Low-voltage distribution network voltage prediction method, system and device and storage medium
By performing equidistant partitioning and local sparse regression of measurement data in the low-voltage distribution network, a sensitivity matrix is established, which solves the voltage fluctuation problem caused by distributed photovoltaic power generation and achieves accurate voltage prediction and stable grid operation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-27
- Publication Date
- 2026-03-31
AI Technical Summary
In low-voltage distribution networks, the high penetration rate of distributed photovoltaics increases the complexity of voltage fluctuations, making it difficult for traditional voltage regulation methods to adapt to dynamic control requirements. Furthermore, existing data-driven methods are inaccurate in predicting data sparsity and multicollinearity.
By acquiring measurement data from low-voltage distribution network nodes, performing equidistant partitioning and local sparse regression, a sensitivity matrix is established. Combined with similarity matching, voltage prediction is performed to achieve accurate voltage prediction.
This ensures the safety and stability of low-voltage distribution network operation and improves the accuracy and efficiency of voltage prediction.
Smart Images
Figure CN121769831A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of voltage prediction technology, and in particular to a method, system, device and storage medium for voltage prediction in low-voltage distribution networks. Background Technology
[0002] Currently, distributed photovoltaic (PV) power has become an important power source component of low-voltage distribution networks. The high penetration rate of distributed PV has significantly altered the unidirectional power flow characteristics of traditional low-voltage distribution networks, resulting in strong uncertainties on both the source and load sides: PV output fluctuates intermittently due to meteorological factors, leading to significant voltage deviations and exacerbating the complexity of voltage fluctuations; the distribution network topology changes frequently with the switching on and off of distributed power sources, significantly increasing the need for dynamic adjustment of operating status. Voltage stability, as a core indicator of the safe operation of low-voltage distribution networks, directly affects power supply reliability and user power quality. Traditional voltage regulation methods based on local measurements and empirical rules are no longer adequate for the dynamic control requirements of high-penetration PV integration scenarios.
[0003] Effectively obtaining the relationship between load and photovoltaic power output fluctuations and system voltage is a crucial prerequisite for solving voltage problems. Directly utilizing the nonlinear relationship between power and voltage to establish an optimization model for distribution network operation control is difficult to solve. While using voltage-power sensitivity to linearly represent the relationship between node power changes and voltage changes can transform complex and difficult-to-solve nonlinear optimization problems into linear optimization problems, enabling rapid and accurate solutions, traditional voltage-power sensitivity calculations require obtaining accurate line parameters and system state variables to calculate the power flow Jacobian matrix, and then obtaining the voltage-power sensitivity through matrix inversion. This calculation method typically suffers from significant calculation errors due to drawbacks such as inaccurate component parameters, untimely information updates, and difficulty in tracking system operating points and related topology changes.
[0004] On the other hand, with the deployment of advanced measurement devices, data-driven methods offer the possibility of circumventing model dependencies, but they also face bottlenecks in real-world applications. Ideal data-driven algorithms, such as ordinary least squares regression, require complete and independent measurement data. However, in reality, due to the high deployment cost of synchronous phasor measurement units (TPMUs) and their sparse distribution in the power distribution network, measurement data is spatially incomplete and contains significant gaps. Furthermore, the inherent spatiotemporal correlation between adjacent distributed photovoltaic systems and loads leads to severe multicollinearity in power measurement data. This not only causes inaccurate regression model results but also renders the estimated sensitivity coefficients physically ambiguous, unstable, and difficult to apply to practical systems. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention provides a method, system, device, and storage medium for predicting voltage in low-voltage distribution networks, thereby achieving the technical effect of accurate voltage prediction in low-voltage distribution networks.
[0006] In a first aspect, the present invention provides a method for predicting voltage in a low-voltage distribution network, the method comprising: Acquire measurement data of each node in a low-voltage distribution network at the current time and at multiple historical times, the measurement data including power data and voltage data; Based on the power data at the current moment and at various historical moments, historical measurement data is selected from the measurement data at historical moments. The historical measurement data is divided into multiple data subsets at equal intervals, and a historical measurement change submatrix corresponding to each data subset is obtained based on the historical measurement data. Local sparse regression is performed on the historical measurement change submatrix corresponding to each data subset to obtain the sensitivity matrix of each data subset; The measurement data at the current moment is updated to obtain the change in power data and the change in voltage data at the current moment. The change in power data at the current moment is matched with each data subset for similarity to obtain the optimal matching data subset. Based on the voltage data change at the current moment and the sensitivity matrix of the optimal matching data subset, the voltage prediction data for each node is obtained.
[0007] Furthermore, the step of filtering historical measurement data from the historical measurement data based on the power data at the current time and various historical times includes: Using the current moment as the current operating point, the distance between the current moment and the first historical moment is calculated based on the difference in power data between different moments, and the distance value is used to determine whether the current moment and the first historical moment belong to the same operating point range. If so, calculate the distance between the current time and the next historical time, and determine whether the current time and the next historical time belong to the same running point range, until a historical time that does not belong to the same running point range is found; If not, then the historical moment that does not belong to the same running point range is taken as the first historical running point, and the next historical running point is determined based on the distance between the first historical running point and other historical moments that have not yet been traversed, until all historical moments are traversed and multiple historical running points are obtained. The measurement data corresponding to each historical operation point is used as historical measurement data.
[0008] Furthermore, the step of obtaining the historical measurement change sub-matrix corresponding to each data subset based on the historical measurement data includes: Based on the historical measurement data, the power change matrix and voltage change matrix are obtained; Feature extraction is performed on the power change matrix to obtain a historical feature matrix, and the sample index of each data subset is obtained based on the extreme value range of the historical feature matrix and the number of data subsets. Based on the sample index, extract the power change submatrix and voltage change submatrix corresponding to each data subset from the power change matrix and the voltage change matrix, respectively.
[0009] Furthermore, the step of performing local sparse regression on the historical measurement change sub-matrices corresponding to each data subset to obtain the sensitivity matrix of each data subset includes: Based on the power change submatrix and voltage change submatrix corresponding to each data subset, a sparse regression prediction model for each data subset is established using the minimum absolute shrinkage and selection operator algorithm. Solve the sparse regression prediction model to obtain the sensitivity coefficient vector of all nodes corresponding to each data subset; The sensitivity coefficient vectors of all nodes corresponding to each data subset are combined to obtain the sensitivity matrix of each data subset.
[0010] Furthermore, the step of performing similarity matching between the current power data change and each data subset to obtain the optimal matching data subset includes: Based on the power change submatrix corresponding to each data subset, the centroid of each data subset is obtained; The similarity is calculated based on the distance between the change in power data at the current moment and the centroid of each data subset, and the data subset with the maximum similarity is taken as the optimal matching data subset.
[0011] Furthermore, the step of obtaining the voltage prediction data for each node based on the voltage data change at the current moment and the sensitivity matrix of the optimal matching data subset includes: Based on the voltage change at the current moment and the sensitivity matrix of the optimal matching data subset, the predicted voltage change value is obtained. Based on the updated voltage data at the current moment and the predicted voltage change value, the voltage prediction data for each node is obtained.
[0012] Furthermore, after the step of obtaining the voltage prediction data for each node, the method further includes: The voltage prediction data of each node is compared with the preset voltage limit. Based on the comparison results, it is determined whether there is a risk of voltage exceeding the limit. If so, an early warning signal is triggered.
[0013] In a second aspect, the present invention provides a low-voltage distribution network voltage prediction system, the system comprising: The data acquisition module is used to acquire measurement data of each node in the low-voltage distribution network at the current time and at multiple historical times. The measurement data includes power data and voltage data. The data filtering module is used to filter historical measurement data from the historical measurement data based on the power data at the current time and at various historical times. The data processing module is used to divide the historical measurement data into equal intervals to obtain multiple data subsets, and to obtain the historical measurement change sub-matrix corresponding to each data subset based on the historical measurement data. The sensitivity calculation module is used to perform local sparse regression on the historical measurement change submatrix corresponding to each data subset to obtain the sensitivity matrix of each data subset. The data matching module is used to update the measurement data at the current moment, obtain the change in power data and voltage data at the current moment, and perform similarity matching between the change in power data at the current moment and each data subset to obtain the optimal matching data subset. The voltage prediction module is used to obtain voltage prediction data for each node based on the voltage data change at the current moment and the sensitivity matrix of the optimal matching data subset.
[0014] Thirdly, embodiments of the present invention also provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.
[0015] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.
[0016] This invention provides a method, system, device, and storage medium for voltage prediction in low-voltage distribution networks. The invention extracts the dominant operating mode of the system from sparse measurement data using dimensionality reduction techniques, ensuring data stability. Through equidistant segmentation and local sparse regression, it achieves time-varying adaptability of the sensitivity matrix, effectively adapting to the rapid changes in the operating status of distribution networks with a high proportion of distributed photovoltaic power. Voltage prediction is performed through online matching and automatic local optimization, ensuring the accuracy of the prediction results and the efficiency of the calculation. This invention, through accurate and efficient voltage prediction, ensures the safety and stability of low-voltage distribution network operation. Attached Figure Description
[0017] Figure 1 This is a flowchart illustrating the low-voltage distribution network voltage prediction method in an embodiment of the present invention; Figure 2 This is a schematic diagram of the power simulation system architecture in the simulation experiment of this invention; Figure 3 This is a comparison chart of the simulation results in the embodiments of the present invention; Figure 4 This is a flowchart illustrating the low-voltage distribution network voltage prediction method in an embodiment of the present invention; Figure 5 This is an internal structural diagram of the computer device in an embodiment of the present invention.
[0018] Figure label: 10. Data acquisition module; 20. Data filtering module; 30. Data processing module; 40. Sensitivity calculation module; 50. Data matching module; 60. Voltage prediction module. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] Please see Figure 1 The first embodiment of the present invention proposes a low-voltage distribution network voltage prediction method, including steps S10 to S60: Step S10: Obtain measurement data of each node in the low-voltage distribution network at the current time and at multiple historical times. The measurement data includes power data and voltage data. Step S20: Based on the power data at the current time and at each historical time, filter out the historical measurement data from the historical measurement data. Step S30: Divide the historical measurement data into equal intervals to obtain multiple data subsets, and obtain the historical measurement change submatrix corresponding to each data subset based on the historical measurement data. Step S40: Perform local sparse regression on the historical measurement change submatrix corresponding to each data subset to obtain the sensitivity matrix of each data subset. Step S50: Update the measurement data at the current moment to obtain the change in power data and the change in voltage data at the current moment. Perform similarity matching between the change in power data at the current moment and each data subset to obtain the optimal matching data subset. Step S60: Based on the voltage data change at the current moment and the sensitivity matrix of the optimal matching data subset, obtain the voltage prediction data for each node.
[0021] This invention provides a voltage prediction method for low-voltage distribution networks containing distributed photovoltaic (PV) power. Currently, intelligent low-voltage distribution substation systems are generally divided into four layers: transformer layer, branch layer, meter box layer, and user layer. This invention installs a synchronous measurement device (PMU) at the user layer and an AEC controller at the PV grid connection point. The AEC controller integrates a computing chip, the PMU, a high-speed power line carrier communication (HPLC) and 4G / 5G communication module, a switch driver module, and relay outputs. The PMU measurement module achieves wide-area synchronization via GPS, accurately extracting voltage, current phasors, and power, reducing load fluctuations and improving control accuracy. The AEC controller integrates HPLC and 4G / 5G communication modules, supporting multiple interfaces for interaction with distributed power sources. The protocol converter adapts to different communication protocols and unifies data formats. Furthermore, the controller is equipped with a switch driver module and relay outputs for switching control of small PV inverters and other low-power power sources. To achieve precise time synchronization within the distribution substation area, this invention utilizes the HPLC carrier communication clock principle, enabling all intelligent integrated monitoring devices in local communication to share the network reference time, achieving precise time synchronization. By deploying synchronous measurement devices (PMUs) and AEC controllers at various nodes such as the user side and photovoltaic grid connection points, it is possible to accurately collect measurement data such as voltage amplitude and power over the current and past periods. Here, power includes active power and reactive power.
[0022] Because the PMU has a high data acquisition frequency and the system operating state does not change much between adjacent measurement times, in order to avoid matrix singularity in subsequent calculations due to the small difference between measurement data between adjacent measurement times, it is necessary to filter the measurement data from multiple historical times. During the filtering process, measurement data from historical times that differ significantly from the current time can be retained, while measurement data with small differences can be filtered out. In a preferred embodiment, the present invention employs the following steps for data filtering: Using the current moment as the current operating point, the distance between the current moment and the first historical moment is calculated based on the difference in power data between different moments, and the distance value is used to determine whether the current moment and the first historical moment belong to the same operating point range. If so, calculate the distance between the current time and the next historical time, and determine whether the current time and the next historical time belong to the same running point range, until a historical time that does not belong to the same running point range is found; If not, then the historical moment that does not belong to the same running point range is taken as the first historical running point, and the next historical running point is determined based on the distance between the first historical running point and other historical moments that have not yet been traversed, until all historical moments are traversed and multiple historical running points are obtained. The measurement data corresponding to each historical operation point is used as historical measurement data.
[0023] In this embodiment, the current time is taken as the current running point, and then the distance between two times is calculated based on the difference between the measurement data of the current time and each historical time. The expression is as follows: In the formula, D m,0 This represents the distance between the current running point and the m-th historical moment. This represents the difference in active power at the i-th node between the m-th historical moment and the current operating point. This represents the reactive power difference at the i-th node between the m-th historical moment and the current operating point. This indicates the system's baseline capacity.
[0024] Multiple historical moments are arranged in descending order of time, with the first historical moment being the closest to the current moment, and so on. Following the distance formula, the distance between the current running point and the first historical moment is first calculated. Then, it is determined whether this distance is less than a preset distance threshold. If it is less, it indicates that the system's operating state between the two moments has not changed significantly and they are within the same running point range. Therefore, the distance between the current running point and the next historical moment is calculated, and the latest distance value is used to determine if they belong to the same running point range, until a historical moment that does not belong to the same running point range as the current running point is found. If the distance is not less than the threshold, it indicates that the system's operating state between the two moments has changed significantly, and therefore they belong to different running point ranges. The historical moment that does not belong to the same running point range as the current running point is designated as the first historical running point. This means that all historical moments before the first historical running point and the current moment belong to the same running point range. Then, the distance between the first historical running point and subsequent unvisited historical moments is calculated. Based on the comparison between the distance value and the distance threshold, the next historical running point is found, and subsequent historical moments are traversed based on the next historical running point until all historical moments have been traversed, finding multiple historical running points. Each historical operating point corresponds to a set of measurement data. The number of measurement data in each set is the same as the number of nodes, meaning it includes measurement data from all nodes. These measurement data corresponding to the historical operating points are then used as historical measurement data. This can be understood as each historical operating point representing the system's operating state over a period of time, i.e., the dominant operating mode. Within this dominant operating mode, the system's operating state changes relatively little and is relatively stable. This data filtering method can identify measurement data with significant changes in the system's operating state, i.e., measurement data under different dominant operating modes, thus avoiding matrix singularity problems during subsequent calculations of the measurement data.
[0025] After selecting the historical measurement data, the dataset is divided into several consecutive subsets of equal time length. Each subset contains historical measurement data on active power, reactive power, and voltage amplitude over a given period. The changes in the historical measurement data within each subset are then analyzed to obtain a corresponding historical measurement change sub-matrix. The specific analysis steps include: Based on the historical measurement data, the power change matrix and voltage change matrix are obtained; Feature extraction is performed on the power change matrix to obtain a historical feature matrix, and the sample index of each data subset is obtained based on the extreme value range of the historical feature matrix and the number of data subsets. Based on the sample index, extract the power change submatrix and voltage change submatrix corresponding to each data subset from the power change matrix and the voltage change matrix, respectively.
[0026] In this embodiment, based on the measurement data of the current operating point and the historical measurement data of each historical operating point, the measurement change data of each historical operating point is obtained by calculating the difference between the measurement data of each historical operating point and the current operating point. The measurement change data includes changes in active power, reactive power, and voltage amplitude. Assuming there are C historical operating points, C sets of measurement change data are obtained. Then, the changes in active power and reactive power of each historical operating point are combined to form a power change matrix, and the changes in voltage amplitude are combined to form a voltage change matrix. The power change matrix X1 can be represented as X1=[△P, △U], and the voltage change matrix is represented as △U, where: In the formula, n represents the number of nodes, C represents the number of historical operating points, i.e., the number of sets of historical measurement data, and ΔP represents a matrix composed of changes in active power. Let represent the change in active power at the nth node in group C, and let ΔQ represent the matrix composed of changes in reactive power. Let ΔU represent the change in reactive power at the nth node in group C, and let ΔU represent the matrix composed of changes in voltage amplitude, i.e., the voltage change matrix. This represents the change in voltage amplitude at the nth node in group C.
[0027] For the power change matrix X1, principal component analysis is used for feature extraction to obtain the historical feature matrix. Specifically, the power change matrix X1 is transformed into a standardized power change matrix X through standardization. The covariance matrix of the standardized matrix X is calculated, and eigenvalues of the covariance matrix are obtained through eigenvalue decomposition. and the corresponding feature vector The eigenvalues are sorted in descending order, and their magnitudes represent the magnitude of the variance of the original data captured by the corresponding principal component direction, that is, the contribution of the principal component to the overall system operation volatility. The eigenvector represents a specific global system operation mode, and its component values represent the participation factors or weight coefficients of each original electrical quantity to the operation mode.
[0028] The cumulative variance contribution rate method is used to select the top k principal components, and the original data is projected onto the new principal component space to obtain the dimensionality-reduced historical feature matrix Z=XV. k Among them, V k This represents a matrix consisting of k eigenvectors.
[0029] Determine the extreme value range of the historical feature matrix Z Assuming there are F data subsets in total, then the length of each subset interval is... for: The sample index contained in the f-th data subset I f It can be represented as: In the formula, c represents the sample index including the group number. Let f represent the actual number of samples in the f-th data subset. For all complete data subsets, we have: For the final incomplete subset of data, .
[0030] Since the rows of the power change matrix, voltage change matrix, and historical feature matrix are in one-to-one correspondence, the power change sub-matrix and voltage change sub-matrix corresponding to each data subset can be extracted from the power change matrix and voltage change matrix respectively, according to the sample index of each data subset. For example, for the f-th data subset, its power change sub-matrix is: In the formula, Let X represent the power change submatrix of the f-th data subset, and let X represent the normalized power change matrix. This represents the matrix indexing operation symbol.
[0031] The above formula represents the extraction of row indices from matrix X.I f All rows, and retain all column data for those rows.
[0032] Similarly, the voltage change submatrix can be represented as: In the formula, This represents the voltage change submatrix of the f-th data subset. This represents the voltage change matrix.
[0033] For each data subset, it is treated as an independent training set. Local sparse regression is performed on its power change submatrix and voltage change submatrix to obtain the sensitivity coefficient vector of each node in that subset with respect to voltage and power. The coefficient vectors of all nodes are then combined to obtain the sensitivity matrix of that data subset. The specific steps include: Based on the power change submatrix and voltage change submatrix corresponding to each data subset, a sparse regression prediction model for each data subset is established using the minimum absolute shrinkage and selection operator algorithm. Solve the sparse regression prediction model to obtain the sensitivity coefficient vector of all nodes corresponding to each data subset; The sensitivity coefficient vectors of all nodes corresponding to each data subset are combined to obtain the sensitivity matrix of each data subset.
[0034] In this embodiment, for the power change submatrix and voltage change submatrix of each data subset, the minimum absolute shrinkage and selection operator algorithm is used to establish a sparse regression prediction model for that subset. The model expression is as follows: In the formula, Let represent the sensitivity coefficient vector of the i-th node in the f-th data subset. Let represent the voltage change vector corresponding to the i-th node in the voltage change submatrix of the f-th data subset. This represents the power change submatrix of the f-th data subset. This represents the feature value corresponding to the f-th data subset. Describing the L2 norm, This represents the L1 norm.
[0035] The above model is solved using conventional algorithms such as coordinate descent, least angle regression, or gradient descent to obtain the sensitivity coefficient vector of each node in the f-th data subset. Then, the sensitivity coefficient vectors of all nodes are combined to obtain the sensitivity matrix of the f-th data subset with respect to voltage and power. In the formula,S f This represents the sensitivity matrix of the f-th data subset. This represents the sensitivity coefficient vector of the nth node in the fth data subset.
[0036] By treating each data subset as an independent training set, and by establishing and solving a sparse regression prediction model for that data subset, the sensitivity matrix of each data subset can be obtained.
[0037] Then, the measurement data at the current moment is updated, and the change in measurement data at the current moment is obtained based on the difference between the updated and unupdated data. This includes changes in power and voltage data, with the power data change encompassing both active and reactive power changes. This can be understood as real-time updating of measurement data, using the difference between the updated and unupdated data as the change in measurement data from the previous moment to the current moment. Then, similarity matching calculations are performed between the current power data change and each data subset to find the optimal matching subset. Specific steps include: Based on the power change submatrix corresponding to each data subset, the centroid of each data subset is obtained; The similarity is calculated based on the distance between the change in power data at the current moment and the centroid of each data subset, and the data subset with the maximum similarity is taken as the optimal matching data subset.
[0038] In this embodiment, firstly, for the power change submatrix of each data subset, the centroid of that data subset is obtained by calculating the mean of each column; then, the power data change at the current moment is used as the real-time feature vector. , where X new Represents the real-time feature vector, ΔP new ΔQ represents the change in active power data at the current moment. new This represents the change in reactive power data at the current moment. The real-time feature vector is standardized using the mean and variance of historical measurement data to obtain a standardized real-time feature vector. The Euclidean distance between the standardized real-time feature vector and the centroid of each data subset is calculated, and the similarity between the vector and each data subset is determined based on the Euclidean distance. The Euclidean distance is inversely proportional to the similarity, that is, the smaller the distance, the higher the similarity. Finally, the data subset with the smallest distance, that is, the highest similarity, is selected as the optimal matching data subset at the current time.
[0039] After finding the optimal matching subset of data, the voltage value at the next time step can be predicted based on the sensitivity matrix and real-time feature vector corresponding to the optimal matching subset of data. The specific steps include: Based on the voltage change at the current moment and the sensitivity matrix of the optimal matching data subset, the predicted voltage change value is obtained. Based on the updated voltage data at the current moment and the predicted voltage change value, the voltage prediction data for each node is obtained.
[0040] In this embodiment, assuming the optimal matching data subset is the g-th data subset, the sensitivity matrix S corresponding to the optimal matching data subset is extracted. g And based on standardized real-time feature vectors Calculate the predicted voltage change value Its formula can be expressed as: The voltage change prediction is actually a matrix containing the voltage change predictions for all nodes. Then, the updated voltage data at the current time step is added to the voltage change predictions to obtain the voltage prediction data for each node at the next time step. In the formula, This represents voltage prediction data. This represents the updated voltage data at the current moment.
[0041] To ensure the safety of the distribution network operation, in a preferred embodiment, after obtaining the voltage prediction data, the present invention also performs voltage limit determination. Specifically, the voltage prediction data is compared with the voltage limit for safe operation of the power grid. Comparison, where U min U represents the lower limit of voltage. max This indicates the upper limit of voltage. If the predicted voltage value at any node exceeds the limit, an early warning signal is triggered. The control system can adjust the power of the node based on the early warning signal and the predicted voltage exceeding the limit to avoid voltage exceeding the limit, thereby improving the safety and reliability of the distribution network operation.
[0042] The following simulation experiment verifies the prediction effect of the voltage prediction method provided by this invention. In this experiment, a simulation platform was built as follows: Figure 2The simulation system shown depicts a low-voltage distribution network with a four-layer structure and 16 nodes. The first layer, node z0, is the balancing node with a constant voltage of 220V, using a reference voltage of 1.0 pu. The second layer includes nodes z1 and z2. The third layer includes nodes z3, z4, z5, and z6. The fourth layer includes eight nodes z7, z8, z9, z10, z11, z12, z13, and z14. A synchronous measurement unit (PMU) is installed at each node in the fourth layer to collect measurement data. Each node in the fourth layer is connected to a photovoltaic (PV) unit and a load, totaling eight PV units (PV1 to PV8), each with the same maximum capacity. The nodes are assumed to be equidistant, with identical cable types and load capacities. The simulation parameters are shown in Table 1. Table 1 Simulation Parameters of Power Simulation System Measurement data from the current moment and several previous historical moments are obtained through a synchronous measurement unit (PMU). Each set of measurement data includes active power, reactive power, and voltage amplitude data. Then, the measurement data is calculated according to the voltage prediction method provided in this invention to obtain the voltage prediction data for the photovoltaic nodes. The actual voltage values of the photovoltaic nodes are then obtained through power flow calculations, resulting in the predicted voltage distribution map and the actual voltage distribution map for all nodes at a given moment. Figure 3 As shown, Figure 3 The horizontal axis represents the photovoltaic node, and the vertical axis represents the voltage value. Figure 3 As can be seen, the predicted voltage almost coincides with the actual voltage, which proves the accuracy of the voltage prediction method provided by this invention.
[0043] This embodiment provides a low-voltage distribution network voltage prediction method. It extracts the dominant system operation mode from sparse measurement data using dimensionality reduction technology, effectively overcoming the modeling difficulties caused by data incompleteness and solving the instability problem of traditional least squares methods under multicollinearity data. Through equidistant segmentation and local sparse regression, continuous operating data is divided into multiple stages, and the sensitivity matrix is independently calculated in each stage, achieving time-varying adaptability of the sensitivity matrix. This solves the problem that a fixed sensitivity matrix cannot adapt to the rapid changes in the operating state of distribution networks with a high proportion of distributed photovoltaic power under high-power interference. A matching algorithm identifies the historical time period most similar to the current operating state in real time, and a dynamic model switching mechanism automatically selects the optimal local model for prediction based on the matching results, ensuring the accuracy of the prediction results while improving computational efficiency.
[0044] Please see Figure 4 Based on the same inventive concept, the second embodiment of the present invention proposes a low-voltage distribution network voltage prediction system, comprising: Data acquisition module 10 is used to acquire measurement data of each node in the low-voltage distribution network at the current time and at multiple historical times. The measurement data includes power data and voltage data. The data filtering module 20 is used to filter historical measurement data from the historical measurement data based on the power data at the current time and at each historical time. The data processing module 30 is used to divide the historical measurement data into equal intervals to obtain multiple data subsets, and to obtain the historical measurement change sub-matrix corresponding to each data subset based on the historical measurement data. Sensitivity calculation module 40 is used to perform local sparse regression on the historical measurement change submatrix corresponding to each data subset to obtain the sensitivity matrix of each data subset. The data matching module 50 is used to update the measurement data at the current moment, obtain the change in power data and the change in voltage data at the current moment, and perform similarity matching between the change in power data at the current moment and each data subset to obtain the optimal matching data subset. The voltage prediction module 60 is used to obtain voltage prediction data for each node based on the voltage data change at the current moment and the sensitivity matrix of the optimal matching data subset.
[0045] The technical features and effects of the low-voltage distribution network voltage prediction system proposed in this embodiment of the invention are the same as those of the method proposed in this embodiment of the invention, and will not be repeated here. Each module in the above-mentioned low-voltage distribution network voltage prediction system can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.
[0046] Furthermore, embodiments of the present invention also propose a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.
[0047] Please see Figure 5The diagram illustrates the internal structure of a computer device in one embodiment. This computer device can specifically be a terminal or a server. The computer device includes a processor, memory, network interface, display, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a low-voltage distribution network voltage prediction method. The display screen of the computer device can be a liquid crystal display (LCD) or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.
[0048] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computing devices may include more or fewer components than those shown in the figure, or combine certain components, or have the same component arrangement.
[0049] Furthermore, embodiments of the present invention also propose a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.
[0050] In summary, the embodiments of this invention propose a low-voltage distribution network voltage prediction method, system, device, and storage medium. The method involves acquiring measurement data of each node in the low-voltage distribution network at the current time and at multiple historical times, including power data and voltage data; filtering historical measurement data from the historical measurement data based on the power data at the current time and at each historical time; dividing the historical measurement data into multiple data subsets at equal intervals, and obtaining a historical measurement change sub-matrix corresponding to each data subset based on the historical measurement data; performing local sparse regression on the historical measurement change sub-matrix corresponding to each data subset to obtain a sensitivity matrix for each data subset; updating the measurement data at the current time to obtain the power data change and voltage data change at the current time; performing similarity matching between the power data change at the current time and each data subset to obtain the optimal matching data subset; and obtaining the voltage prediction data for each node based on the sensitivity matrix between the voltage data change at the current time and the optimal matching data subset. This invention extracts the dominant system operation mode from sparse measurement data using dimensionality reduction techniques, ensuring data stability. Through equidistant segmentation and local sparse regression, it achieves time-varying adaptability of the sensitivity matrix, effectively adapting to the rapid changes in the operating status of distribution networks with a high proportion of distributed photovoltaic power. Voltage prediction via online matching and automatic local optimization not only ensures the accuracy of the prediction results but also improves computational efficiency. This invention, through accurate and efficient voltage prediction, ensures the safety and stability of low-voltage distribution network operation.
[0051] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.
[0052] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of this invention, and these improvements and substitutions should also be considered within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the scope of the claims.
Claims
1. A low voltage distribution network voltage prediction method, characterized in that, The method comprises the following steps: acquiring measurement data of each node in a low-voltage distribution network at a current time and a plurality of historical times, wherein the measurement data comprises power data and voltage data; screening historical measurement data from the measurement data at the historical times according to the power data at the current time and the historical times; performing equidistant division on the historical measurement data to obtain a plurality of data subsets, and obtaining a historical measurement change sub-matrix corresponding to each data subset according to the historical measurement data; performing local sparse regression on the historical measurement change sub-matrix corresponding to each data subset to obtain a sensitivity matrix of each data subset; updating the measurement data at the current time to obtain a power data change and a voltage data change at the current time, performing similarity matching between the power data change at the current time and each data subset to obtain an optimal matching data subset; and obtaining voltage prediction data of each node according to the voltage data change at the current time and the sensitivity matrix of the optimal matching data subset.
2. The low voltage distribution network voltage prediction method of claim 1, wherein, The step of screening the historical measurement data from the measurement data at the historical times according to the power data at the current time and the historical times comprises the following steps: taking the current time as a current operating point, calculating a distance value between the current time and a first historical time according to a difference between the power data at different times, and judging whether the current time and the first historical time belong to a same operating point range according to the distance value; if yes, calculating a distance value between the current time and a next historical time, and judging whether the current time and the next historical time belong to the same operating point range until a historical time not belonging to the same operating point range is found; if no, taking the historical time not belonging to the same operating point range as a first historical operating point, and judging a next historical operating point according to a distance value between the first historical operating point and other historical times not yet traversed until all the historical times are traversed to obtain a plurality of historical operating points; and taking measurement data corresponding to each historical operating point as the historical measurement data.
3. The low voltage distribution network voltage prediction method of claim 1, wherein, The step of obtaining the historical measurement change sub-matrix corresponding to each data subset according to the historical measurement data comprises the following steps: obtaining a power change matrix and a voltage change matrix according to the historical measurement data; performing feature extraction on the power change matrix to obtain a historical feature matrix, and obtaining a sample index of each data subset according to an extreme value range of the historical feature matrix and a number of the data subsets; and extracting a power change sub-matrix and a voltage change sub-matrix corresponding to each data subset from the power change matrix and the voltage change matrix respectively according to the sample index.
4. The low voltage distribution network voltage prediction method of claim 3, wherein, The step of performing local sparse regression on the historical measurement change sub-matrix corresponding to each data subset to obtain the sensitivity matrix of each data subset comprises the following steps: adopting a least absolute shrinkage and selection operator algorithm to establish a sparse regression prediction model of each data subset according to a power change sub-matrix and a voltage change sub-matrix corresponding to each data subset; solving the sparse regression prediction model to obtain a sensitivity coefficient vector of all nodes corresponding to each data subset. Combine the sensitivity coefficient vectors of all nodes corresponding to each data subset to obtain a sensitivity matrix of each data subset.
5. The low voltage distribution network voltage prediction method of claim 3, wherein, The step of performing similarity matching between the power data change amount at the current moment and each data subset to obtain an optimal matching data subset comprises: Obtain a centroid of each data subset according to the power change sub-matrix corresponding to each data subset; Calculate a similarity according to the distance between the power data change amount at the current moment and the centroid of each data subset, and take the data subset corresponding to the maximum similarity as the optimal matching data subset.
6. The low voltage distribution network voltage prediction method of claim 1, wherein, The step of obtaining voltage prediction data of each node according to the voltage data change amount at the current moment and the sensitivity matrix of the optimal matching data subset comprises: Obtain a voltage change prediction value according to the voltage data change amount at the current moment and the sensitivity matrix of the optimal matching data subset; Obtain voltage prediction data of each node according to the updated voltage data at the current moment and the voltage change prediction value.
7. The low voltage distribution network voltage prediction method of claim 1, wherein, After the step of obtaining voltage prediction data of each node, the method further comprises: Compare the voltage prediction data of each node with a preset voltage limit value, judge whether there is a voltage out-of-limit risk according to the comparison result, and if there is, trigger a warning signal.
8. A low voltage distribution network voltage prediction system, characterized by, The system is applied to the method of any one of claims 1 to 7, and the system comprises: A data acquisition module configured to acquire measurement data of each node in the low-voltage power distribution network at a current moment and a plurality of historical moments, the measurement data comprising power data and voltage data; A data screening module configured to screen historical measurement data from the measurement data at the historical moments according to the power data at the current moment and the historical moments; A data processing module configured to perform equidistant division on the historical measurement data to obtain a plurality of data subsets, and obtain a historical measurement change sub-matrix corresponding to each data subset according to the historical measurement data; A sensitivity calculation module configured to perform local sparse regression on the historical measurement change sub-matrix corresponding to each data subset respectively to obtain a sensitivity matrix of each data subset; A data matching module configured to update the measurement data at the current moment to obtain a power data change amount and a voltage data change amount at the current moment, perform similarity matching between the power data change amount at the current moment and each data subset to obtain an optimal matching data subset; A voltage prediction module configured to obtain voltage prediction data of each node according to the voltage data change amount at the current moment and the sensitivity matrix of the optimal matching data subset.
9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 7.