A real-time angle difference monitoring and intelligent decision method and system for rapid loop closing and load reversal of a power distribution network

Abstract

This invention discloses a real-time phase angle difference monitoring and intelligent decision-making method and system for rapid loop-closing load switching in distribution networks, relating to the field of distribution network technology. The method includes: constructing a phase angle measurement matrix based on the three-phase voltage time series corresponding to the distribution network; obtaining an observation admittance matrix based on the phase angle measurement matrix; calculating a dry observation admittance matrix when the distribution network is under a dry reference condition, and obtaining a condensation topology latching matrix based on the dry observation admittance matrix and the observation admittance matrix; constructing a strategy evaluation matrix based on the condensation topology latching matrix; and obtaining a compensation observation admittance matrix based on the target state row in the strategy evaluation matrix. This invention improves the safety of the rapid loop-closing load switching process in distribution networks.

Description

Technical Field

[0001] This invention relates to the field of power distribution network technology, and in particular to a real-time angle difference monitoring and intelligent decision-making method and system for rapid loop-closing load switching in power distribution networks. Background Technology

[0002] With the continuous expansion of power distribution networks and the increasing demands for power supply reliability, fast loop-closing load transfer technology has become a common operating mode in urban power distribution networks and industrial park power distribution networks. Fast loop-closing load transfer refers to the operation method of establishing an electrical connection between two lines supplied by different power sources through a short-term parallel connection, completing the load transfer without power interruption, and then restoring the normal operating structure by disconnecting the lines. This operating mode is widely used in maintenance transfer, power switching, load optimization and distribution, and emergency handling of sudden faults, especially in outdoor overhead lines, pole-mounted switches, prefabricated substations, and multi-branch feeder structures. In actual operation, the key to loop-closing operation is to accurately grasp the voltage phase angle difference on both sides of the loop-closing point to control the magnitude of the circulating current that may be generated at the moment of loop closure, avoiding inrush current, protection malfunction, or increased equipment stress due to excessive phase difference. In the operating environment of outdoor power distribution equipment, when the air humidity increases and reaches a certain level, condensation can easily form on the surfaces of insulators, poles, supports, and voltage sampling points. Under the influence of an electric field, the water stains may connect along a specific path, forming additional conductive channels, thereby changing the original electrical equivalent admittance distribution structure and affecting the accuracy of voltage phase angle measurement.

[0003] In existing technologies, the acquisition of loop phase angle difference typically relies on offline power flow calculations or estimations based on single-point voltage measurements combined with synchronization devices. While this meets basic requirements under dry environmental conditions and ideal equipment status, condensation or moisture on outdoor equipment surfaces alters the electrical proportional relationships between nodes, causing a structural shift in the observed admittance matrix and resulting in systematic errors in voltage phase angle calculations. Because existing methods lack mechanisms for identifying and correcting admittance structural shifts, it is difficult to distinguish between phase differences caused by changes in the actual power grid operating conditions and measurement deviations caused by additional conductive paths in the environment. This leads to risks of inaccurate phase angle judgment, distorted circulating current assessment, and even misjudgments during rapid loop closure and load shedding decisions, impacting the safety and reliability of the distribution network operation. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies that make it difficult to distinguish between phase differences caused by changes in the actual operating state of the power grid and measurement deviations caused by additional conductive channels in the environment. Therefore, this invention proposes a real-time phase difference monitoring and intelligent decision-making method and system for rapid loop-closing and load switching in distribution networks.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: A real-time angle difference monitoring and intelligent decision-making method for rapid loop closing and load switching in distribution networks includes: S1. Based on the three-phase voltage time series corresponding to the distribution network, a phase angle measurement matrix is ​​constructed; S2. Based on the phase angle measurement matrix, obtain the observation admittance matrix; S3. When the distribution network is under dry reference conditions, calculate the dry observation admittance matrix, and based on the dry observation admittance matrix and the observation admittance matrix, obtain the condensation topology latch matrix. S4. Construct a strategy evaluation matrix based on the condensation topology latch matrix, and obtain the compensation observation admittance matrix based on the target state row in the strategy evaluation matrix. S5. Calculate the reward value of reinforcement learning based on the compensation observation admittance matrix, and update the policy evaluation matrix based on the reward value to obtain the target angle difference compensation matrix. S6. Based on the target angle difference compensation matrix, perform fast loop-closing load reversal control on the loop-closing interconnection switch.

[0006] Preferably, a phase angle measurement matrix is ​​constructed based on the three-phase voltage time series corresponding to the distribution network, including: Nodes in the distribution network are selected to obtain multiple measurement nodes; Three-phase voltages are sampled at the measurement nodes to obtain a three-phase voltage time series; The three-phase voltage time series is subjected to discrete Fourier transform to obtain the phase angles of the first phase voltage, the second phase voltage, and the third phase voltage. A phase angle measurement matrix is ​​constructed based on the phase angles of the first, second, and third phase voltages.

[0007] Preferably, the observation admittance matrix is ​​obtained based on the phase angle measurement matrix, including: Three-phase current is sampled at the measurement nodes to obtain a three-phase current time series; The three-phase current time series is processed by discrete Fourier transform to obtain the first phase current phasor, the second phase current phasor, and the third phase current phasor; Construct a current matrix based on the first phase current phasor, the second phase current phasor, and the third phase current phasor; Performing a generalized inverse operation on the phase angle measurement matrix yields the generalized inverse matrix of the phase angle measurement matrix; The observation admittance matrix is ​​obtained by performing matrix multiplication on the generalized inverse matrix of the phase angle measurement matrix and the current matrix.

[0008] Preferably, when the distribution network is under dry reference operating conditions, the calculation of the dry observation admittance matrix includes: Determine that the power distribution network is in a dry reference operating condition with no condensation and a stable environment; Obtain the drying phase angle measurement matrix and drying current matrix under the drying reference operating conditions; The drying observation admittance matrix is ​​calculated based on the drying current matrix and the drying phase angle measurement matrix.

[0009] Preferably, based on the dryness observation admittance matrix and the observation admittance matrix, the condensation topology latch matrix is ​​obtained, including: The average admittance matrix is ​​obtained based on the observed admittance matrix; Based on the dry observation admittance matrix, the reference admittance matrix is ​​obtained; The condensation topology latch matrix is ​​obtained by subtracting the average admittance matrix from the reference admittance matrix.

[0010] Preferably, constructing a strategy evaluation matrix includes: Based on historical operating data of the distribution network, determine the set of angle difference compensation coefficients; Based on the set of angle difference compensation coefficients, a candidate angle difference compensation matrix is ​​constructed; The element distribution of the condensation topology latch matrix is ​​classified to obtain the condensation topology type; A strategy evaluation matrix is ​​constructed based on the candidate angle difference compensation matrix and different condensation topology types.

[0011] Preferably, the compensation observation admittance matrix is ​​obtained based on the target state row in the strategy evaluation matrix, including: Based on the condensation topology latch matrix, the target state row in the strategy evaluation matrix is ​​determined; Based on the target state row, the column index of the strategy evaluation matrix is ​​selected to obtain the candidate angle difference compensation matrix corresponding to the target state row; The candidate angle difference compensation matrix and phase angle measurement matrix corresponding to the target state row are added together to obtain the compensation phase angle measurement matrix; Calculate the generalized inverse matrix of the compensation phase angle measurement matrix; The compensation observation admittance matrix is ​​obtained by performing matrix multiplication on the generalized inverse matrix of the current matrix and the compensation phase angle measurement matrix.

[0012] Preferably, obtaining the target angle difference compensation matrix includes: The admittance residual matrix is ​​obtained by subtracting the compensation observation admittance matrix from the reference admittance matrix. The norm of the admittance residual matrix is ​​calculated to obtain the residual index; The reward value for reinforcement learning is calculated based on the residual index. Based on the return value, the strategy evaluation matrix is ​​updated to obtain the updated strategy evaluation matrix; Perform a column maximum search on the updated strategy evaluation matrix to obtain the target column index; The candidate angle difference compensation matrix corresponding to the target column index is used as the target angle difference compensation matrix.

[0013] Preferably, based on the target angle difference compensation matrix, rapid loop-closing load switching is performed on the loop-closing interconnector switch, including: The phase angle measurement matrix and the target angle difference compensation matrix corresponding to the current sampling time are added together to obtain the online compensation phase angle measurement matrix; Extract the phase angle data corresponding to the measurement nodes in the closed loop from the online compensated phase angle measurement matrix; Calculate the real-time angle difference based on phase angle data; Based on real-time angle difference, a fast closed-loop load reversal control strategy is generated. The fast loop-closing load reversal control strategy performs fast loop-closing load reversal control on the loop-closing interconnection switch.

[0014] To address the aforementioned problems, this invention also provides a real-time angle difference monitoring and intelligent decision-making system for rapid loop-closing load switching in distribution networks, the system comprising: The phasor acquisition module is used to construct a phase angle measurement matrix based on the three-phase voltage time series corresponding to the distribution network. The admittance modeling module is used to obtain the observation admittance matrix based on the phase angle measurement matrix; The condensation modeling module is used to calculate the dry observation admittance matrix when the distribution network is under dry reference conditions, and to obtain the condensation topology latch matrix based on the dry observation admittance matrix and the observation admittance matrix. The enhanced decision-making module is used to construct a strategy evaluation matrix based on the condensation topology latch matrix, and to obtain the compensation observation admittance matrix based on the target state row in the strategy evaluation matrix. The compensation generation module is used to calculate the reward value of reinforcement learning based on the compensation observation admittance matrix, and update the policy evaluation matrix based on the reward value to obtain the target angle difference compensation matrix. The parallel control module is used to perform rapid loop-closing load reversal control on the loop-closing interconnection switch based on the target angle difference compensation matrix.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention constructs a dry observation admittance matrix as a stable benchmark and compares and analyzes it with the real-time observation admittance matrix to form a condensation topology latch matrix. This identifies the distribution of additional leakage admittance caused by environmental condensation and achieves a quantitative characterization of admittance structure offset. Compared with the method of relying solely on single-point phase angle measurement or offline power flow calculation, it can accurately distinguish between the deviation caused by actual operating changes and environmental additional conductive channels in complex outdoor environments, and improve the accuracy of closed loop angle difference judgment.

[0016] 2. A strategy evaluation matrix is ​​constructed based on the condensation topology type, and a reinforcement update mechanism is established by using the residual index between the compensation observation admittance matrix and the reference admittance matrix. This enables the angle difference compensation process to have adaptive capabilities. By evaluating and iteratively updating different candidate compensation matrices online, the optimal compensation scheme can be gradually selected, reducing admittance residuals and improving the consistency between the compensated phase angle and the actual power grid state, thereby enhancing the reliability of decision-making.

[0017] 3. Based on the corrected online phase angle data, calculate the real-time angle difference on both sides of the loop and generate a fast loop-closing load switching control strategy. By combining physical modeling with intelligent decision-making, the loop-closing operation is based on the corrected real electrical relationship, which effectively reduces the risk of misjudgment caused by environmental influences and improves the safety and stability of the fast loop-closing load switching process of the distribution network. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network, provided in an embodiment of the present invention. Figure 2 This is a functional block diagram of a real-time angle difference monitoring and intelligent decision-making system for rapid loop-closing load switching in a distribution network, provided as an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0020] This embodiment provides a real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network. (See also...) Figure 1 Specifically, including: S1. Based on the three-phase voltage time series corresponding to the distribution network, a phase angle measurement matrix is ​​constructed; In an embodiment of the present invention, a phase angle measurement matrix is ​​constructed based on the three-phase voltage time series corresponding to the distribution network, including: Nodes in the distribution network are selected to obtain multiple measurement nodes; A power distribution network refers to a power distribution network structure consisting of substations, feeders, switching equipment, and load nodes, used to transmit electrical energy from the power source side to each power consumption node; node selection refers to selecting several representative electrical connection points in the power distribution network as measurement locations; the measurement node refers to an electrical connection point that is actually connected to a voltage sampling device and can acquire voltage signals.

[0021] Specifically, the primary wiring data and operational topology information of the distribution network are obtained. This operational topology information includes the connection relationships between each substation outgoing lines and each feeder, as well as the current open / closed status of each sectionalizing switch and tie switch. Based on this operational topology information, the feeder range relevant to the proposed loop-closing load transfer operation is determined, and the electrical connection points on both sides of the loop-closing tie switch within this feeder range are identified as mandatory nodes. Tracing back from the mandatory nodes towards the power supply side, the corresponding power supply point is selected as the upstream node, corresponding to the power supply side bus or outgoing terminal. Tracing back from the mandatory nodes towards the load side, the points are identified as sectionalizing switches, important load branches, or terminal nodes, and locations sensitive to load distribution changes are selected as downstream nodes. On-site verification is performed on each candidate node to confirm that it has a live connection location suitable for installing a voltage sampling device, and that it meets the requirements for fixed installation and safe distance. A unique node number is assigned to each measurement node, and the correspondence between the node number and the on-site equipment number, the name of the line to which it belongs, and the location of the tower or transformer substation is recorded to form a measurement node list.

[0022] Three-phase voltages are sampled at the measurement nodes to obtain a three-phase voltage time series; Three-phase voltage sampling refers to the process of continuously acquiring the three-phase AC voltage signal at the measurement node according to a set sampling frequency; three-phase voltage time series refers to the set of instantaneous three-phase voltage data arranged in chronological order during the continuous sampling process.

[0023] Specifically, voltage sampling circuits are set up on the three-phase conductors of each measurement node. These circuits convert the primary-side three-phase voltage into an acquireable low-voltage signal using a voltage transformer or capacitive voltage divider, and then introduce these three low-voltage signals into sampling channels. The sampling channels are checked for phase identification, confirming that the conductor phase corresponding to each channel matches the node number. The range and polarity of the sampling channels are also checked to ensure that the sampled values ​​are consistent with the direction of change of the primary-side voltage. The sampling channels of each measurement node are connected to a synchronous sampling device, which outputs a unified sampling time marker, aligning the sampling at each measurement node to the same sampling time. During sampling, the instantaneous values ​​of the three-phase voltage are continuously collected at the unified time and stored in chronological order to form a three-phase voltage time series.

[0024] The three-phase voltage time series is subjected to discrete Fourier transform to obtain the phase angles of the first phase voltage, the second phase voltage, and the third phase voltage. Discrete Fourier Transform refers to a mathematical processing method that performs frequency domain decomposition on the three-phase voltage time series to extract the amplitude and phase information of the fundamental component; the phase angles of the first phase voltage, the second phase voltage, and the third phase voltage refer to the phase angle values ​​of the corresponding three-phase voltage fundamental components relative to a unified reference phase.

[0025] Specifically, for each measurement node, the instantaneous three-phase voltage values ​​aligned at the same sampling time in the three-phase voltage time series are read, and the instantaneous voltage values ​​of each phase are arranged into a discrete sequence in chronological order. The fundamental frequency corresponding to the discrete sequence is determined. The fundamental frequency is determined by the operating frequency of the distribution network and converted from the time interval corresponding to the sampling time mark. Using one fundamental cycle as the calculation window, the discrete sequence is segmented. Each segment contains multiple consecutive sampling points and covers a complete fundamental cycle. DC component removal processing is performed on each segment of the discrete sequence. The DC component removal processing involves calculating the mean of the segment and subtracting the mean from each sampling point in the segment. A discrete Fourier transform is performed on the DC component-removed discrete sequence to calculate the complex spectrum value of the segment at the fundamental frequency. The complex spectrum value consists of a real part and an imaginary part. The real part is obtained by multiplying each sampling point with the fundamental cosine sequence and summing the results. The imaginary part is obtained by multiplying each sampling point with the fundamental sine sequence and summing the results. The phase angle is calculated based on the real and imaginary parts of the complex spectrum value. The phase angle is obtained by arctangent operation and normalized according to a unified phase reference to ensure that the phase angles of different measurement nodes are comparable. Thus, the first phase voltage phase angle, the second phase voltage phase angle and the third phase voltage phase angle are obtained respectively and associated with the corresponding measurement node number and sampling time mark for storage.

[0026] A phase angle measurement matrix is ​​constructed based on the phase angles of the first, second, and third phase voltages.

[0027] A phase angle measurement matrix is ​​a matrix-like data structure formed by arranging the phase angles of the three-phase voltages corresponding to each measurement node in the order of the measurement node numbers. It is used to reflect the phase distribution state of multiple measurement nodes at the same time.

[0028] Specifically, for each sampling time, the three-phase voltage phase angle data of all measurement nodes at that sampling time are collected, and the measurement nodes are numbered and sorted according to the aforementioned measurement node list to form a fixed row order. The first phase voltage phase angle of the i-th measurement node at the sampling time is written into the i-th row and first column of the phase angle measurement matrix, the second phase voltage phase angle of the i-th measurement node at the sampling time is written into the i-th row and second column of the phase angle measurement matrix, and the third phase voltage phase angle of the i-th measurement node at the sampling time is written into the i-th row and third column of the phase angle measurement matrix, thereby obtaining a phase angle measurement matrix composed of multiple rows corresponding to the number of measurement nodes and three columns corresponding to the number of nodes.

[0029] S2. Based on the phase angle measurement matrix, obtain the observation admittance matrix; In an embodiment of the present invention, the observation admittance matrix is ​​obtained based on the phase angle measurement matrix, including: Three-phase current is sampled at the measurement nodes to obtain a three-phase current time series; Three-phase current sampling refers to the process of continuously collecting the instantaneous current values ​​flowing through the three-phase conductors at the measurement node according to a unified sampling time; three-phase current time series refers to the set of three-phase current instantaneous value data arranged in chronological order.

[0030] Specifically, for each measurement node, a sampling position for the three-phase current is determined. The sampling position is the location of the current path of the three-phase conductors at that measurement node. A current sampling loop is set up at the sampling position. The current sampling loop uses a current transformer or Rogowski coil to acquire the primary side three-phase current and outputs a secondary side signal proportional to the primary current. The three-phase secondary side signals are then connected to the sampling channels respectively. Phase correspondence verification is performed on the sampling channels to ensure a one-to-one correspondence between the sampling channels and the three-phase conductors. The sampling channels of each measurement node are connected to a sampling device at a unified sampling time, enabling each measurement node to synchronously sample the three-phase current at the same sampling time. The instantaneous values ​​of the three-phase current at each sampling time are continuously acquired and stored in chronological order as a three-phase current time series.

[0031] The three-phase current time series is processed by discrete Fourier transform to obtain the first phase current phasor, the second phase current phasor, and the third phase current phasor; The first phase current phasor, the second phase current phasor, and the third phase current phasor refer to the complex representations formed by the fundamental components of the corresponding three-phase currents. The complex representations are composed of amplitude and phase, and are used to characterize the magnitude and relative phase of the phase current.

[0032] Specifically, for each measurement node, the instantaneous three-phase current values ​​aligned at the same sampling time in the three-phase current time series are read, and each instantaneous phase current value sequence is arranged into a discrete sequence in chronological order. The fundamental frequency corresponding to the discrete sequence is determined, which is determined by the operating frequency of the distribution network and converted from the time interval corresponding to the sampling time mark. The discrete sequence is segmented using one fundamental cycle as the calculation window. Each segment contains multiple consecutive sampling points and covers a complete fundamental cycle. DC component removal processing is performed on each segment of the discrete sequence. The DC component removal processing involves calculating the mean of the discrete sequence segment and subtracting the mean from each sampling point in the segment. A discrete Fourier transform is performed on the discrete sequence after removing the DC component. The complex spectrum value is calculated at the fundamental frequency. The real part of the complex spectrum value is obtained by multiplying the discrete sequence with the fundamental cosine sequence point by point and summing the results. The imaginary part of the complex spectrum value is obtained by multiplying the discrete sequence with the fundamental sine sequence point by point and summing the results. The complex spectrum value is output as the current phasor of the corresponding phase, thereby obtaining the first phase current phasor, the second phase current phasor, and the third phase current phasor, respectively.

[0033] Construct a current matrix based on the first phase current phasor, the second phase current phasor, and the third phase current phasor; A current matrix is ​​a matrix-like data structure formed by arranging the three-phase current phasors corresponding to each measurement node in the order of the measurement node number. It is used to centrally express the current distribution state of multiple measurement nodes at the same time.

[0034] Specifically, the three-phase current phasor data of each measurement node at the same sampling time are collected. The three-phase current phasor data correspond to the first phase current phasor, the second phase current phasor, and the third phase current phasor, respectively, and are one-to-one with the measurement node number. The measurement node numbers are sorted according to the measurement node list to determine the row order of the current matrix. The three-phase current phasors of each measurement node are written in the row order. The first phase current phasor of each measurement node is written into the first column of the current matrix, the second phase current phasor of each measurement node is written into the second column of the current matrix, and the third phase current phasor of each measurement node is written into the third column of the current matrix, thus forming a three-column matrix arranged in the order of measurement node numbers.

[0035] Performing a generalized inverse operation on the phase angle measurement matrix yields the generalized inverse matrix of the phase angle measurement matrix; The generalized inverse matrix refers to the inverse mapping matrix obtained by performing linear algebraic operations on the phase angle measurement matrix. It is used to establish a linear relationship between voltage phase angle and current in the sense of minimum error.

[0036] Specifically, the phase angle measurement matrix at the same sampling time as the current matrix is ​​obtained. Its row order is verified to match the measurement node numbering order, and its column order is verified to match the three-phase sequence. The phase angle measurement matrix is ​​transposed to obtain the transposed matrix. Matrix multiplication is performed between the transposed matrix and the phase angle measurement matrix to obtain the product matrix. The invertibility of the product matrix is ​​determined. If the product matrix is ​​invertible, its inverse is obtained. Then, matrix multiplication is performed between the inverse matrix and the transposed matrix to obtain the generalized inverse matrix of the phase angle measurement matrix. If the product matrix is ​​not invertible, it is regularized to obtain an invertible matrix. Then, its inverse is obtained. Finally, matrix multiplication is performed between the inverse matrix and the transposed matrix to obtain the generalized inverse matrix of the phase angle measurement matrix.

[0037] The observation admittance matrix is ​​obtained by performing matrix multiplication on the generalized inverse matrix of the phase angle measurement matrix and the current matrix.

[0038] The observation admittance matrix refers to the nodal admittance representation obtained by multiplying the generalized inverse matrix of the phase angle measurement matrix with the current matrix. It is used to describe the equivalent electrical connection relationship between measurement nodes.

[0039] Specifically, the generalized inverse matrix of the phase angle measurement matrix and the current matrix corresponding to the same sampling time are obtained. The sequence of measurement node numbers in both matrices is verified to be consistent. The three-phase column order of the current matrix is ​​verified to be consistent with the three-phase column order of the phase angle measurement matrix. The column order of the generalized inverse matrix of the phase angle measurement matrix is ​​verified to be consistent with the row order of the current matrix. Matrix multiplication is performed on the generalized inverse matrix of the phase angle measurement matrix and the current matrix. This matrix multiplication includes summing the element-wise products of each row of the generalized inverse matrix of the phase angle measurement matrix and each column of the current matrix according to the multiplication rules. The product elements corresponding to each row and column are then written into the admittance matrix data structure according to their row and column positions to form the observation admittance matrix.

[0040] S3. When the distribution network is under dry reference conditions, calculate the dry observation admittance matrix, and based on the dry observation admittance matrix and the observation admittance matrix, obtain the condensation topology latch matrix. In an embodiment of the present invention, when the distribution network is under a dry reference operating condition, the calculation of the dry observation admittance matrix includes: Determine that the power distribution network is in a dry reference operating condition with no condensation and a stable environment; Dry reference condition refers to the operating state when no condensation, water film or other damp layer forms on the surface of power distribution equipment and the external meteorological conditions are stable. It is used as a benchmark state for comparing electrical parameters. No condensation means that there are no water droplets or continuous water film on the surface of conductors, insulators and auxiliary structures caused by changes in air humidity. Stable environment means that environmental factors such as temperature, humidity and wind speed change slowly and without sudden changes over a continuous period of time.

[0041] Specifically, on-site visual inspections are conducted on outdoor switches, insulators, bushings, and voltage sampling points corresponding to the measurement nodes to confirm the absence of water droplets, fog-like deposits, or continuous water films on their surfaces, and the inspection time and location are recorded. Meteorological observation data related to the distribution network are collected simultaneously. This data includes at least air temperature, relative humidity, wind speed, and precipitation. It is determined that the meteorological data has not shown abrupt changes over a continuous period, and that no precipitation, frost, or condensation has occurred. Simultaneously, the operational status records of the distribution network are acquired. These records include switch on / off status, feeder load changes, and voltage levels. It is determined that the operational status records have not shown frequent switching operations or significant load changes over a continuous period. The time period that satisfies the visual inspection results, the stability of the meteorological observation data, and the stability of the operational status records is defined as the dry reference operating condition time period, and the start and end times of this time period are associated and stored with the corresponding measurement node list.

[0042] Obtain the drying phase angle measurement matrix and drying current matrix under the drying reference operating conditions; The drying phase angle measurement matrix refers to the matrix data structure formed by arranging the three-phase voltage phase angles of each measurement node in the order of node number under the drying reference operating condition; the drying current matrix refers to the matrix data structure formed by arranging the three-phase current phasors of each measurement node in the order of node number under the drying reference operating condition.

[0043] Specifically, within the specified drying reference operating condition time period, three-phase voltage time series and three-phase current time series are collected at each measurement node at a unified sampling time. A Discrete Fourier Transform (DFT) is performed on the three-phase voltage time series to obtain the three-phase voltage phase angles, and a DFT is performed on the three-phase current time series to obtain the three-phase current phasors. The three-phase voltage phase angles of each measurement node at the same sampling time are written into a matrix according to the measurement node number order to form a drying phase angle measurement matrix. The row order of the drying phase angle measurement matrix is ​​consistent with the measurement node number order, and the column order is consistent with the three-phase phase order. Similarly, the three-phase current phasors of each measurement node at the same sampling time are written into a matrix according to the same measurement node number order to form a drying current matrix. The row order of the drying current matrix is ​​consistent with the measurement node number order, and the column order is consistent with the three-phase phase order.

[0044] The drying observation admittance matrix is ​​calculated based on the drying current matrix and the drying phase angle measurement matrix.

[0045] The dry observation admittance matrix refers to the admittance matrix obtained by performing matrix multiplication between the generalized inverse matrix of the phase angle measurement matrix and the current matrix under the dry reference operating condition. It is used to represent the equivalent electrical connection relationship between each measurement node and the proportional relationship between current and voltage under non-condensation conditions.

[0046] Specifically, the drying phase angle measurement matrix and the drying current matrix at the same sampling time are obtained. A generalized inverse operation is performed on the drying phase angle measurement matrix to obtain the generalized inverse matrix. The generalized inverse operation includes performing a transpose operation on the drying phase angle measurement matrix to obtain a transpose matrix, performing matrix multiplication on the transpose matrix and the drying phase angle measurement matrix to obtain a product matrix, performing an invertibility check on the product matrix and performing an inverse operation on the invertible product matrix to obtain an inverse matrix, and then performing matrix multiplication on the inverse matrix and the transpose matrix to obtain the generalized inverse matrix of the drying phase angle measurement matrix. If the product matrix is ​​not invertible, regularization is performed on the product matrix to obtain an invertible matrix, and then the inverse operation is performed and multiplied with the transpose matrix to obtain the generalized inverse matrix. After obtaining the generalized inverse matrix of the drying phase angle measurement matrix, a matrix multiplication operation is performed on it with the drying current matrix. The matrix multiplication operation performs element-wise multiplication and summation on each row of the generalized inverse matrix and each column of the drying current matrix according to the matrix multiplication rules and writes the summation to the corresponding row and column positions to form the drying observation admittance matrix.

[0047] When conductive water traces form at locations such as outdoor pole-mounted switch posts, insulators, and voltage sampling points, additional leakage admittance related to location is superimposed between existing nodes, causing a shift in the spatial distribution of the observed admittance. This shift can only be accurately identified when compared with a stable, unaffected reference admittance structure. Therefore, by calculating the dry observation admittance matrix using the dry phase angle measurement matrix and the dry current matrix under dry reference conditions, a standard admittance distribution reflecting the inherent electrical topology of the distribution network can be obtained. When subsequent real-time observation admittance matrices are differentially analyzed with this reference matrix, the additional admittance topology changes caused by the condensation water trace network can be identified, thus providing a reliable physical reference basis for real-time phase angle correction and intelligent loop-closing decision-making.

[0048] Additional leakage admittance refers to the equivalent admittance component formed by additional conductive paths caused by environmental factors outside the original conductors and insulation structure of the distribution network. When the surface of insulators, bushings, poles, or metal supports forms continuous or semi-continuous water trails due to moisture or condensation, these water trails provide additional current leakage paths between adjacent conductors or between conductors and ground. This causes the current that originally flowed only through the designed conductive path to be diverted, thereby creating a new current ratio between nodes or between nodes and ground. This admittance component generated by non-designed structures and superimposed on the original network admittance is called additional leakage admittance. Its magnitude and distribution are related to the spatial connectivity and coverage location of the water trails.

[0049] In an embodiment of the present invention, a condensation topology latch matrix is ​​obtained based on the dryness observation admittance matrix and the observation admittance matrix, including: The average admittance matrix is ​​obtained based on the observed admittance matrix; The average admittance matrix refers to the admittance matrix obtained by summing the corresponding elements of the observed admittance matrix over multiple consecutive sampling times and dividing by the number of sampling times. It is used to reflect the stable distribution of nodal admittance relationships within a certain time range.

[0050] Specifically, a sequence of observation admittance matrices corresponding to multiple consecutive sampling times is obtained, and the consistency of the observation admittance matrix sequence is checked to confirm that the set of measurement node numbers corresponding to each observation admittance matrix is ​​consistent and the row and column order is consistent. The observation admittance matrix sequence is summed element by element according to the corresponding row and column positions to obtain the admittance summation matrix. The element-by-element summation is to add the admittance elements at the same row and column position in turn at each sampling time. Scalar division is performed on the admittance summation matrix, and each element of the admittance summation matrix is ​​divided by the number of sampling times involved in the summation to obtain the average admittance matrix.

[0051] Based on the dry observation admittance matrix, the reference admittance matrix is ​​obtained; The reference admittance matrix refers to a stable baseline admittance matrix determined based on the dry observation admittance matrix after time screening and consistency verification, and is used as a standard for subsequent comparisons.

[0052] Specifically, a sequence of drying observation admittance matrices formed during the drying reference operating condition time period is obtained. The sequence is then summed element-wise according to corresponding row and column positions to obtain a drying admittance summation matrix. A scalar division operation is then performed on the drying admittance summation matrix to obtain the drying average admittance matrix. This drying average admittance matrix is ​​then output as the reference admittance matrix.

[0053] The condensation topology latch matrix is ​​obtained by subtracting the average admittance matrix from the reference admittance matrix.

[0054] The condensation topology latch matrix refers to the difference matrix obtained by subtracting the average admittance matrix from the reference admittance matrix. It is used to represent the spatial distribution of the additional admittance formed by condensation traces and the changes in the node connection relationship that remain relatively stable over a certain period of time.

[0055] Specifically, the average admittance matrix and reference admittance matrix formed under the same measurement node numbering order are obtained. The number of rows and columns of the two matrices are checked to be consistent, and the measurement node numbering order and three-phase arrangement order of the two matrices are checked to be consistent. The two matrices are subtracted element by element according to the matrix subtraction rule, that is, the reference admittance matrix element at the same row and column position is subtracted from the element at the corresponding position of the average admittance matrix. The difference is written into the corresponding row and column position to form the admittance difference matrix. The completeness of the admittance difference matrix is ​​checked. After confirming that all row and column elements have been calculated, the admittance difference matrix is ​​output as the condensation topology latch matrix.

[0056] Specifically, based on the linear relationship between node current and node voltage in the distribution network, the electrical coupling relationship between nodes can be represented by the node admittance matrix. When the network is in a dry and stable state, the node admittance matrix is ​​determined only by the conductors, cables, transformer windings, and the designed ground insulation structure. When condensation forms on the equipment surface and constitutes a continuous or semi-continuous conductive channel, it is equivalent to superimposing new parallel admittance branches or additional admittance to ground on the original network structure, which changes the proportional relationship between node current and node voltage, thereby changing the corresponding elements in the observed admittance matrix. Since the dry reference admittance matrix represents the inherent admittance structure under the condition of no additional leakage channels, while the average admittance matrix represents the actual admittance structure including the influence of additional leakage channels in the current time period, the difference matrix obtained by subtracting the two element by element is equivalent to the additional admittance matrix. This additional admittance matrix reflects the node positions connected by the water trace conductive path in space, and presents a relatively stable state in time due to the connectivity maintenance of the water trace network. Therefore, this difference matrix can characterize the conductive topology formed by condensation water traces and its connectivity that remains unchanged over a certain period of time, and is thus defined as the condensation topology latching matrix.

[0057] S4. Construct a strategy evaluation matrix based on the condensation topology latch matrix, and obtain the compensation observation admittance matrix based on the target state row in the strategy evaluation matrix. In an embodiment of the present invention, constructing a strategy evaluation matrix includes: Based on historical operating data of the distribution network, determine the set of angle difference compensation coefficients; Historical operating data refers to the set of continuously sampled data such as voltage phase angle, current phasor, admittance matrix and environmental conditions recorded during the long-term operation of the distribution network. It is used to reflect the changes in electrical relationships between nodes under different operating conditions. The set of phase angle compensation coefficients refers to a set of coefficient data obtained based on the analysis of historical operating data to correct the phase angle measurement deviation. These coefficients are used to offset the phase offset caused by the change in additional admittance.

[0058] The phase angle measurement matrix, current matrix, and their timestamps collected and stored at each measurement node during historical operation are acquired and arranged in chronological order to form a historical data sequence. The sequence of measurement node numbers and the sequence of the three-phase columns are verified to be consistent. Data segments with missing sampling points or inconsistent phase sequences are deleted. Then, taking the time period confirmed as the dry reference condition in the historical data as the range, the observation admittance matrix for each sampling moment within that time period is calculated. These observation admittance matrices are summed element-wise and then divided by the number of samplings to obtain the reference admittance matrix. Subsequently, the following processing is performed on each sampling moment in the historical data sequence: first, the observation admittance matrix for that sampling moment is calculated from the phase angle measurement matrix and current matrix; then, the observation admittance matrix is ​​subtracted element-wise from the reference admittance matrix to obtain the admittance residual matrix, while the phase angle measurement matrix for that sampling moment is retained as the object to be corrected. Next, the phase angle correction is determined with the goal of minimizing the overall energy of the admittance residual matrix. Specifically, the phase angle measurement matrix is ​​set as an unknown quantity according to the measurement node number and the three-phase sequence. The phase angle correction is added element-wise to the phase angle measurement matrix to obtain the compensated phase angle measurement matrix. Based on the compensated phase angle measurement matrix and the current matrix at the sampling time, the compensated observation admittance matrix is ​​recalculated. The compensated observation admittance matrix is ​​then subtracted element-wise from the reference admittance matrix to obtain the compensation residual matrix. The norm of the compensation residual matrix is ​​used as the evaluation quantity. Then, a least-squares solution is established for the unknown phase angle correction. A linear algebraic solution is used to write the relationship between the compensation residual matrix and the phase angle correction as a system of linear equations. Through generalized inverse operations, a set of phase angle corrections that minimizes the evaluation quantity is obtained. This set of phase angle corrections is then arranged into a vector of phase angle compensation coefficients according to the measurement node number and the three-phase sequence. Finally, the angle difference compensation coefficient vectors obtained from all historical sampling times are summarized, and duplicates are removed and merged according to the measurement node number and the three-phase order to form an angle difference compensation coefficient set.

[0059] Based on the set of angle difference compensation coefficients, a candidate angle difference compensation matrix is ​​constructed; The candidate angle difference compensation matrix refers to the matrix structure formed by arranging the set of angle difference compensation coefficients according to the measurement node number and the three-phase order, which is used to perform compensation calculations on the phase angle measurement matrix.

[0060] Specifically, a set of angle difference compensation coefficients and a list of measurement nodes are obtained. The list of measurement nodes includes the order of measurement node numbers and the three-phase order. Each set of compensation coefficient vectors in the set of angle difference compensation coefficients is read. Each compensation coefficient vector contains the three-phase compensation amounts of each measurement node arranged in the order of measurement node numbers. The row order of the matrix is ​​determined according to the measurement node number order, and the column order of the matrix is ​​determined according to the three-phase order. The first-phase compensation amount of each measurement node is written into the first column of the candidate angle difference compensation matrix, the second-phase compensation amount of each measurement node is written into the second column of the candidate angle difference compensation matrix, and the third-phase compensation amount of each measurement node is written into the third column of the candidate angle difference compensation matrix, forming a candidate angle difference compensation matrix. The above matrix filling process is repeated for all compensation coefficient vectors in the set of angle difference compensation coefficients to obtain multiple candidate angle difference compensation matrices.

[0061] The element distribution of the condensation topology latch matrix is ​​classified to obtain the condensation topology type; Condensation topology type refers to the conductive connectivity type obtained by classifying the spatial distribution pattern of the additional admittance elements in the condensation topology latch matrix, which is used to represent the connectivity structure of water trace network between different nodes.

[0062] Specifically, the sequence of condensation topology latch matrices calculated within the same time frame is obtained. The sequence of measurement node numbers for each condensation topology latch matrix is ​​verified to be consistent. Element positions are extracted from each condensation topology latch matrix, and the positions of elements with non-zero absolute values ​​and their corresponding row and column number combinations are recorded to form an element distribution set. This element distribution set is then mapped to a set of node connection relationships according to the measurement node numbers. The element distribution set is then merged based on connectivity relationships, grouping condensation topology latch matrices with the same or similar connection relationships into the same category. A unique category identifier is assigned to each category, and this category identifier is output as the condensation topology type. The condensation topology type is then associated and stored with the corresponding condensation topology latch matrix element distribution set.

[0063] A strategy evaluation matrix is ​​constructed based on the candidate angle difference compensation matrix and different condensation topology types.

[0064] The strategy evaluation matrix refers to the evaluation matrix formed by quantifying the applicability of different candidate angle difference compensation matrices under different condensation topology types. It is used to characterize the effect of each compensation strategy on restoring the consistency of admittance structure under different condensation topology conditions.

[0065] Specifically, a set of candidate angle difference compensation matrices and their index identifiers are obtained, as well as a set of condensation topology types and their category identifiers. A matrix data structure with condensation topology type as row index and candidate angle difference compensation matrix as column index is established as the storage container for the strategy evaluation matrix. Then, for each condensation topology type, historical sample data corresponding to that condensation topology type is obtained. The historical sample data includes the phase angle measurement matrix, current matrix, reference admittance matrix, and observation admittance matrix calculated from the phase angle measurement matrix and current matrix under that type. The consistency between the measurement node number order and the three-phase sequence of the historical sample data is checked. Subsequently, an evaluation calculation is performed on each candidate angle difference compensation matrix. The candidate angle difference compensation matrix is ​​added to the phase angle measurement matrix in the historical sample data to obtain the compensation phase angle measurement matrix. The compensation observation admittance matrix is ​​calculated based on the compensation phase angle measurement matrix and the corresponding current matrix. Then, the compensation observation admittance matrix is ​​subtracted element-wise from the reference admittance matrix to obtain the admittance residual matrix. The norm of the admittance residual matrix is ​​calculated to obtain the residual index. After obtaining the residual indices for all historical samples under the same condensation topology type, a summation operation is performed and divided by the number of samples to obtain the average residual index of the candidate angle difference compensation matrix under the condensation topology type. The average residual index is then written into the strategy evaluation matrix according to the row position of the corresponding condensation topology type and the column position of the candidate angle difference compensation matrix. The above calculation and writing process is repeated for all condensation topology types until all row and column elements of the strategy evaluation matrix are assigned values, thereby obtaining the strategy evaluation matrix used for subsequent strategy selection and updating.

[0066] In an embodiment of the present invention, the compensation observation admittance matrix is ​​obtained based on the target state row in the policy evaluation matrix, including: Based on the condensation topology latch matrix, the target state row in the strategy evaluation matrix is ​​determined; The target state row refers to the row of data in the strategy evaluation matrix that corresponds to the current condensation topology latch matrix, and is used to represent the evaluation results of each candidate compensation scheme under the current condensation connectivity state.

[0067] Specifically, the condensation topology latch matrix within the current time range is obtained, along with a list of condensation topology types. This list records the element distribution sets corresponding to each condensation topology type. Element position extraction is performed on the current condensation topology latch matrix to obtain the current element distribution set. This current element distribution set contains the positions of elements in the condensation topology latch matrix whose absolute values ​​are not zero, along with their corresponding row and column number combinations. The current element distribution set is then matched one by one with each element distribution set in the condensation topology type list. The matching criteria include the consistency of the element position sets and the consistency of the node connection relationships mapped from the element positions. The condensation topology type with the highest matching degree is selected, and its row index position in the policy evaluation matrix is ​​determined. The entire row of data corresponding to the row index position is then designated as the target state row.

[0068] Based on the target state row, the column index of the strategy evaluation matrix is ​​selected to obtain the candidate angle difference compensation matrix corresponding to the target state row; Specifically, the strategy evaluation matrix and the candidate angle difference compensation matrix set are obtained. The column order of the strategy evaluation matrix corresponds to the index order of the candidate angle difference compensation matrix set. The evaluation values ​​of each column in the target state row are read, and the column position with the optimal evaluation value in the target state row is determined. This column position is then output as the column index. Based on the column index, the corresponding candidate angle difference compensation matrix is ​​selected from the candidate angle difference compensation matrix set, and this candidate angle difference compensation matrix is ​​output as the compensation matrix corresponding to the target state row.

[0069] The candidate angle difference compensation matrix and phase angle measurement matrix corresponding to the target state row are added together to obtain the compensation phase angle measurement matrix; The compensation phase angle measurement matrix is ​​the matrix obtained by adding the candidate phase angle difference compensation matrix and the phase angle measurement matrix element by element, and is used to represent the corrected voltage phase angle distribution.

[0070] Specifically, obtain the phase angle measurement matrix corresponding to the current sampling time, obtain the candidate angle difference compensation matrix corresponding to the target state row, verify that the measurement node number order of the two is consistent, verify that the three-phase order of the two is consistent, and perform element-by-element addition operation on the two matrices according to the matrix addition rule, that is, add the elements of the candidate angle difference compensation matrix and the elements of the phase angle measurement matrix at the same row and column position, and write the summation result into the corresponding row and column position to form the compensation phase angle measurement matrix.

[0071] Calculate the generalized inverse matrix of the compensation phase angle measurement matrix; The generalized inverse matrix refers to the inverse mapping matrix obtained by performing linear algebraic operations on the compensation phase angle measurement matrix. It is used to establish a linear relationship between voltage phase angle and current in the sense of minimum error.

[0072] Specifically, the compensation phase angle measurement matrix is ​​obtained while maintaining the measurement node numbering order and the three-phase order. The compensation phase angle measurement matrix is ​​input into a linear algebra operation process. The transpose operation is performed on the compensation phase angle measurement matrix to obtain the transpose matrix. Matrix multiplication is performed on the transpose matrix and the compensation phase angle measurement matrix to obtain the product matrix. The invertibility of the product matrix is ​​checked. If the product matrix is ​​invertible, the inverse operation is performed on the product matrix to obtain the inverse matrix. Then, matrix multiplication is performed on the inverse matrix and the transpose matrix to obtain the generalized inverse matrix of the compensation phase angle measurement matrix. If the product matrix is ​​not invertible, regularization is performed on the product matrix to obtain an invertible matrix. The inverse operation is performed on the invertible matrix to obtain the inverse matrix. Finally, matrix multiplication is performed on the inverse matrix and the transpose matrix to obtain the generalized inverse matrix of the compensation phase angle measurement matrix.

[0073] The compensation observation admittance matrix is ​​obtained by performing matrix multiplication on the generalized inverse matrix of the current matrix and the compensation phase angle measurement matrix.

[0074] The compensated observation admittance matrix refers to the equivalent admittance relationship matrix of the nodes after the phase angle of the measured node voltage is corrected. It is used to represent the electrical coupling strength and current distribution ratio between each measured node and between the node and ground under the current operating state. Each element in the matrix reflects the correspondence between the node voltage change and the node current change under the condition that the phase angle error is corrected. It can characterize the comprehensive admittance structure under the combined action of conductors, cables, transformer windings and possible additional leakage channels.

[0075] Specifically, obtain the current matrix corresponding to the current sampling time, obtain the generalized inverse matrix of the compensation phase angle measurement matrix corresponding to the same sampling time, verify that the measurement node numbering order of the two is consistent, verify that the three-phase order of the current matrix is ​​consistent with the three-phase order of the compensation phase angle measurement matrix, perform row-by-row and column-by-column product summation operation on the current matrix and the generalized inverse matrix according to the matrix multiplication rules, and write each product element into the corresponding row and column position to form the compensation observation admittance matrix.

[0076] Under steady-state operation, the power distribution network satisfies Kirchhoff's current law and Ohm's law, exhibiting a linear relationship between node currents and voltages. This linear relationship can be represented by the node admittance matrix, where the node current equals the product of the node admittance matrix and the node voltage vector. When phase angle measurements are inaccurate, the admittance relationship established based on the measured voltage phase angle deviates from the true admittance distribution. After phase angle compensation, the resulting compensated phase angle measurement matrix more closely approximates the true phase distribution of the node voltages. By performing a linear algebraic inverse mapping operation on the compensated phase angle measurement matrix, a mapping matrix satisfying the linear relationship between current and voltage can be obtained with minimal error. Multiplying the current matrix with this inverse mapping matrix is ​​equivalent to deducing the admittance coefficients between nodes based on the linear proportional relationship between current and compensated voltage. Therefore, the result obtained by multiplying the current matrix with the generalized inverse matrix of the compensated phase angle measurement matrix is ​​the equivalent node admittance matrix satisfying Kirchhoff's current law, i.e., the compensated observation admittance matrix.

[0077] S5. Calculate the reward value of reinforcement learning based on the compensation observation admittance matrix, and update the policy evaluation matrix based on the reward value to obtain the target angle difference compensation matrix. In an embodiment of the present invention, obtaining the target angle difference compensation matrix includes: The admittance residual matrix is ​​obtained by subtracting the compensation observation admittance matrix from the reference admittance matrix. The admittance residual matrix is ​​the difference matrix obtained by subtracting the compensated observation admittance matrix from the reference admittance matrix element by element. It is used to represent the admittance offset and its spatial distribution relative to the dry reference under the current operating condition.

[0078] Specifically, the compensation observation admittance matrix and the reference admittance matrix corresponding to the same sampling time are obtained, and element-by-element subtraction is performed according to the matrix subtraction rules. That is, the element of the compensation observation admittance matrix at the same row and column position is subtracted from the element of the reference admittance matrix at the corresponding position, and the difference is written into the corresponding row and column position to form the admittance residual matrix.

[0079] The norm of the admittance residual matrix is ​​calculated to obtain the residual index; The residual index refers to the scalar value obtained by performing norm operation on the admittance residual matrix. It is used to measure the overall degree of admittance shift and reflect the quality of compensation.

[0080] Specifically, the admittance residual matrix is ​​obtained and its elements are expanded into a residual element sequence in row and column order. The magnitude of each element in the residual element sequence is calculated, and the magnitudes of each element are squared and summed to obtain the residual energy sum. The square root operation is performed on the residual energy sum to obtain the residual index.

[0081] The reward value for reinforcement learning is calculated based on the residual index. The reward value in reinforcement learning refers to the evaluation quantity obtained by mapping the residual index. It is used to quantify the effect of the current angle difference compensation selection on admittance bias suppression and serves as the basis for policy updates.

[0082] Specifically, the residual index corresponding to the current sampling time is obtained, and the residual index sequence formed in historical samples under the same condensation topology is obtained. The residual index sequence corresponds one-to-one with the index identifier of the candidate angle difference compensation matrix. The residual index sequence is statistically processed to calculate the mean and standard deviation of the residual index sequence, which are used to characterize the typical level and dispersion of the residual index under the condensation topology. The current residual index is differentially divided with the mean to obtain the residual change, and the residual change is divided by the standard deviation to obtain the normalized residual change, so that the residuals under different time periods and different admittance dimensions are comparable. The normalized residual change is sign-mapped to obtain the reward direction quantity, so that the decrease of residual corresponds to positive reward and the increase of residual corresponds to negative reward. The magnitude of the reward direction quantity and the normalized residual change quantity are multiplied to obtain the reward value of reinforcement learning.

[0083] Based on the return value, the strategy evaluation matrix is ​​updated to obtain the updated strategy evaluation matrix; The updated strategy evaluation matrix refers to the matrix obtained by correcting the evaluation elements in the strategy evaluation matrix corresponding to the current condensation topology and the current compensation scheme under the influence of the reward value. It is used to reflect the latest applicability of the compensation scheme under the current condensation connectivity state.

[0084] Specifically, the method obtains the reward value corresponding to the current sampling time, the condensation topology type identifier and the candidate angle difference compensation matrix index identifier associated with the reward value, and the current strategy evaluation matrix. The strategy evaluation matrix uses the condensation topology type identifier to determine the row position and the candidate angle difference compensation matrix index identifier to determine the column position. Based on the condensation topology type identifier, the target state row is located in the strategy evaluation matrix, and based on the candidate angle difference compensation matrix index identifier, the target column position is located in the strategy evaluation matrix. This locates the evaluation element in the strategy evaluation matrix corresponding to the current state and the current candidate angle difference compensation matrix. The current value of the evaluation element is read, and a weighted fusion operation is performed between the reward value and the current value of the evaluation element. The weighted fusion operation assigns weights to the current value of the evaluation element and the reward value respectively, and then sums them to obtain an updated value. The weights are determined by the learning rate and remain unchanged during the method execution. The updated value is written back to the position of the evaluation element, keeping the other elements of the strategy evaluation matrix unchanged, resulting in the updated strategy evaluation matrix.

[0085] Specifically, the steps for weighted fusion of the reward value and the current value of the evaluation element are as follows: First, obtain the reward value corresponding to the current sampling time. Second, read the current value of the evaluation element corresponding to the target state row and selected column index in the policy evaluation matrix. Third, obtain the learning rate, which is set during method initialization and remains a fixed value throughout the entire operation. Fourth, assign a first weight (one minus the learning rate) to the current value of the evaluation element and assign a second weight (the learning rate) to the reward value. Fifth, calculate the product of the current value of the evaluation element and the first weight, and the product of the reward value and the second weight. Sixth, sum these two products to obtain the updated value. Finally, write the updated value into the corresponding target state row and selected column index position in the policy evaluation matrix to replace the original current value of the evaluation element, thus completing a weighted fusion update based on the reward value.

[0086] Perform a column maximum search on the updated strategy evaluation matrix to obtain the target column index; The target column index refers to the column position with the best evaluation value selected within the target state row of the updated strategy evaluation matrix, which is used to indicate the number of the compensation scheme to be selected.

[0087] Specifically, the updated strategy evaluation matrix is ​​obtained, and the row position of the target state associated with the current sampling time is obtained. The row position of the target state is determined by the condensation topology type identifier. All column evaluation values ​​of the target state row in the updated strategy evaluation matrix are read and formed into an evaluation value sequence in column order. A traversal comparison operation is performed on the evaluation value sequence, recording the current maximum evaluation value and its corresponding column position. When a new column evaluation value is found that is greater than the recorded maximum evaluation value, the maximum evaluation value is updated to the new column evaluation value and the column position is updated to the new column position until the traversal is complete. The column positions recorded after the traversal are completed are output as the target column index.

[0088] The candidate angle difference compensation matrix corresponding to the target column index is used as the target angle difference compensation matrix.

[0089] The target angle difference compensation matrix refers to the candidate angle difference compensation matrix pointed to by the target column index, which is used as the final matrix for compensating and correcting the phase angle measurement matrix.

[0090] Specifically, a set of candidate angle difference compensation matrices is obtained, and it is confirmed that the numbering order of the candidate angle difference compensation matrix set corresponds one-to-one with the column order of the strategy evaluation matrix. Based on the target column index, the candidate angle difference compensation matrix with the corresponding number is located in the set of candidate angle difference compensation matrices, and the candidate angle difference compensation matrix is ​​read as the output matrix. The output matrix is ​​defined as the target angle difference compensation matrix.

[0091] Under steady-state conditions in a distribution network, the phase angles of node current and voltage satisfy a linear mapping relationship. When there is a deviation in the phase angle, the derived observed admittance matrix will deviate from the dry reference admittance matrix. The role of the phase angle compensation matrix is ​​to correct the phase angle, so that the linear relationship between current and voltage approximates the true admittance distribution again. By calculating the residual between the compensated observed admittance matrix and the reference admittance matrix, and converting this residual into a reward value to update the strategy evaluation matrix, the compensation effect is essentially measured by the magnitude of the admittance deviation. When a candidate phase angle compensation matrix minimizes the difference between the compensated observed admittance matrix and the reference admittance matrix, its corresponding reward value will tend to be optimal after multiple updates, and the evaluation value of that column in the strategy evaluation matrix will be maximized. Performing a column maximum search on the updated strategy evaluation matrix and selecting the corresponding candidate phase angle compensation matrix is ​​equivalent to selecting the correction matrix that can most effectively eliminate the additional admittance offset and restore the original node admittance relationship among all candidate compensation schemes. Therefore, the selected candidate phase angle compensation matrix is ​​the target phase angle compensation matrix.

[0092] In high-humidity environments, outdoor pole-mounted switches, insulators, and external sampling units form a spatially connected and stable conductive water trail network. This network is equivalent to adding a set of position-dependent additional leakage admittances to the original node admittance matrix, causing a structural shift in the equivalent admittance matrix of the power grid to exhibit a fixed spatial pattern over a period of time. This shift is reflected as a systematic error in the phase angle calculation results through the voltage and current measurement links. During rapid loop closing and load switching, the loop closing decision depends on the accuracy of the voltage phase angle difference between the two sides. If the structural admittance shift caused by the condensation topology latching effect is not corrected, the real-time angle difference between the two sides of the loop closing point will deviate from the actual electrical state, leading to inaccurate or even misjudged circulating current estimation. Therefore, by evaluating various candidate compensation schemes and selecting the compensation matrix that can eliminate the influence of admittance structural shift to the greatest extent, the target angle difference compensation matrix is ​​obtained to restore the consistency between the voltage phase angle and the actual power grid admittance, thereby providing a reliable real-time angle difference basis for rapid loop closing and load switching in the distribution network.

[0093] S6. Based on the target angle difference compensation matrix, perform fast loop-closing load reversal control on the loop-closing interconnection switch.

[0094] In an embodiment of the present invention, based on the target angle difference compensation matrix, fast loop-closing load switching is performed on the loop-closing interconnection switch, including: The phase angle measurement matrix and the target angle difference compensation matrix corresponding to the current sampling time are added together to obtain the online compensation phase angle measurement matrix; The online compensation phase angle measurement matrix refers to the matrix obtained by adding the target phase angle difference compensation matrix to the phase angle measurement matrix at the current sampling time element by element, which is used to reflect the spatial distribution of the corrected node voltage phase angle.

[0095] Specifically, the phase angle measurement matrix at the current sampling time is obtained, and it is confirmed that the order of measurement node numbers in each row and the order of three phases in each column of the phase angle measurement matrix are determined and remain unchanged. The target angle difference compensation matrix is ​​obtained, and it is confirmed that the order of measurement node numbers and the order of three phases in the target angle difference compensation matrix are consistent with the phase angle measurement matrix. The element-by-element addition operation is performed according to the matrix addition rule, that is, the elements of the target angle difference compensation matrix and the elements of the phase angle measurement matrix in the same row and column are summed, and the summation result is written into the corresponding row and column positions to form the online compensation phase angle measurement matrix.

[0096] Extract the phase angle data corresponding to the measurement nodes in the closed loop from the online compensated phase angle measurement matrix; The phase angle data corresponding to the measurement node in the closed loop branch refers to the voltage phase angle value corresponding to the measurement node on both sides of the fitting loop tie switch, which is used to characterize the electrical phase state on both sides of the closed loop point.

[0097] Specifically, a node list for the closed loop branch is obtained. This node list consists of measurement nodes already deployed on the lines at both ends of the closed loop tie switch. Each measurement node is assigned its row position and phase column position in the online compensation phase angle measurement matrix. Based on the node list, the corresponding row is located in the online compensation phase angle measurement matrix, and the phase angle values ​​corresponding to the first, second, and third phases in that row are read to form the three-phase phase angle data of that measurement node. The above reading process is repeated for each measurement node on both sides of the closed loop tie switch. The three-phase phase angle data of the measurement nodes on both sides are collected into two sets of phase angle data sets, and the phase angle data sets are associated and stored with the current sampling time marker and the closed loop branch identifier.

[0098] Calculate the real-time angle difference based on phase angle data; Real-time phase difference refers to the phase difference between corresponding phase voltages on both sides of a closed loop branch, used to measure the magnitude of the circulating current that may be generated at the moment of loop closure.

[0099] Specifically, the three-phase phase angle data sets of the measurement nodes on both sides of the closed loop branch are obtained, and the measurement nodes on both sides are grouped according to the electrical ends of the closed loop tie switch to form a power supply side phase angle data set and a load side phase angle data set. For each phase, angle difference calculation is performed. First, the phase angle value of that phase is read from the power supply side phase angle data set and the phase angle value of that phase is read from the load side phase angle data set. The difference between the two values ​​is calculated to obtain the initial angle difference. Then, the absolute value of the initial angle difference is taken to obtain the non-negative angle difference. The non-negative angle difference is then calculated... Phase normalization processing is performed to ensure that the angle difference falls within half a cycle. The normalization processing includes determining whether the non-negative angle difference exceeds half a cycle phase. If it exceeds half a cycle phase, the non-negative angle difference is subtracted from the cycle phase to obtain the normalized angle difference. If it does not exceed half a cycle phase, the non-negative angle difference remains unchanged. After obtaining the normalized angle differences for each of the three phases, the real-time angle difference output format is determined according to the requirements of the closed-loop safety assessment. The output format includes outputting the three-phase angle differences separately or selecting the maximum value among the three-phase angle differences as the real-time angle difference.

[0100] Based on real-time angle difference, a fast closed-loop load reversal control strategy is generated. The fast loop-closing load transfer control strategy refers to the switching operation and load transfer scheme generated based on the real-time angle difference and line operating status.

[0101] Specifically, real-time angle difference and operational information related to the closed loop branch are acquired. This operational information includes the current open / closed state of the closed loop tie switch, voltage amplitudes on both sides of the closed loop branch, frequency information on both sides of the closed loop branch, and load information of related lines in the closed loop branch. The real-time angle difference is combined with this operational information to form a set of decision input data required for the closed loop operation. Based on the constraint rules of the distribution network closed loop operation, a control strategy is generated. This control strategy includes the operation command type, operation sequence, and target path description for load transfer of the closed loop tie switch. The control strategy is then structured and output to generate a control strategy record for execution. This control strategy record includes the device identifier of the closed loop tie switch, command issuance sequence, command issuance time marker, and feeder switching description related to load transfer. The control strategy record is then associated and stored with the real-time angle difference data and the current sampling time marker for subsequent rapid closed loop load transfer control execution.

[0102] The fast loop-closing load reversal control strategy performs fast loop-closing load reversal control on the loop-closing interconnection switch.

[0103] A loop tie switch is a switching device that connects two power supply lines from different sources. Its closing action is used to establish an electrical parallel channel and realize the transfer of load between different power sources.

[0104] Specifically, firstly, the rapid loop-closing load transfer control strategy record is acquired. This record includes at least the loop-closing tie switch equipment identifier, target feeder or target branch identifier, command sequence, command type, and corresponding timestamp. Then, the operating status of the loop-closing tie switch is read, including its open / close position signal, remote signaling status, and related interlocking status information. The read results are then compared with the equipment identifier in the control strategy record for consistency. A distribution queue is generated based on the command sequence in the control strategy record, arranging the loop-closing command, disconnection command, and related load transfer switch commands sequentially. Each command is appended with its corresponding equipment identifier, operation type, and operation object identifier. Next, a closing control command is sent to the loop-closing tie switch. After the command is sent, the open / close position signal of the loop-closing tie switch is read to confirm the closing state has been formed. Simultaneously, the electrical quantity information after closing is read, including the voltage amplitude on both sides of the loop branch, three-phase current, and real-time phase angle difference data. This electrical quantity information is then associated with and stored using the current sampling time marker. Subsequently, following the order recorded in the control strategy, subsequent switching commands related to load transfer are sent to transfer the load along the target feeder or target branch. After each subsequent command is executed, the open / close position signal of the corresponding switch and the branch current change information are read to confirm that the load transfer path has been established. Finally, after the load transfer is completed, a tripping command is sent to the designated loop-breaking switch to release the temporary parallel operation state, and the open / close position signal of the loop-breaking switch is read to confirm the completion of the disconnection. This forms a closed-loop record of a rapid loop-closing load transfer operation. The closed-loop record includes the order of issuance of each command, the execution result, the changes of key electrical quantities over time, and the final state of the loop-closing tie switch. This record is used to prove that the rapid loop-closing load transfer control of the loop-closing tie switch has been completed and to provide a basis for subsequent operation analysis.

[0105] like Figure 2 The diagram shown is a functional block diagram of a real-time angle difference monitoring and intelligent decision-making system for rapid loop-closing load switching in a distribution network, provided by an embodiment of the present invention.

[0106] In this embodiment, the functions of each module / unit are as follows: The phasor acquisition module is used to construct a phase angle measurement matrix based on the three-phase voltage time series corresponding to the distribution network. The admittance modeling module is used to obtain the observation admittance matrix based on the phase angle measurement matrix; The condensation modeling module is used to calculate the dry observation admittance matrix when the distribution network is under dry reference conditions, and to obtain the condensation topology latch matrix based on the dry observation admittance matrix and the observation admittance matrix. The enhanced decision-making module is used to construct a strategy evaluation matrix based on the condensation topology latch matrix, and to obtain the compensation observation admittance matrix based on the target state row in the strategy evaluation matrix. The compensation generation module is used to calculate the reward value of reinforcement learning based on the compensation observation admittance matrix, and update the policy evaluation matrix based on the reward value to obtain the target angle difference compensation matrix. The parallel control module is used to perform rapid loop-closing load reversal control on the loop-closing interconnection switch based on the target angle difference compensation matrix.

[0107] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network, characterized in that, Includes the following steps: S1. Based on the three-phase voltage time series corresponding to the distribution network, a phase angle measurement matrix is ​​constructed; S2. Based on the phase angle measurement matrix, obtain the observation admittance matrix; S3. When the distribution network is under dry reference conditions, calculate the dry observation admittance matrix, and based on the dry observation admittance matrix and the observation admittance matrix, obtain the condensation topology latch matrix. S4. Construct a strategy evaluation matrix based on the condensation topology latch matrix, and obtain the compensation observation admittance matrix based on the target state row in the strategy evaluation matrix. S5. Calculate the reward value of reinforcement learning based on the compensation observation admittance matrix, and update the policy evaluation matrix based on the reward value to obtain the target angle difference compensation matrix. S6. Based on the target angle difference compensation matrix, perform fast loop-closing load reversal control on the loop-closing interconnection switch.

2. The real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network according to claim 1, characterized in that, Based on the three-phase voltage time series corresponding to the distribution network, a phase angle measurement matrix is ​​constructed, including: Nodes in the distribution network are selected to obtain multiple measurement nodes; Three-phase voltages are sampled at the measurement nodes to obtain a three-phase voltage time series; The three-phase voltage time series is subjected to discrete Fourier transform to obtain the phase angles of the first phase voltage, the second phase voltage, and the third phase voltage. A phase angle measurement matrix is ​​constructed based on the phase angles of the first, second, and third phase voltages.

3. The real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network according to claim 2, characterized in that, Based on the phase angle measurement matrix, the observation admittance matrix is ​​obtained, including: Three-phase current is sampled at the measurement nodes to obtain a three-phase current time series; The three-phase current time series is processed by discrete Fourier transform to obtain the first phase current phasor, the second phase current phasor, and the third phase current phasor; Construct a current matrix based on the first phase current phasor, the second phase current phasor, and the third phase current phasor; Performing a generalized inverse operation on the phase angle measurement matrix yields the generalized inverse matrix of the phase angle measurement matrix; The observation admittance matrix is ​​obtained by performing matrix multiplication on the generalized inverse matrix of the phase angle measurement matrix and the current matrix.

4. The real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network according to claim 1, characterized in that, When the distribution network is under dry reference operating conditions, the dry observation admittance matrix is ​​calculated, including: Determine that the power distribution network is in a dry reference operating condition with no condensation and a stable environment; Obtain the drying phase angle measurement matrix and drying current matrix under the drying reference operating conditions; The drying observation admittance matrix is ​​calculated based on the drying current matrix and the drying phase angle measurement matrix.

5. The real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network according to claim 1, characterized in that, Based on the dryness observation admittance matrix and the observation admittance matrix, the condensation topology latch matrix is ​​obtained, including: The average admittance matrix is ​​obtained based on the observed admittance matrix; Based on the dry observation admittance matrix, the reference admittance matrix is ​​obtained; The condensation topology latch matrix is ​​obtained by subtracting the average admittance matrix from the reference admittance matrix.

6. The real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network according to claim 1, characterized in that, Construct a strategy evaluation matrix, including: Based on historical operating data of the distribution network, determine the set of angle difference compensation coefficients; Based on the set of angle difference compensation coefficients, a candidate angle difference compensation matrix is ​​constructed; The element distribution of the condensation topology latch matrix is ​​classified to obtain the condensation topology type; A strategy evaluation matrix is ​​constructed based on the candidate angle difference compensation matrix and different condensation topology types.

7. The real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network according to claim 3, characterized in that, The compensated observation admittance matrix is ​​obtained based on the target state row in the policy evaluation matrix, including: Based on the condensation topology latch matrix, the target state row in the strategy evaluation matrix is ​​determined; Based on the target state row, the column index of the strategy evaluation matrix is ​​selected to obtain the candidate angle difference compensation matrix corresponding to the target state row; The candidate angle difference compensation matrix and phase angle measurement matrix corresponding to the target state row are added together to obtain the compensation phase angle measurement matrix; Calculate the generalized inverse matrix of the compensation phase angle measurement matrix; The compensation observation admittance matrix is ​​obtained by performing matrix multiplication on the generalized inverse matrix of the current matrix and the compensation phase angle measurement matrix.

8. The real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network according to claim 5, characterized in that, The target angle difference compensation matrix is ​​obtained, including: The admittance residual matrix is ​​obtained by subtracting the compensation observation admittance matrix from the reference admittance matrix. The norm of the admittance residual matrix is ​​calculated to obtain the residual index; The reward value for reinforcement learning is calculated based on the residual index. Based on the return value, the strategy evaluation matrix is ​​updated to obtain the updated strategy evaluation matrix; Perform a column maximum search on the updated strategy evaluation matrix to obtain the target column index; The candidate angle difference compensation matrix corresponding to the target column index is used as the target angle difference compensation matrix.

9. The real-time angle difference monitoring and intelligent decision-making method for rapid loop-closing load switching in a distribution network according to claim 1, characterized in that, Based on the target angle difference compensation matrix, fast loop closing and load switching control is performed on the loop tie switch, including: The phase angle measurement matrix and the target angle difference compensation matrix corresponding to the current sampling time are added together to obtain the online compensation phase angle measurement matrix; Extract the phase angle data corresponding to the measurement nodes in the closed loop from the online compensated phase angle measurement matrix; Calculate the real-time angle difference based on phase angle data; Based on real-time angle difference, a fast closed-loop load reversal control strategy is generated. The fast loop-closing load reversal control strategy performs fast loop-closing load reversal control on the loop-closing interconnection switch.

10. A system for real-time angle difference monitoring and intelligent decision-making in a distribution network rapid loop-closing load switching method as described in any one of claims 1-9, characterized in that, The system includes: The phasor acquisition module is used to construct a phase angle measurement matrix based on the three-phase voltage time series corresponding to the distribution network. The admittance modeling module is used to obtain the observation admittance matrix based on the phase angle measurement matrix; The condensation modeling module is used to calculate the dry observation admittance matrix when the distribution network is under dry reference conditions, and to obtain the condensation topology latch matrix based on the dry observation admittance matrix and the observation admittance matrix. The enhanced decision-making module is used to construct a strategy evaluation matrix based on the condensation topology latch matrix, and to obtain the compensation observation admittance matrix based on the target state row in the strategy evaluation matrix. The compensation generation module is used to calculate the reward value of reinforcement learning based on the compensation observation admittance matrix, and update the policy evaluation matrix based on the reward value to obtain the target angle difference compensation matrix. The parallel control module is used to perform rapid loop-closing load reversal control on the loop-closing interconnection switch based on the target angle difference compensation matrix.

Images