CN117955121A - Power distribution network reactive voltage control method based on data driving and source load uncertainty - Google Patents
Power distribution network reactive voltage control method based on data driving and source load uncertainty Download PDFInfo
- Publication number
- CN117955121A CN117955121A CN202410356221.6A CN202410356221A CN117955121A CN 117955121 A CN117955121 A CN 117955121A CN 202410356221 A CN202410356221 A CN 202410356221A CN 117955121 A CN117955121 A CN 117955121A
- Authority
- CN
- China
- Prior art keywords
- model
- network
- data
- reactive
- power
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000009826 distribution Methods 0.000 title claims abstract description 39
- 238000005457 optimization Methods 0.000 claims abstract description 37
- 230000008569 process Effects 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 16
- 239000003990 capacitor Substances 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000011217 control strategy Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000009827 uniform distribution Methods 0.000 claims description 3
- 230000003213 activating effect Effects 0.000 claims description 2
- 230000006870 function Effects 0.000 description 13
- 230000008859 change Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 9
- 230000009471 action Effects 0.000 description 6
- 230000001105 regulatory effect Effects 0.000 description 4
- 238000010248 power generation Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/30—Reactive power compensation
Landscapes
- Control Of Electrical Variables (AREA)
Abstract
The invention discloses a reactive voltage control method of a power distribution network based on uncertainty of data driving and source load, which belongs to the technical field of operation optimization of a power system.
Description
Technical Field
The invention belongs to the technical field of operation optimization of power systems, and particularly relates to a reactive voltage control method of a power distribution network with double time scales based on data driving and source load uncertainty.
Background
The traditional power generation mode of power generation by using fossil energy causes continuous deterioration of the climate environment, and simultaneously the total amount of available primary energy is continuously reduced. The problem can be effectively avoided by using clean energy to generate power, the installed capacity of the power supply represented by wind power and photovoltaic has obvious rising trend in recent years, and along with the continuous development of the renewable energy power generation field, a large number of distributed power supplies such as wind power and photovoltaic are connected into the power distribution network, so that remarkable benefits are brought to a power system and an ecological environment. Because the distributed power supply has strong randomness and volatility, a large number of distributed power supplies are connected into the power distribution network, so that a large amount of influence is brought to the safe and stable operation of the power system, and meanwhile, the power distribution network can possibly have voltage violations and other problems along with the increase of load demands and the improvement of volatility, therefore, the distributed power supply and other controllable equipment in the power distribution network are required to be controlled, and the purpose of reducing the negative influence of the distributed power supply on the power distribution network is achieved.
With the continuous development of the machine learning field, many researches introduce a data-driven model into a voltage optimization control strategy and obtain ideal results. The traditional voltage optimization control method is realized based on a physical model, but the solution model is non-convex, so that the solution difficulty is high, and the method can face the problems of complex modeling, high dimensionality, slow solution and the like for a complex network or a network with more controllable devices. The data driving model has strong fitting capability and has the advantage of offline training on-line application, and can solve the problem of difficult solution. In addition, as the action characteristics of controllable equipment in the power distribution network are different, the voltage control problem on different time scales exists, and the control equipment on different time scales are mutually influenced, the unified modeling and solving through the traditional physical model are difficult.
Therefore, a new technical solution is needed in the prior art to solve the above-mentioned problems.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: the method comprises the steps of dividing a control strategy into two stages through different time scales, namely day-ahead scheduling and real-time daily optimization control, wherein the day-ahead scheduling in the first stage adopts a centralized optimization method to model, so that network loss is minimized while voltage safety is met, the real-time daily optimization control in the second stage is based on a data driving model, the model realizes a voltage regulating function in an area, the data driving model is trained offline by using a power supply, load data and a voltage regulating strategy generated based on WGAN-GP model, and a better voltage regulating effect is obtained and network loss is reduced through a two-stage double-time-scale control method.
The reactive voltage control method of the power distribution network based on the uncertainty of data driving and source load comprises the following steps,
Step one, an improved generation countermeasure network WGAN-GP model is established, real sample data x is input, WGAN-GP model parameters are set, and the model is trained to the maximum iteration times;
in the method, in the process of the invention, Representing a loss function consisting of generator G and arbiter D together, E [. Cndot. ] representing a mathematical expectation, being a gradient operator,/>Random interpolation for real data and generated data; λ represents the weight of the gradient penalty term; /(I)For 2 norms, D (-) represents a discriminant model, G (-) represents a generator model, the generator and the discriminant are realized through a neural network, the generator and the discriminant model are built by adopting a deep convolution network, z is input noise data, x is sample data conforming to real historical data P data (x), and/>For sample distribution of real data,/>Is a random value subject to a uniform distribution of [0,1 ];
Step two, a centralized reactive voltage optimization model is established, the centralized reactive voltage optimization method carries out optimal power flow calculation in a centralized controller through photovoltaic and load data in each hour to obtain a result of reactive output of an on-load tap-changer OLTC, a capacitor bank CB and a photovoltaic PV, and carries out controllable equipment dispatching in a long time scale to reduce active power loss in a power distribution network;
step three, a distributed reactive voltage control model is established, wherein the controller establishes a network model based on a data-driven deep neural network DNN structure, and the network model is as follows:
in the method, in the process of the invention, 、/>And/>、/>The weight and bias matrix of the ith layer-1 and the ith layer are respectively, h i,in and h i,out are respectively the input and output of the ith layer network, C is the total layer number in the network, and/ >For activating a function, when i=c, h C,out is the output result of the current network output layer;
Taking x and y as the input and output of the whole network, beta represents the mapping relation between DNN network input and output, and simplifying the network model into
Based on WGAN-GP generated source load data, carrying out MISOCP model solving for a plurality of times to obtain a controllable equipment voltage regulation result, storing the solving result as sample data of a training area controller, and determining input and output characteristics of the model according to a formula (5);
in the method, in the process of the invention, Respectively represent j-th PV active force and capacity in Zn area,/>Reactive load for Zn and nodes in the area adjacent to Zn,/>Reactive output power for the ith PV in the Zn region;
And training the deep neural network model offline, putting the trained regional control into online application, and combining with centralized distributed control to form a reactive voltage optimization control strategy with double time scales.
The optimal power flow calculation model is as follows:
the formulas (6) and (7) are node active and reactive power balance equations, the formula (8) is a voltage relation between two connected nodes, and the formula (9) represents a relation between branch current, power and node voltage;
in the method, in the process of the invention, 、/>First-segment active power transmitted by ij branch at t moment,/>Active power and reactive power transmitted by the lines of the nodes j and k at the moment t are respectively; /(I)Branch current, resistance and reactance; the active power and the reactive power are loaded for the j node at the moment t; /(I) The active power and the reactive power output by the photovoltaic output by the j node at the moment t and the reactive power output by the capacitor bank are respectively; /(I)Is the voltage of the i-node at time t,The voltage of the j node at the time t is represented; /(I)The branch end and head end node sets with the node j as the head end and end nodes respectively.
Through the design scheme, the invention has the following beneficial effects: according to the reactive voltage control method of the power distribution network based on the uncertainty of data driving and source load, in the first stage, centralized optimization realizes the dispatching of slow-adjusting equipment (OLTC, CB) through daily optimization, and in the second stage, through information exchange between regional controllers, reactive optimization control of photovoltaic in a region is realized, and network loss is reduced while voltage out-of-limit is avoided. The method effectively relieves the problem of voltage out-of-limit caused by source load fluctuation in the power distribution network, the distributed control can reduce the calculated amount and the communication cost, and compared with the centralized control, the execution time of the distributed control in the same period is reduced by 96.52%. The method can realize the mutual coordination among different time scales of the control equipment, and the optimization control of the short time scale can compensate the error of the long time scale scheduling result, compared with a centralized optimization method, the method meets the voltage safety constraint and reduces the network loss by 47.9% in the same time period.
Drawings
The invention is further described below with reference to the drawings and detailed description.
Fig. 1 is a flow chart of a reactive voltage control method of a power distribution network based on data driving and source load uncertainty.
Fig. 2 is a block diagram of a reactive voltage control method for generating an countermeasure network for a power distribution network based on data driving and source load uncertainty.
Fig. 3 is a deep neural network structure diagram of the reactive voltage control method of the power distribution network based on data driving and source load uncertainty.
Fig. 4 is a diagram illustrating an exemplary IEEE33 node architecture for accessing a controllable device according to an embodiment of the present invention.
Fig. 5 is a graph of photovoltaic output versus load daily variation in accordance with an embodiment of the present invention.
Fig. 6 is a switching state diagram of the capacitor bank at each moment according to the embodiment of the present invention.
Fig. 7 is a graph showing the shift change of OLTC according to an embodiment of the present invention.
FIG. 8 is a graph illustrating a node voltage distribution diagram after centralized optimization in accordance with an embodiment of the present invention.
Fig. 9 is a diagram of a photovoltaic generation data box in accordance with an embodiment of the present invention.
FIG. 10 is a diagram of a load creation data box according to an embodiment of the present invention.
FIG. 11 is a graph of a loss function of a WGAN-GP generator in accordance with an embodiment of the present invention.
Fig. 12 is a graph showing the convergence of a DNN area controller according to an embodiment of the present invention.
FIG. 13 is a graph showing the voltage distribution after period optimization from 9:00 to 12:00 according to the embodiment of the present invention.
Detailed Description
The reactive voltage control method of the power distribution network based on the uncertainty of data driving and source load, as shown in fig. 1, comprises the following steps:
step one, building and training WGAN-GP models according to distributed photovoltaic and load historical data x, and generating a source load scene set according to the trained models. Fig. 2 is a diagram of a structure of a generated countermeasure network.
The WGAN-GP model is expressed as:
in the method, in the process of the invention, Representing a loss function consisting of generator G and arbiter D together, E [. Cndot. ] representing a mathematical expectation, being a gradient operator,/>Random interpolation for real data and generated data; λ represents the weight of the gradient penalty term; /(I)In the invention, a deep convolution network is adopted to establish the generator and the discriminant model, z is input noise data, x is sample data conforming to real historical data P data (x), and the model is composed of a model of 2 norms, D (-) represents a model of the discriminant, G (-) represents a model of the generator and the discriminant are usually realized through a neural networkFor sample distribution of real data,/>Is a random value subject to a uniform distribution of 0, 1.
Setting WGAN-GP model parameters, and determining the network structures of the generator G and the discriminator D in WGAN-GP, and the maximum iteration times and the learning rate of the model.
And inputting the real data x into WGAN-GP model for iterative training, and generating sample data of load real data distribution.
And secondly, establishing a centralized reactive voltage optimization model, wherein the centralized reactive voltage optimization is to perform optimal power flow calculation in a centralized controller through photovoltaic and load data of each hour to obtain results of OLTC, CB and PV reactive output, so as to perform long-time scale controllable equipment dispatching, ensure voltage quality and reduce active power loss in a power distribution network. And adopting a branch flow model as flow constraint of the power distribution network, wherein the branch flow model is expressed by the following formula:
the formulas (6) and (7) are node active and reactive power balance equations, the formula (8) is a voltage relation between two connected nodes, and the formula (9) represents a relation between branch current, power and node voltage.
In the method, in the process of the invention,、/>First-segment active power transmitted by ij branch at t moment,/>Active power and reactive power transmitted by the lines of the nodes j and k at the moment t are respectively; /(I)Branch current, resistance and reactance; the active power and the reactive power are loaded for the j node at the moment t; /(I) The active power and the reactive power output by the photovoltaic output by the j node at the moment t and the reactive power output by the capacitor bank are respectively; /(I)Is the voltage of the i-node at time t,The voltage of the j node at the time t is represented; /(I)The branch end and head end node sets with the node j as the head end and end nodes respectively.
To ensure voltage quality, constraints are established to ensure that the voltage is within a safe range.
In the method, in the process of the invention,、/>Respectively minimum and maximum limit values of voltage safety constraint,/>The per-unit value of the voltage of the i node at the moment t.
Photovoltaic capacity limitation constraints are established.
In the method, in the process of the invention,、/>Active power and reactive power output of ith photovoltaic at time t,/>, respectivelyIs the maximum capacity of the ith photovoltaic.
Operational constraints of the capacitor bank are established.
In the method, in the process of the invention,For the maximum number of actions of the ith capacitor bank in the T period,/>And/>The state and the capacity of the kth capacitor of the ith capacitor bank, respectively,/>The reactive output power of the ith capacitor bank.
And establishing the operation constraint of the on-load tap-changer.
In the method, in the process of the invention,For the t moment OLTC gear state,/>For the maximum number of actions of OLTC within T period,/>Is the per unit value of the primary side voltage,/>Is the per unit value of the secondary side voltage,/>Voltage variation for one gear.
In order to facilitate the solution, the model is subjected to convex transformation, and equations (6) - (9), (14) - (15) give constraint for solving the optimal power flow of the power distribution network with controllable equipment, but due to the existence of quadratic terms and absolute values, the current optimization problem belongs to mixed integer nonlinear programming (MINLP), the solution difficulty is high, and in order to simplify the optimization problem, the second-order cone relaxation principle and the linearization method are utilized to convert the optimization problem into mixed integer second-order cone programming (MISOCP), so that the solution can be conveniently carried out by a solver. Converting formulae (8) - (15) into formulae (16) - (21), specifically expressed as follows:
in the formulas (16) to (21), Representing the square of ij branch current at time t,/>Is the square of the voltage at inode at time t,/>Is the square of the voltage at node j at time t,/>Are binary variables that all assist in linearization,/>Represents the state change of CB at times t and t+1,/>Is an integer variable between [ -10,10] and represents the tap change at times t and t+1,/>For the maximum OLTC tap position, h is the OLTC tap position variable, and the value range is [0 ]/>]，/>Is an auxiliary integer variable between [ -10,10] representing the gear change state of OLTC.
The MISOCP optimization problem constraint after conversion is formed by formulas (6) - (8), (11) - (12), (16) - (21) through second order cone conversion and linearization.
An objective function is established. The objective of this optimization problem is to minimize the total network loss while ensuring that the voltage is not violated, the objective function is shown in equation (22).
And thirdly, establishing a distributed reactive voltage optimization model.
And establishing a regional controller model based on the deep neural network, setting model parameters, and determining a network structure, the maximum iteration times and the learning rate.
The controller is based on a data-driven Deep Neural Network (DNN), the network structure of the DNN is simpler, and the calculation method comprises the following steps:
in the method, in the process of the invention, The weight and bias matrix of the ith layer-1 and the ith layer respectively, h i,in and h i,out are the input and output of the ith layer network respectively, C is the total layer number in the network,/>To activate the function, when i=c, h C,out is the output result of the current network output layer.
The above process is simplified into a function to be expressed, x and y are the input and output of the whole network, beta represents the mapping relation between the input and output of the DNN network, the model is expressed as formula (4), the structure diagram of the model is shown in fig. 3, in the figure, x i represents the ith input variable of the network, and y i represents the ith output variable of the network.
Training sample data is obtained. The DNN model training requires a large amount of sample data x and y, the source of the sample data is realized by calculating the optimal power flow in the power distribution network, the photovoltaic generated by the WGAN-GP model, a load sample and a real sample are input into the power distribution network with controllable equipment, each sample is repeatedly calculated, and the state and the power flow result of the controllable equipment are stored.
The photovoltaic active power output level in the area and the reactive load level of the adjacent area are selected as DNN input, each PV reactive power in the area is taken as DNN output, and a sample set used for training DNN can be selected from an off-line calculated optimal power flow result and expressed as follows by a formula:
in the method, in the process of the invention, Respectively represent j-th PV active force and capacity in Zn area,/>Reactive load for Zn and nodes in the area adjacent to Zn,/>Reactive output power for the ith PV in the Zn region.
Based on WGAN-GP generated source load data, carrying out MISOCP model solving for a plurality of times to obtain a controllable equipment voltage regulation result, storing the solving result as sample data of a training area controller, and determining input and output characteristics of the model according to a formula (5).
A loss function is established. The parameter updating of DNN is realized by calculating a loss function and counter-propagating gradient, wherein the loss function of the network is the mean square error of the reactive power output by the network and the actual optimal reactive power, and the mean square error is expressed as the following formula:
Wherein the method comprises the steps of Representing the optimal reactive power output samples calculated by the optimal power flow.
And training the deep neural network model offline, putting the trained regional control into online application, and combining with centralized distributed control to form a reactive voltage optimization control strategy with double time scales.
The simulation analysis of the method of the invention is as follows:
Table 1: controllable equipment parameter in power distribution network
| Apparatus and method for controlling the operation of a device | Parameters (parameters) | Position of |
| OLTC | ±10×1% | 1 |
| CB | 3*50kvar | 16，22 |
| PV | 1000kVA | 3，6，10，14，18，19，25，28，33 |
The embodiment of the invention adopts an improved IEEE33 node model, and FIG. 4 is an improved IEEE33 node model, and CB, OLTC and a plurality of distributed photovoltaics are connected on the basis of the original IEEE33 node model. The specific parameters are shown in Table 1. In the figure, the network is divided into three areas, each area is provided with an area controller, and when the first-stage day-ahead dispatching is performed, the centralized controller dispatches controllable equipment in the whole network, and then the second-stage day-ahead real-time optimized control is realized through the area controllers. In the calculation example MISOCP model adopts CPLEX solver to solve.
In the calculation example, the photovoltaic power supply and Load data of a certain day are selected as the basis of analysis, the normalized data curve is shown in fig. 5, in the graph, PV represents a photovoltaic output curve, and Load represents a Load change curve. For the centralized optimization process, the scheduling period is set to three hours, i.e. the scheduling times in a day is 8. Considering that the too frequent actions of the OLTC and the CB can affect the service life of equipment, setting the maximum action times in one day to be 5 times, and carrying out centralized optimization solution on the improved IEEE33 node model based on the given parameters and the proposed MISOCP centralized optimization model. Fig. 6 and 7 are state change curves of the capacitor bank and OLTC at different times, respectively. By combining the photovoltaic and load change curves of fig. 5, it can be seen that the photovoltaic output and load level are lower in the early morning period, the OLTC state is 0 at this time, the voltage can meet the safety requirement in the state without adjustment, and the CB and photovoltaic reactive power output is smaller. The photovoltaic output and the load level are increased simultaneously in the midday period, the voltage of the end node is lower at the moment, the CB and the photovoltaic reactive power output are increased, meanwhile, the OLTC is switched to a lower gear, the load level is higher but the photovoltaic active output is 0 in the night period, the voltage of the end node of the line is further reduced, more reactive power support is needed, and the CB and the photovoltaic reactive output are further increased. The centralized optimization realizes the long-time scale dispatching of the slow regulating device in the day before, but in the dispatching interval, voltage violations can occur due to the fluctuation and randomness of the photovoltaic power supply and the load, and the reactive power output of the photovoltaic power supply needs to be optimally controlled in a short time scale.
In order to realize real-time distributed control in a day and enable a regional controller based on a data driving model to have good performance, a WGAN-GP model is used for generating a large number of samples which have relevance and accord with real data distribution, the generation results are shown in figures 9 and 10, 300 groups of generated source load data samples are displayed in the figures, and the sampling time interval is 15min, namely 96 data points in a day. Fig. 9 is a box plot of photovoltaic generation data, and it can be seen that the overall distribution of the data sets has a trend of photovoltaic change over the day, while fig. 10 is the load generation data, the main change intervals of which are concentrated in a smaller range. FIG. 11 is a plot of the loss function of WGAN-GP model generator during training, approaching 0.001 when Epoch reaches 5000, illustrating model convergence.
Based on the generated data, carrying out MISOCP model solving for multiple times on the basis of 15min of sampling points to obtain a controllable equipment voltage regulation result, storing the solving result, obtaining input and output samples of training data through a formula (5), and training a data driving model DNN, wherein fig. 12 is a training iteration convergence curve of the area 1 controller. The resulting short time scale distributed control results are shown in fig. 12, which corresponds to 9 in the centralized optimization day: 00-12: the distributed control voltage distribution for the 00 period is shown in fig. 13. In the period, node voltages are all in a safety range, the distribution of the voltages is improved along with the gradual increase of the photovoltaic output, and the frequent action of the PV reactive power is controlled by the regional controller, so that the network loss of the power distribution network is further reduced while the voltage distribution is reasonable, and the centralized optimization network loss is reduced by 47.9% compared with that of the first stage in the period. In the intra-day real-time optimization control of the second stage, the time required for calculation by the centralized optimization is 5.1821s, the time executed by the data-driven area controller is 0.1801s, and the calculation time is reduced by 96.52% compared with the centralized control.
Claims (2)
1. The power distribution network reactive voltage control method based on data driving and source load uncertainty is characterized by comprising the following steps of: comprises the steps of,
Step one, an improved generation countermeasure network WGAN-GP model is established, real sample data x is input, WGAN-GP model parameters are set, and the model is trained to the maximum iteration times;
，
in the method, in the process of the invention, Representing a loss function consisting of generator G and arbiter D together, E [. Cndot. ] representing a mathematical expectation, being a gradient operator,/>Random interpolation for real data and generated data; λ represents the weight of the gradient penalty term; /(I)For 2 norms, D (-) represents a discriminant model, G (-) represents a generator model, the generator and the discriminant are realized through a neural network, the generator and the discriminant model are built by adopting a deep convolution network, z is input noise data, x is sample data conforming to real historical data P data (x), and/>For sample distribution of real data,/>Is a random value subject to a uniform distribution of [0,1 ];
Step two, a centralized reactive voltage optimization model is established, the centralized reactive voltage optimization method carries out optimal power flow calculation in a centralized controller through photovoltaic and load data in each hour to obtain a result of reactive output of an on-load tap-changer OLTC, a capacitor bank CB and a photovoltaic PV, and carries out controllable equipment dispatching in a long time scale to reduce active power loss in a power distribution network;
step three, a distributed reactive voltage control model is established, wherein the controller establishes a network model based on a data-driven deep neural network DNN structure, and the network model is as follows:
，
in the method, in the process of the invention, 、/>And/>、/>The weight and bias matrix of the ith layer-1 and the ith layer respectively, h i,in and h i,out are the input and output of the ith layer network respectively, C is the total layer number in the network,/>For activating a function, when i=c, h C,out is the output result of the current network output layer;
Taking x and y as the input and output of the whole network, beta represents the mapping relation between DNN network input and output, and simplifying the network model into
，
Based on WGAN-GP generated source load data, carrying out MISOCP model solving for a plurality of times to obtain a controllable equipment voltage regulation result, storing the solving result as sample data of a training area controller, and determining input and output characteristics of the model according to a formula (5);
，
in the method, in the process of the invention, Respectively represent j-th PV active force and capacity in Zn area,/>Reactive load for Zn and nodes in the area adjacent to Zn,/>Reactive output power for the ith PV in the Zn region;
And training the deep neural network model offline, putting the trained regional control into online application, and combining with centralized distributed control to form a reactive voltage optimization control strategy with double time scales.
2. The method for controlling reactive voltage of a power distribution network based on uncertainty of data driving and source load according to claim 1, wherein the method is characterized by comprising the following steps: the optimal power flow calculation model is as follows:
，
the formulas (6) and (7) are node active and reactive power balance equations, the formula (8) is a voltage relation between two connected nodes, and the formula (9) represents a relation between branch current, power and node voltage;
in the method, in the process of the invention, 、/>The first active power transmitted by the ij branch at the time t is respectively; /(I)Active power and reactive power transmitted by the lines of the nodes j and k at the moment t are respectively; /(I)Branch current, resistance and reactance; the active power and the reactive power are loaded for the j node at the moment t; /(I) The active power and the reactive power output by the photovoltaic output by the j node at the moment t and the reactive power output by the capacitor bank are respectively; /(I)Is the voltage of the i-node at time t,The voltage of the j node at the time t is represented; /(I)The branch end and head end node sets with the node j as the head end and end nodes respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410356221.6A CN117955121B (en) | 2024-03-27 | Power distribution network reactive voltage control method based on data driving and source load uncertainty |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410356221.6A CN117955121B (en) | 2024-03-27 | Power distribution network reactive voltage control method based on data driving and source load uncertainty |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117955121A true CN117955121A (en) | 2024-04-30 |
| CN117955121B CN117955121B (en) | 2024-06-28 |
Family
ID=
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114977321A (en) * | 2022-06-07 | 2022-08-30 | 国网山东省电力公司莱芜供电公司 | Power distribution network optimal scheduling method considering source load uncertainty |
| CN115632406A (en) * | 2022-12-21 | 2023-01-20 | 国网天津市电力公司城东供电分公司 | Reactive voltage control method and system based on digital-mechanism fusion drive modeling |
| CN116388274A (en) * | 2023-04-04 | 2023-07-04 | 华北电力大学 | New energy and multi-time-scale flexible resource collaborative allocation method in active power distribution network |
| US20230211675A1 (en) * | 2021-12-30 | 2023-07-06 | Sustainable Energy Technologies, Inc. | Supercapacitor to electrochemical hybrid system with smart self-discharge capability |
| CN117291292A (en) * | 2023-08-25 | 2023-12-26 | 天津大学 | Distribution network distribution robust optimization scheduling method based on condition generation countermeasure network |
| US20240054267A1 (en) * | 2019-10-03 | 2024-02-15 | Vestas Wind Systems A/S | Method for planning a layout of a renewable energy site |
| CN117574218A (en) * | 2023-10-20 | 2024-02-20 | 国网山东省电力公司威海供电公司 | Electric power and electric quantity balancing method based on data driving under multidimensional uncertain condition |
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20240054267A1 (en) * | 2019-10-03 | 2024-02-15 | Vestas Wind Systems A/S | Method for planning a layout of a renewable energy site |
| US20230211675A1 (en) * | 2021-12-30 | 2023-07-06 | Sustainable Energy Technologies, Inc. | Supercapacitor to electrochemical hybrid system with smart self-discharge capability |
| CN114977321A (en) * | 2022-06-07 | 2022-08-30 | 国网山东省电力公司莱芜供电公司 | Power distribution network optimal scheduling method considering source load uncertainty |
| CN115632406A (en) * | 2022-12-21 | 2023-01-20 | 国网天津市电力公司城东供电分公司 | Reactive voltage control method and system based on digital-mechanism fusion drive modeling |
| CN116388274A (en) * | 2023-04-04 | 2023-07-04 | 华北电力大学 | New energy and multi-time-scale flexible resource collaborative allocation method in active power distribution network |
| CN117291292A (en) * | 2023-08-25 | 2023-12-26 | 天津大学 | Distribution network distribution robust optimization scheduling method based on condition generation countermeasure network |
| CN117574218A (en) * | 2023-10-20 | 2024-02-20 | 国网山东省电力公司威海供电公司 | Electric power and electric quantity balancing method based on data driving under multidimensional uncertain condition |
Non-Patent Citations (3)
| Title |
|---|
| QINGYUAN JI: "GraphPro: A Graph-based Proactive Prediction Approach for Link Speeds on Signalized Urban Traffic Network", 2022 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN, AND CYBERNETICS (SMC), 18 November 2022 (2022-11-18) * |
| 王文超，庞丹: "考虑电价型需求响应的交直流混合配电网优化调度", 电网技术, vol. 43, no. 5, 31 May 2019 (2019-05-31) * |
| 胡旭光;马大中;郑君;张化光;王睿;: "基于关联信息对抗学习的综合能源系统运行状态分析方法", 自动化学报, no. 09, 15 September 2020 (2020-09-15) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chai et al. | Hierarchical distributed voltage optimization method for HV and MV distribution networks | |
| Erlich et al. | Optimal dispatch of reactive sources in wind farms | |
| CN113363997A (en) | Reactive voltage control method based on multi-time scale and multi-agent deep reinforcement learning | |
| CN110690732A (en) | Photovoltaic reactive power partition pricing power distribution network reactive power optimization method | |
| CN113098022B (en) | Wind farm grid-connected reactive power regulation method, device, equipment and storage medium | |
| CN114597969B (en) | Power distribution network double-layer optimization method considering intelligent soft switch and virtual power plant technology | |
| KR102133897B1 (en) | Method for controlling a voltage and reactive power based on real time measurement, program for the same | |
| CN115313403A (en) | Real-time voltage regulation and control method based on deep reinforcement learning algorithm | |
| Zangeneh et al. | A survey: Fuzzify parameters and membership function in electrical applications | |
| CN112784475A (en) | Multi-agent technology-based multi-level voltage coordination control method for power distribution network | |
| CN114583710B (en) | Wind farm reactive voltage optimization control method based on data driving modeling | |
| Yin et al. | Sequential reconfiguration of unbalanced distribution network with soft open points based on deep reinforcement learning | |
| Hu et al. | Deep reinforcement learning based coordinated voltage control in smart distribution network | |
| CN113872213B (en) | Autonomous optimization control method and device for power distribution network voltage | |
| CN109494750B (en) | Hierarchical distributed voltage optimization control method for high and medium voltage distribution network | |
| CN107947166A (en) | Dispatching method and device when a kind of multipotency microgrid based on dynamic matrix control becomes | |
| CN113422371B (en) | Distributed power supply local voltage control method based on graph convolution neural network | |
| Chen et al. | Real-time volt/var optimization for distribution systems with photovoltaic integration | |
| CN112671047B (en) | Active power distribution network reconstruction and reactive power joint robust optimization method considering limit scene | |
| CN113394770A (en) | Interconnected micro-grid group frequency complete distribution type optimization control method and system | |
| CN117955121B (en) | Power distribution network reactive voltage control method based on data driving and source load uncertainty | |
| Modha et al. | Minimization of active power loss for optimum reactive power dispatch using pso | |
| CN117955121A (en) | Power distribution network reactive voltage control method based on data driving and source load uncertainty | |
| CN116667361A (en) | Multi-stage voltage automatic control optimization method containing new energy based on model predictive control | |
| CN116979611A (en) | Hierarchical optimization scheduling method for source network load storage |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |