A method for calculating the voltage of missing nodes in a distribution network measurement

Abstract

This invention relates to the field of power grid data repair technology, specifically a method for calculating the voltage of nodes with missing distribution network measurements. By analyzing the power deviation problem in distribution network measurement configuration, a mechanism model for power deviation allocation is studied and embedded into a neural network to construct a hybrid driving mode. The neural network is trained offline using historical measurement data to achieve online inference of voltage phasors for nodes with missing distribution network measurements. The calculation and derivation problem is transformed into an allocation coefficient optimization problem through the current matching mechanism equation, while satisfying power flow equation constraints. Finally, the voltage phasor of any node is quickly solved using a mechanism-data hybrid driving neural network. This invention effectively solves the voltage calculation problem under conditions of missing distribution network measurements, improving computational efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of power grid data repair technology. Background Technology

[0002] As the end point of the power system, the distribution network plays a crucial role in supplying and distributing electricity to users and connecting distributed power sources. However, the current measurement coverage of intermediate nodes in the distribution network is low, making it difficult to achieve full coverage and resulting in opacity in the operation of the distribution network. At the same time, with the distributed access of new energy sources, voltage overruns at key nodes in the distribution network occur in some areas, and even photovoltaic backfeeding to the main grid occurs, affecting the safe and stable operation of the power grid. In addition, the distribution network has numerous lines and equipment, making it difficult and costly to obtain accurate models. Furthermore, when model parameters are biased, the optimal solution obtained by the optimization method is no longer reliable. Traditional model-based optimization methods are difficult to meet actual needs, and data-driven methods must be introduced to compensate for model errors. Researching computational derivation methods for missing measurement nodes in distribution networks is of significant importance. Existing technologies have proposed the concept of boundary power for distribution network measurements, transforming state estimation into a weight allocation problem; however, they lack optimization methods that incorporate power flow constraints. Some studies have explored lightweight synchronous phasor D-PMU configuration architectures for distribution networks, effectively addressing the redundancy issue associated with missing measurements, but these methods are not yet suitable for nationwide application across the entire power grid. Other approaches have designed corresponding state estimation models for different distribution network measurement conditions, constructing nonlinear dynamic systems to accelerate state estimation convergence; however, these methods are highly sensitive to initial conditions. Therefore, there is an urgent need to propose a practical and engineering-appropriate method for deriving the voltage of missing measurement nodes in distribution networks. Summary of the Invention

[0003] To overcome the problems that existing models cannot meet new requirements and that the calculation method for missing nodes in distribution network measurements is not applicable to all regions, this invention provides a method for calculating the voltage of missing nodes in distribution network measurements.

[0004] The technical solution adopted by this invention to achieve the above objectives is: a method for calculating the voltage of missing nodes in a distribution network, comprising the following steps:

[0005] S1. Collect real-time measurement data of nodes in the distribution network system, including voltage, current and power;

[0006] S2. Based on the collected data, use the current mechanism equation to calculate the relationship between the load of the node, the inflow and outflow power of the node, and the deviation power under any end node;

[0007] S3. Initialize the voltage at the first node and assign initial values ​​to the voltages at other nodes. Calculate the initial values ​​of the initial deviation power and the initial value of the deviation power distribution coefficient.

[0008] S4. Establish a mechanism-data hybrid driving neural network, inputting the relationship between the load of each node and the inflow and outflow power of that node, as well as the initial value of the deviation power allocation coefficient, into the neural network;

[0009] S5. Given load data, calculate branch power and current through a forward calculation, calculate the voltage of each node by back-calculating from the first node, and determine whether the voltage difference between two adjacent iterations is less than the given convergence index.

[0010] S6. After satisfying the convergence index, substitute the optimal solution of the calculated deviation power distribution coefficient into the existing power flow calculation equation to solve for the voltage phasor of any node.

[0011] Preferably, step S2 specifically involves: calculating any end Under the node, the load of the node , With the power flowing into and out of this node , The relationship between the power and the deviation power:

[0012] ;

[0013] in, This represents the active load of the node at the 15-minute level. This represents the reactive load at the 15-minute level for this node. This represents the magnitude of the voltage at that node. Let be the phase angle of the voltage at that node. This is the active power deviation distribution coefficient at this node. This is the reactive power deviation distribution coefficient at this node. The active power deviation is the power. This refers to reactive power deviation.

[0014] Calculate the measurement relationships of secondary distribution coverage equipment:

[0015] ;

[0016] in, This represents the active power load of the node on a second-by-second scale. This represents the reactive load on a second-level scale for this node;

[0017] Calculate the phase angle of the voltage at the first node:

[0018] ;

[0019] in, The phase angle of the first node voltage. For the benefit of network loss, For network loss and no effect, The current at the first-end node is in the order of seconds. The voltage at the first node is in the range of seconds.

[0020] Preferably, step S3 includes:

[0021] Initialize the voltage at the first node, assign initial values ​​to the voltages at other nodes, and calculate the initial deviation power:

[0022] ;

[0023] in, This is the initial active power deviation power allocation coefficient at this node. This is the initial reactive power deviation distribution coefficient at this node. It is the active power of the end load. It is the reactive power of the end load;

[0024] Calculate the iterative value of the deviation power and the iterative update of the 15-minute measured load node:

[0025] ;

[0026] Obtain the relationship between the load of each node and the inflow and outflow power of that node, as well as the deviation power allocation coefficient. , The initial value.

[0027] Preferably, step S4 includes: establishing a mechanism-data hybrid driven neural network, inputting the relationship between the load of each node and the inflow and outflow power of that node, as well as the initial value of the deviation power allocation coefficient into the neural network, the neural network input being the measurement information of the line and the beginning and end ends, during the training process, using the deviation power and the deviation of the known measurement value as the loss function, and incorporating the calculation process of forward and backward power flow, so that the data-driven solution of the allocation coefficient automatically satisfies the power flow constraints and mechanism process, and different neural network weight coefficients are trained according to different feeders, so as to obtain the optimal solution of the deviation power allocation coefficient when the measurement inputs at the beginning and end ends are obtained.

[0028] Preferably, step S5 includes: given load data, calculating branch power and current through a forward calculation, calculating the voltage of each node by back-calculating from the first node, and determining whether the voltage difference between two adjacent iterations is less than a given convergence criterion.

[0029] ;

[0030] in, Given a convergence index;

[0031] If the convergence criterion is met, stop; otherwise, repeat steps S3-S5 until the convergence criterion is met.

[0032] The beneficial effects of this invention are as follows:

[0033] This invention constructs a distribution network calculation and deduction system from a mechanism-data fusion-driven perspective, solving the problem of insufficient secondary coverage in power distribution systems. Addressing the issue of asynchronous measurements at the beginning and end of existing distribution networks leading to impaired power flow calculations, it presents a calculation process combining deviation power allocation and power flow forward and backward iteration. The mechanistic equations are constructed as a loss function for a neural network, with the mechanism incorporated into the loss function, ensuring that each iteration of the neural network approximates the mechanistic equations. Training with existing measurement data improves the speed of weight coefficient calculation, achieving offline training and online inference capabilities. This effectively solves the voltage calculation problem under conditions of missing distribution network measurements, improving computational efficiency and accuracy. Attached Figure Description

[0034] Figure 1 This is an overall flowchart of an embodiment of the present invention;

[0035] Figure 2 This is a schematic diagram of the neural network structure according to an embodiment of the present invention. Detailed Implementation

[0036] Embodiments of the present invention provide a method for calculating the voltage of missing nodes in a distribution network, such as... Figure 1 As shown, it includes the following steps:

[0037] S1. Collect real-time measurement data from the end nodes of the distribution network system, such as voltage, current and power, including real-time current measurement data of pole-mounted switches and branch lines at the second level and near real-time measurement data of the end distribution substation transformer assessment table at the 15-minute level.

[0038] S2. Based on the above data, use the current mechanism equation to calculate the value at any end. Under the node, the load of the node , With the power flowing into and out of this node , The relationship between the power and the deviation power:

[0039] ;

[0040] in, This represents the active load of the node at the 15-minute level. This represents the reactive load at the 15-minute level for this node. This represents the magnitude of the voltage at that node. Let be the phase angle of the voltage at that node. This is the active power deviation distribution coefficient at this node. This is the reactive power deviation distribution coefficient at this node. The active power deviation is the power. This refers to reactive power deviation.

[0041] Calculate the measurement relationships of secondary distribution coverage equipment:

[0042] ;

[0043] in, This represents the active power load of the node on a second-by-second scale. This represents the reactive load on a second-level scale for this node;

[0044] Calculate the phase angle of the voltage at the first node:

[0045] ;

[0046] in, The phase angle of the first node voltage. For the benefit of network loss, For network loss and no effect, The current at the first-end node is in the order of seconds. The voltage at the first node is in the range of seconds.

[0047] S3. Initialize the voltage at the first node and assign initial values ​​to the voltages at other nodes, and calculate the initial deviation power:

[0048] ;

[0049] in, This is the initial active power deviation power allocation coefficient at this node. This is the initial reactive power deviation distribution coefficient at this node. It is the active power of the end load. It is the reactive power of the end load. and It reflects the actual electrical energy consumed at the end of the distribution network;

[0050] Calculate the iterative value of the deviation power and the iterative update of the 15-minute measured load node:

[0051] ;

[0052] Obtain the relationship between the load of each node and the inflow and outflow power of that node, as well as the deviation power allocation coefficient. , The initial value;

[0053] S4. Establish a mechanism-data hybrid driven neural network to establish the relationship between the load of each node and the inflow and outflow power of that node, as well as the deviation power allocation coefficient. , The initial values ​​are input into the neural network, and the neural network structure is as follows: Figure 2As shown, the input consists of measurement information from the line and its two ends. During training, the deviation power and the deviation from the known measurement values ​​are used as the loss function, and the calculation process of forward and backward power flow is incorporated. This allows the data-driven solution of the allocation coefficients to automatically satisfy power flow constraints and mechanisms. Different neural network weight coefficients are trained for different feeders to obtain the allocation coefficients of the deviation power when the measurement inputs from the two ends are received. , The optimal solution;

[0054] S5, Given load data By performing a forward pass operation, the branch power and current are calculated. Then, by working backward from the first node, the voltage of each node is calculated. Finally, it is determined whether the voltage difference between two adjacent iterations is less than a given convergence criterion.

[0055] ;

[0056] in, Given a convergence index;

[0057] If the convergence criterion is met, stop; otherwise, repeat steps S3-S5 until the convergence criterion is met.

[0058] S6. After satisfying the convergence index, the distribution coefficient of the calculated deviation power will be... , Substituting the optimal solution into the existing power flow calculation equations:

[0059]

[0060] in, and They are nodes Active power and reactive power, For nodes The voltage phasor, and They are nodes With nodes Mutual conductance and mutual susceptance between them For nodes With nodes The voltage phase angle difference;

[0061] Solving for the voltage phasor at any node .

[0062] This invention has been described through embodiments. Those skilled in the art will understand that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of this invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, this invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of this invention.

Claims

1. A method for calculating the voltage of missing nodes in a distribution network measurement, characterized in that, Includes the following steps: S1. Collect real-time measurement data of nodes in the distribution network system, including voltage, current and power; S2. Based on the collected data, use the current mechanism equation to calculate the relationship between the load of the node, the inflow and outflow power of the node, and the deviation power at any end node; Calculate arbitrary end Under the node, the load of the node , With the power flowing into and out of this node , The relationship between the power and the deviation power: ； in, This represents the active load of the node at the 15-minute level. This represents the reactive load at the 15-minute level for this node. This represents the magnitude of the voltage at that node. Let be the phase angle of the voltage at that node. This is the active power deviation distribution coefficient at this node. This is the reactive power deviation distribution factor at this node. The active power deviation is the power. This refers to reactive power deviation. Calculate the measurement relationships of secondary distribution coverage equipment: ； in, This represents the active power load of the node on a second-by-second scale. This represents the reactive load on a second-level scale for this node; Calculate the phase angle of the voltage at the first node: ； in, The phase angle of the voltage at the first node. Contributing to the network loss For network loss and no effect, The current at the first-end node is in the order of seconds. The voltage at the first node is in the range of seconds. S3. Initialize the voltage at the first node and assign initial values ​​to the voltages at other nodes. Calculate the initial values ​​of the initial deviation power and the initial value of the deviation power distribution coefficient. Initialize the voltage at the first node and assign initial values ​​to the voltages at other nodes, then calculate the initial deviation power: ； in, This is the initial active power deviation power allocation coefficient at this node. This is the initial reactive power deviation distribution coefficient at this node. It is the active power of the end load. It is the reactive power of the end load; Calculate the iterative value of the deviation power and the iterative update of the 15-minute measured load node: ； Obtain the relationship between the load of each node and the inflow and outflow power of that node, as well as the deviation power allocation coefficient. , The initial value; S4. Establish a mechanism-data hybrid driving neural network, inputting the relationship between the load of each node and the inflow and outflow power of that node, as well as the initial value of the deviation power allocation coefficient, into the neural network; S5. Given load data, calculate branch power and current through a forward calculation, calculate the voltage of each node by back-calculating from the first node, and determine whether the voltage difference between two adjacent iterations is less than the given convergence index. S6. After satisfying the convergence index, substitute the optimal solution of the calculated deviation power distribution coefficient into the existing power flow calculation equation to solve for the voltage phasor of any node.

2. The method for calculating the voltage of missing nodes in a distribution network according to claim 1, characterized in that, Step S4 includes: establishing a mechanism-data hybrid driven neural network, inputting the relationship between the load of each node and the inflow and outflow power of that node, as well as the initial value of the deviation power allocation coefficient into the neural network. The neural network input is the measurement information of the line and the beginning and end ends. During the training process, the deviation between the deviation power and the known measurement value is used as the loss function, and the calculation process of forward and backward power flow is incorporated, so that the data-driven solution of the allocation coefficient automatically satisfies the power flow constraints and mechanism process. Different neural network weight coefficients are trained according to different feeders to obtain the optimal solution of the deviation power allocation coefficient when the measurement inputs at the beginning and end ends are obtained.

3. The method for calculating the voltage of missing nodes in a distribution network according to claim 1, characterized in that, Step S5 includes: given load data, calculating branch power and current through a forward calculation, calculating the voltage of each node by back-calculating from the first node, and determining whether the voltage difference between two adjacent iterations is less than a given convergence criterion. ； in, Given a convergence index; If the convergence criterion is met, stop; otherwise, repeat steps S3-S5 until the convergence criterion is met.

Images