithm in the present application has high calculation efficiency while ensuring the accuracy of the optimization result, and also indicating that the algorithm in the present application can adapt to large-scale examples while ensuring the optimization control effect and maintaining high calculation efficiency.

[0163] In summary, compared with the prior art, based on the above embodiments, the present application can achieve the following beneficial technical effects:

[0164] (1) Compared with the traditional day-ahead-day-in double-layer control architecture for distributed photovoltaic clusters, the present application provides a high penetration rate distributed photovoltaic feeder area closed-loop regulation method and device, which is a double-layer collaborative support framework of high penetration rate distributed photovoltaic feeder-area through full-process closed-loop control, that is, a "bottom-up-bottom" regulation framework, which further expands the regulation level to the low-voltage side, enhances the closed-loop interaction logic between the medium and low-voltage levels, and specifically aggregates the resources of the area through the area aggregation mechanism, centrally regulates the feeder through the feeder regulation mechanism, and performs control in the control mode of area decomposition execution through the area regulation mechanism, realizes the closed-loop precise control of the medium and low-voltage levels, fully excavates the active / reactive power active support potential of distributed photovoltaic, and realizes the real-time and efficient collaborative optimization regulation method of "centralized calculation-distributed execution" through the low-voltage distributed photovoltaic cluster aggregation operation domain and instruction distribution mechanism, and fully solves the problems of insufficient timeliness and lack of feedback in the execution of the traditional method.

[0165] (2) To solve the problem of the traditional NSGA-III algorithm in initialization uniformity and reference point dynamicity, which leads to the decline of convergence-distribution performance in high-dimensional multi-objective optimization and the limitation of real-time performance, LHS-HNSGA-III is proposed. Firstly, the Latin hypercube sampling is used to replace the random initialization, which generates individuals with low difference and high coverage in the decision space, so as to quickly approach the real front and reduce redundant evaluation. Secondly, the adaptive updating strategy of reference points based on historical information is embedded in the environmental selection stage: the distribution density and evolution direction of the discovered non-dominated solutions are memorized by the elite archive, and the redundant reference points in the aggregation area are periodically removed and new reference points are inserted in the sparse area, so as to realize the dynamic matching of the reference set and the Pareto front shape. Experiments show that in the real-time scheduling example of distributed photovoltaic power in distribution network, the improved algorithm has a significant increase in computational efficiency and a rapid decrease in single iteration time compared with the original algorithm, which can fully adapt to the multi-objective optimization scene with high requirements for real-time performance and accuracy in the real-time regulation scene of new energy.

[0166] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiment methods. The storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM) or a random access memory (RAM), etc.

[0167] The technical features of the above-mentioned embodiments can be combined arbitrarily. To make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described, but as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application. The above-mentioned embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be considered as a limitation of the scope of the present application. Those skilled in the art can make some modifications and improvements without departing from the concept of the present application, which are all within the protection scope of the present application.

Claims

1. A closed-loop control method for high-penetration distributed photovoltaic (PV) feeder areas, used for the control of medium- and low-voltage distribution networks, wherein the upper layer of the medium- and low-voltage distribution network is a medium-voltage feeder-level system, and the lower layer consists of several low-voltage distribution areas containing high-penetration distributed PV systems connected to different locations of the feeder, characterized in that... The method includes the following steps: A feeder control mechanism is introduced at the upper layer, and a distribution area aggregation mechanism and a distribution area control mechanism are introduced at the lower layer. The feeder control mechanism, the distribution area aggregation mechanism, and the distribution area control mechanism form a closed-loop control system for the entire process, thereby obtaining a two-layer collaborative support framework for high-penetration distributed photovoltaic power generation with a closed-loop control system. The closed-loop control system is as follows: the distribution area aggregation mechanism obtains power flow information from the distribution area control mechanism and generates distribution area control information including the boundary of the distribution area operating domain, which is then transmitted to the feeder control mechanism and the distribution area control mechanism; the feeder control mechanism generates feeder command information based on the distribution area control information and sends it to the distribution area control mechanism; the distribution area control mechanism allocates the output of distributed photovoltaic power generation in the distribution area based on the command information and the distribution area control information, updates the power flow information in the distribution area, and uploads it to the distribution area aggregation mechanism. Based on the two-layer collaborative support framework, an upper-layer feeder control model is established for the feeder control mechanism, a lower-layer transformer area control model is established for the transformer area aggregation mechanism, and a lower-layer transformer area aggregation model is established for the transformer area control mechanism. After combination, a multi-stage distributed photovoltaic multi-objective optimization control model is obtained. Solve the multi-stage distributed photovoltaic multi-objective optimization control model to obtain the optimal control scheme for each stage.

2. The closed-loop control method for high-penetration distributed photovoltaic feeder areas according to claim 1, characterized in that, The aforementioned low-voltage distribution areas include at least three low-voltage distribution areas equipped with adjustable distributed photovoltaics and whose photovoltaic penetration rate decreases sequentially, and one low-voltage distribution area without distributed photovoltaics; the upper and lower layers exchange power through a transformer gate.

3. The closed-loop control method for high-penetration distributed photovoltaic feeder areas according to claim 1, characterized in that, In the step of establishing an upper-level feeder control model based on a dual-layer collaborative support framework, a lower-level transformer area control model based on a feeder control mechanism, and a lower-level transformer area aggregation model based on a transformer area aggregation mechanism, and then combining them to obtain a multi-stage distributed photovoltaic multi-objective optimization control model, the lower-level transformer area aggregation model adopts a lower-level transformer area dynamic domain aggregation model, and the establishment steps include: The upper bound evaluation function and the lower bound evaluation function of the adjustable interval are constructed with the goal of maximizing the sum of the adjustable range of the upper and lower bounds of the active power at the transformer outlet and the adjustable range of the upper and lower bounds of the reactive power at the transformer outlet. Based on the voltage over-limit situation of each node in the transformer area, set the voltage over-limit penalty value of the transformer area and establish the transformer area voltage penalty function one; Based on the adjustable interval upper bound evaluation function, the adjustable interval lower bound evaluation function, and the transformer area voltage penalty function, a weighting coefficient is introduced and combined with the interval search direction to obtain the first objective function; Set the first constraint condition for the first objective function to obtain the dynamic domain aggregation model of the lower-level transformer area; the first constraint condition includes: node voltage, line current, network power flow and adjustable power of distributed photovoltaic.

4. The closed-loop control method for high-penetration distributed photovoltaic feeder areas according to claim 1, characterized in that, In the step of establishing an upper-level feeder control model based on a dual-layer collaborative support framework, a lower-level transformer area control model based on a feeder control mechanism, and a lower-level transformer area aggregation model based on a transformer area control mechanism, and then combining them to obtain a multi-stage distributed photovoltaic multi-objective optimization control model, the upper-level feeder control model adopts an upper-level feeder multi-mode optimization control model, and the establishment steps include: A feeder self-balance evaluation function is established with the objective of minimizing the sum of the absolute values ​​of active power and reactive power at the feeder outlet. A function for evaluating feeder voltage offset is established with the goal of minimizing the sum of the offsets between the voltage at each node and the reference voltage. Based on the voltage exceedance situation of each node in the feeder, the feeder voltage exceedance penalty value is set in stages, and the feeder voltage penalty function is established; Set the overload penalty value for the feeder transformer area based on the load rate of each transformer area, and establish the overload penalty function for the feeder transformer area. A function for evaluating the load balance of feeder transformer areas is established with the goal of minimizing the distance between the load rate of each transformer area and the average load rate of all transformer areas. Based on the feeder self-balance evaluation function, feeder voltage deviation evaluation function, feeder voltage penalty function, feeder substation area overload penalty function, and feeder substation area load rate balance evaluation function, a weighting coefficient is introduced and combined with the selected feeder main optimization objective and feeder operation mode to obtain the second objective function. A second constraint is set for the second objective function to obtain the multi-mode optimization and control model of the upper feeder; the second constraint includes: node voltage, line current, network power flow and adjustable power of the transformer area.

5. The closed-loop control method for high-penetration distributed photovoltaic feeder areas according to claim 1, characterized in that, In the step of establishing an upper-level feeder control model based on a dual-layer collaborative support framework, a lower-level transformer area control model based on a feeder control mechanism, and a lower-level transformer area aggregation model based on a transformer area aggregation mechanism, and then combining them to obtain a multi-stage distributed photovoltaic multi-objective optimization control model, the lower-level transformer area control model adopts a lower-level transformer area low-voltage photovoltaic control model, and the establishment steps include: Based on the active and reactive power at the output of the transformer area and the preset active and reactive power commands, the deviation between the current output power of the transformer area and the set value is obtained, and a command approximation evaluation function is established with the goal of minimizing the approximation degree. Based on the voltage over-limit situation of each node in the transformer area, set the voltage over-limit penalty value of the transformer area and establish the transformer area voltage penalty function 2; The third objective function is obtained by combining the instruction approximation evaluation function and the transformer area voltage penalty function with weighting coefficients. A third constraint is set for the third objective function to obtain the low-voltage photovoltaic control model for the lower-level distribution area; the third constraint includes: node voltage, line current, network power flow and adjustable power of distributed photovoltaic.

6. The closed-loop control method for high-penetration distributed photovoltaic feeder areas according to claim 1, characterized in that, In the step of solving the multi-stage distributed photovoltaic multi-objective optimization control model and obtaining the optimal control scheme for each stage, the improved NSGA-III algorithm is used to solve the multi-stage distributed photovoltaic multi-objective optimization control model. The improved NSGA-III algorithm is specifically obtained by introducing an adaptive Latin hypercube sampling initialization mechanism and a reference point update strategy based on historical information to improve the NSGA-III algorithm.

7. The closed-loop control method for high-penetration distributed photovoltaic feeder areas according to claim 1, characterized in that, In the step of solving the multi-stage distributed photovoltaic multi-objective optimization control model and obtaining the optimal control scheme for each stage, the optimization algorithm is used to solve and optimize. When the iteration is completed, the boundary of the operating domain of the transformer area, the feeder control command and the distributed photovoltaic power output control command are output in stages, which are respectively used as the optimal control scheme for each stage.

8. The closed-loop control method for high-penetration distributed photovoltaic feeder areas according to claim 6, characterized in that, In the step of solving the multi-stage distributed photovoltaic multi-objective optimization control model and obtaining the optimal control scheme for each stage, a feeder control command verification step is set: analyze whether the feeder control command and the boundary of the transformer area operation domain are within the boundary. If not, update the parameter information of the model and iterate the feeder control command again. If so, then a feeder control command output is generated.

9. A closed-loop control device for a high-penetration distributed photovoltaic (PV) feeder area, used for the control of a medium- and low-voltage distribution network, wherein the upper layer of the medium- and low-voltage distribution network is a medium-voltage feeder-level system, and the lower layer consists of several low-voltage distribution areas containing high-penetration distributed PV systems connected to different locations of the feeder, characterized in that... The device includes: The supporting framework construction module is used to introduce a feeder control mechanism at the upper layer and a transformer area aggregation mechanism and a transformer area control mechanism at the lower layer. The feeder control mechanism, transformer area aggregation mechanism, and transformer area control mechanism form a closed-loop control system for the entire process, thereby obtaining a two-layer collaborative support framework for feeder-transformer area high-penetration distributed photovoltaic power generation with a closed-loop control system. The closed-loop control system is as follows: the transformer area aggregation mechanism obtains transformer area power flow information from the transformer area control mechanism and generates transformer area control information containing the transformer area operating domain boundary, which is then transmitted to the feeder control mechanism and the transformer area control mechanism; the feeder control mechanism generates feeder command information based on the transformer area control information and sends it to the transformer area control mechanism; the transformer area control mechanism allocates the output of the distributed photovoltaic power generation system based on the command information and the transformer area control information, updates the transformer area power flow information, and uploads it to the transformer area aggregation mechanism. The regulation model construction module is used to establish an upper-level feeder regulation model based on the two-layer collaborative support framework, a lower-level transformer area regulation model based on the feeder regulation mechanism, and a lower-level transformer area aggregation model based on the transformer area regulation mechanism. After combination, a multi-stage distributed photovoltaic multi-objective optimization regulation model is obtained. The regulation scheme output module is used to solve the multi-stage distributed photovoltaic multi-objective optimization regulation model and obtain the optimal regulation scheme for each stage.

Images