Power distribution network carrying capacity evaluation system based on dynamic correction

Abstract

The application relates to the technical field of power distribution network evaluation, and discloses a power distribution network carrying capacity evaluation system based on dynamic correction. The system comprises five modules, namely, a dynamic data acquisition module, a multi-dimensional state space construction module, a safety domain analysis module, a partition coupling degree calculation module and a carrying capacity evaluation engine. The dynamic data acquisition module obtains power injection sequence, voltage deviation rate sequence and uncontrollable parameter fluctuation data of each partition node of the power distribution network; the multi-dimensional state space construction module maps the above sequences to generate a linearized power flow state space model containing a power-voltage Jacobian matrix; the safety domain analysis module generates a dynamic safety operation constraint set according to the boundary of the uncontrollable parameter fluctuation correction model; the partition coupling degree calculation module quantifies the electrical independence index by using a spectral radius; and the carrying capacity evaluation engine constructs an opportunity constraint optimization model, outputs the maximum accessible capacity of photovoltaic power in each partition and a safety supply control strategy set, and improves the evaluation accuracy and practicability.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power distribution network evaluation, in particular to a power distribution network carrying capacity evaluation system based on dynamic correction. BACKGROUND

[0002] With the continuous increase of the penetration rate of distributed photovoltaic and other new energy in the power distribution network, the operation state of the power distribution network presents significant volatility and uncertainty, and the traditional power distribution network carrying capacity evaluation method gradually exposes many limitations.

[0003] In terms of data collection, the traditional method relies on fixed-period static data collection, which is difficult to capture the dynamic changes of key operating parameters such as power injection of each subarea node and voltage deviation rate of the power distribution network in real time, especially in the scenarios of new energy output fluctuation and load mutation. Static data cannot timely reflect the real operation state of the power distribution network, resulting in the loss of accurate data basis for subsequent carrying capacity evaluation.

[0004] For state space construction, the traditional method often uses a simplified linear model, which does not fully consider the complex mapping relationship between node power injection and voltage deviation rate, and lacks comprehensive description of the power flow state of the power distribution network. This simplified model ignores many influencing factors, resulting in a large deviation between the constructed state space model and the actual operation of the power distribution network, and cannot provide accurate model support for carrying capacity evaluation.

[0005] In terms of safety domain analysis, the traditional method usually determines the safety domain of the power distribution network based on fixed operating parameters and static constraint conditions, and fails to effectively integrate uncontrollable parameter fluctuation data. However, in actual operation, the fluctuation of new energy output caused by changes in wind speed and light intensity, as well as the load fluctuation caused by changes in user electricity consumption behavior, will constantly change the safe operation boundary of the power distribution network. Static safety domain analysis method is difficult to adapt to such dynamic changes, which may lead to conservative evaluation results or safety hazards.

[0006] In the calculation of subarea coupling degree, the traditional method mostly ignores the electrical correlation characteristics between each subarea of the power distribution network, and fails to effectively quantify the electrical independence between subareas. Each subarea of the power distribution network does not run completely independently, and the coupling relationship between power exchange and voltage influence between subareas will directly affect the carrying capacity evaluation results of each subarea. Lack of accurate calculation of such coupling degree will make the carrying capacity evaluation at the subarea level inconsistent with the actual situation, and further affect the reasonable determination of photovoltaic access capacity.

[0007] In terms of the carrying capacity evaluation engine, the conventional evaluation model adopts a deterministic optimization method, fails to fully integrate dynamic safety operation constraints and electrical independence indicators, and is difficult to maximize the photovoltaic access capacity of each subzone under the premise of ensuring the safe operation of the distribution network. Meanwhile, the conventional model cannot output a targeted safety supply control strategy set, is difficult to guide the actual operation and regulation of the distribution network, and cannot meet the refined and dynamic needs of the distribution network for carrying capacity evaluation under high penetration of new energy. SUMMARY

[0008] The present application aims to provide a distribution network carrying capacity evaluation system based on dynamic correction to solve the problems raised in the background art.

[0009] To achieve the above-mentioned purpose, the present application provides a distribution network carrying capacity evaluation system based on dynamic correction, which comprises:

[0010] A dynamic data acquisition module is configured to acquire a real-time operation parameter set of each subzone of the distribution network, wherein the real-time operation parameter set comprises a node power injection sequence, a voltage deviation rate sequence, and uncontrollable parameter fluctuation data.

[0011] A multi-dimensional state space construction module is configured to perform complete dimensionality mapping processing on the node power injection sequence and the voltage deviation rate sequence, and generate a linearized power flow state space model of the distribution network, wherein the linearized power flow state space model comprises a power-voltage Jacobian matrix and a control variable simplified mapping table.

[0012] A safety domain analysis module is configured to perform boundary correction processing on the linearized power flow state space model according to the uncontrollable parameter fluctuation data, and generate a dynamic safety operation constraint set, wherein the dynamic safety operation constraint set comprises a subzone power limit curve and a node voltage safety threshold.

[0013] A subzone coupling degree calculation module is configured to calculate an electrical independence indicator between each subzone based on the power-voltage Jacobian matrix, wherein the electrical independence indicator is quantified by the spectral radius of the power sensitivity matrix between subzones.

[0014] A carrying capacity evaluation engine is configured to integrate the dynamic safety operation constraint set and the electrical independence indicator, construct an opportunity constraint optimization model at the subzone level, and output the maximum accessible capacity of photovoltaic for each subzone and a safety supply control strategy set.

[0015] Preferably, the multi-dimensional state space construction module comprises:

[0016] A nonlinear power flow analysis unit is configured to input the node power injection sequence into a preset nonlinear power flow equation, and generate an initial voltage distribution field and a power loss gradient field.

[0017] a dimension-increasing mapping unit, configured to perform Taylor series expansion processing on the initial voltage distribution field to extract a complete dimension-increasing state space basis composed of second-order and higher-order terms;

[0018] a control variable compression unit, configured to dynamically update a control variable simplified mapping table recording dimension-reducing association rules of active-reactive control variables according to the power loss gradient field;

[0019] a Jacobian matrix generation unit, configured to calculate a sparsification correction coefficient of a power-voltage Jacobian matrix based on the complete dimension-increasing state space basis and the dimension-reducing association rules.

[0020] Preferably, the security domain analysis module comprises:

[0021] a parameter perturbation simulation unit, configured to generate a plurality of groups of random perturbation scenarios according to the uncontrollable parameter fluctuation data, each group of perturbation scenarios containing distributed photovoltaic output fluctuation and load mutation increment;

[0022] a boundary searching unit, configured to sequentially load the random perturbation scenarios in the linearized power flow state space model, and search for a feasible region boundary point of a power injection space by using a golden section method;

[0023] a threshold fitting unit, configured to perform least squares fitting on the feasible region boundary point and a preset node voltage safety benchmark to generate a slope adjustment amount of a partition power limit curve and an offset compensation value of a node voltage safety threshold.

[0024] Preferably, the partition coupling degree calculation module comprises:

[0025] a sensitivity matrix construction unit, configured to extract mutual admittance parameters of cross-partition node pairs from the power-voltage Jacobian matrix, and construct a block diagonal submatrix of an inter-partition power sensitivity matrix;

[0026] a spectrum analysis unit, configured to calculate a maximum eigenvalue modulus of the block diagonal submatrix, and compare the maximum eigenvalue modulus with a preset electrical decoupling benchmark value to output an inter-partition coupling strength level;

[0027] an independent partition identification unit, configured to mark a corresponding partition as an electrically independent partition when the coupling strength level is lower than a dynamic decoupling threshold, and generate a partition division topology structure diagram.

[0028] Preferably, the carrying capacity evaluation engine comprises:

[0029] an uncertainty modeling unit, configured to convert the uncontrollable parameter fluctuation data into a probability density function of photovoltaic output and a Markov transition matrix of load change;

[0030] a segmental analysis unit, configured to divide the power distribution network into a plurality of load capacity evaluation segments according to the partition power limit curve, each segment corresponding to a different level of linearized power flow equation simplification;

[0031] an optimization solving unit, configured to embed the probability density function and the Markov transition matrix in the chance-constrained optimization model, and solve the maximum accessible capacity of each partition by using a branch and bound method;

[0032] a strategy generating unit, configured to output a differentiated control instruction sequence for each electrically independent partition according to a superimposed result of the security supply control strategy set and the partition division topology graph.

[0033] Preferably, the system further comprises a real-time correction module, configured to periodically collect deviation data of an actual photovoltaic access capacity and a theoretical maximum accessible capacity of the power distribution network, and generate a weight correction coefficient of the chance-constrained optimization model;

[0034] The real-time correction module comprises:

[0035] a deviation analysis unit, configured to calculate a root mean square error of an actual photovoltaic consumption rate and a predicted consumption rate, and adjust a confidence interval width of the probability density function according to the root mean square error;

[0036] a model iteration unit, configured to feed back the weight correction coefficient to the segmental analysis unit, and trigger dynamic reorganization of the level of linearized power flow equation simplification.

[0037] Preferably, the system further comprises a topology adaptive module, configured to recalculate a coupling relationship between the electrical independence index and the dynamic safe operation constraint set when detecting a topology structure change event of the power distribution network;

[0038] The topology adaptive module comprises:

[0039] an event response unit, configured to identify a topology change flag triggered by a newly added distributed power supply access point or a line reconfiguration operation;

[0040] a fast reconfiguration unit, configured to locally update mutual admittance parameters related to a changed area in the power-voltage Jacobian matrix, and synchronously adjust node membership of the partition division topology graph.

[0041] Preferably, the system further comprises a multi-time scale coordination module, configured to perform hierarchical synchronization on execution periods of the dynamic data acquisition module and the load capacity evaluation engine according to differences in update frequencies of the power distribution network operation states;

[0042] The multi-time scale coordination module comprises:

[0043] A fast-slow channel separation unit is configured to divide the node power injection sequence into a millisecond-level sampling channel and a minute-level sampling channel.

[0044] A clock alignment unit is configured to establish a timestamp mapping relationship between different sampling channels in the chance-constrained optimization model, and ensure the timeliness consistency of the dynamic safe operation constraint set.

[0045] Preferably, the system further comprises a data-driven compensation module configured to start a non-complete dimension-increasing data-driven compensation mechanism when it is detected that the fitting residual of the linearized power flow state space model exceeds a threshold value.

[0046] The data-driven compensation module comprises:

[0047] A residual monitoring unit is configured to compare the absolute deviation between the actual measured voltage value and the model predicted voltage value.

[0048] A compensation activation unit is configured to embed the power-voltage mapping relationship in the historical operation data in the form of a lookup table into the Jacobian matrix generation unit when the absolute deviation continuously exceeds the residual tolerance.

[0049] Preferably, the system further comprises a dynamic weight adjustment module configured to adjust the weight distribution of each sub-area in the chance-constrained optimization model in real time according to the change of the operation condition of the power distribution network.

[0050] The dynamic weight adjustment module comprises:

[0051] A condition recognition unit is configured to detect a load mutation event or a distributed power output abnormal event of the current power distribution network.

[0052] A weight optimization unit is configured to dynamically calculate the weight coefficient of each sub-area in the chance-constrained optimization model based on the severity of the event, wherein the weight coefficient of the key load concentration area is positively correlated with the voltage sensitivity.

[0053] Compared with the prior art, the present application has the following advantages:

[0054] The dynamic data acquisition module can acquire the node power injection sequence, the voltage deviation rate sequence and the uncontrollable parameter fluctuation data of each sub-area of the power distribution network in real time, breaking the limitations of traditional static data acquisition. Real-time data acquisition can timely capture the dynamic changes in the operation process of the power distribution network. Whether it is the random fluctuation of new energy output or the sudden change of load, it can be quickly perceived and recorded, providing comprehensive and accurate data input for subsequent state space construction, safety domain analysis and carrying capacity evaluation, ensuring that the entire evaluation process is always based on the real operation state of the power distribution network.

[0055] The multi-dimensional state space construction module fully maps the node power injection sequence and the voltage deviation rate sequence to generate a linearized power flow state space model containing a power-voltage Jacobian matrix and a control variable simplified mapping table. This fully mapping processing method fully explores the complex correlation between the node power injection and the voltage deviation rate. Compared with the traditional simplified model, the linearized power flow state space model can more comprehensively and accurately depict the power flow operating state of the distribution network. The power-voltage Jacobian matrix can clearly reflect the sensitive relationship between power change and voltage change, and the control variable simplified mapping table provides convenience for subsequent model calculation and analysis, so that the state space model has high accuracy and good practicability, and lays a reliable model foundation for subsequent security domain analysis and carrying capacity evaluation.

[0056] The security domain analysis module performs boundary correction processing on the linearized power flow state space model according to the uncontrollable parameter fluctuation data to generate a dynamic security operation constraint set containing a partition power limit curve and a node voltage safety threshold. This module fully considers the uncertainty factors in the operation of the distribution network, adjusts the security operation boundary in real time through the uncontrollable parameter fluctuation data, and changes the disadvantage that the traditional static security domain analysis cannot adapt to dynamic changes of parameters. The dynamic security operation constraint set can be updated in real time with the fluctuation of uncontrollable parameters, so that the security domain of the distribution network always matches the actual operating environment, avoiding both the low evaluation result of carrying capacity caused by too conservative static constraints and the security risk caused by not considering parameter fluctuations, and ensuring that the distribution network can be in a safe and controllable state under various operating scenarios.

[0057] The partition coupling degree calculation module calculates the electrical independence index between each partition based on the power-voltage Jacobian matrix, and quantifies it through the spectral radius of the partition power sensitivity matrix, effectively solving the problem of ignoring the partition coupling relationship in the traditional method. The electrical independence index can accurately reflect the electrical correlation degree between each partition, and the spectral radius quantization method makes this correlation have a clear numerical representation. By accurately mastering the electrical independence between each partition, the mutual influence between partitions can be fully considered when evaluating the carrying capacity of each partition, avoiding the evaluation deviation caused by ignoring the coupling relationship, and making the carrying capacity evaluation results of each partition more consistent with the actual operating characteristics of the distribution network, providing an important basis for the reasonable allocation of subsequent photovoltaic access capacity.

[0058] The carrying capacity evaluation engine fuses the dynamic safe operation constraint set and the electrical independence index, constructs an opportunity constraint optimization model at the partition level, and outputs the maximum accessible capacity of each partition and the safety supply control strategy set. The opportunity constraint optimization model can seek the optimal photovoltaic access capacity scheme under the consideration of various uncertain factors, which not only guarantees the safe operation of the distribution network, but also maximizes the photovoltaic access potential of each partition. Compared with the traditional deterministic optimization model, this model is more suitable for the operation characteristics of the distribution network under high penetration of new energy, and the output of the maximum accessible capacity of photovoltaic has higher practicability and rationality. At the same time, the safety supply control strategy set provides specific operation guidance for the operation and control of the distribution network, so that the distribution network can maintain stable operation and guarantee the safety and reliability of power supply through effective control strategies after accessing a large amount of photovoltaic, and further improve the access capacity and operation management level of the distribution network to new energy. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 The timing diagram of the distribution network carrying capacity evaluation system based on dynamic correction described in the application;

[0060] Figure 2 The flowchart for the work of the multi-dimensional state space construction module;

[0061] Figure 3 The multi-stage key parameter data flowchart for the distribution network carrying capacity evaluation;

[0062] Figure 4 The flowchart for the work of the carrying capacity evaluation engine. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0064] Please refer to Figure 1The application provides a power distribution network carrying capacity evaluation system based on dynamic correction, which comprises the following steps: acquiring real-time operation parameter sets of each subarea of the power distribution network through a dynamic data acquisition module, wherein the sets comprise a node power injection amount sequence, a voltage deviation rate sequence and uncontrollable parameter fluctuation data; performing complete dimension increasing mapping processing on the node power injection amount sequence and the voltage deviation rate sequence by a multi-dimensional state space construction module to generate a linearized power flow state space model of the power distribution network, wherein the model comprises a power-voltage Jacobian matrix and a control variable simplified mapping table; performing boundary correction processing on the linearized power flow state space model according to the uncontrollable parameter fluctuation data by a security domain analysis module to generate a dynamic security operation constraint set, wherein the set comprises a subarea power limit curve and a node voltage safety threshold; calculating an electrical independence index between each subarea based on the power-voltage Jacobian matrix by a subarea coupling degree calculation module, wherein the index is quantified through a spectral radius of a subarea power sensitivity matrix; and fusing the dynamic security operation constraint set and the electrical independence index by a carrying capacity evaluation engine to construct an opportunity constraint optimization model at a subarea level and output a maximum accessible capacity of photovoltaic and a safety supply control strategy set of each subarea. The system realizes dynamic evaluation and correction of the carrying capacity of the power distribution network through modular design, and ensures that the evaluation result is adapted to real-time operation conditions.

[0065] Example 1: see Figure 2The operation of the multi-dimensional state space construction module starts with the processing of the node power injection sequence provided by the dynamic data acquisition module. The nonlinear power flow analysis unit imports this sequence into the preset nonlinear power flow equation for solving. The nonlinear power flow equation adopts a power balance equation in polar coordinate form, and the solving process uses an improved Newton-Raphson algorithm to enhance convergence. The rated voltage of the distribution network is used as the initial estimated value of the node voltage during algorithm initialization. During the iterative calculation process, the norm of the power imbalance is monitored in real time. When the norm is less than the preset convergence precision threshold, it is determined that the calculation has converged, thereby generating an initial voltage distribution field covering all nodes in the network. The distribution field contains the voltage amplitude and phase angle information of each node. At the same time, the unit calculates the total power loss in the network based on the converged power flow solution, and further derives the power loss gradient field through the adjoint network method or direct partial differentiation method. The gradient field quantifies the sensitivity of the total loss to the small change of the node injection power. The dimensionality mapping unit then performs in-depth processing on the obtained initial voltage distribution field. The core operation is to perform Taylor series expansion, which takes the current system operating point, i.e., the initial voltage distribution field, as the base point. Not only the first-order term relied on traditional linearization is retained, but also the second-order term, third-order term, and even higher-order terms are systematically extracted. These high-order terms collectively constitute a fully dimensionally upgraded state space base. The expansion order of the Taylor series is not fixed, but is adaptively adjusted according to the deviation of the current operating point from the rated state. When the voltage deviation rate sequence shows that the system operating state deviates far from the rated point, a higher-order expansion is automatically used to more accurately capture the nonlinear characteristics. The construction of the fully dimensionally upgraded state space base is essentially to form a tensor structure containing high-order partial derivatives of voltage with respect to power.

[0066] The control variable compression unit dynamically maintains a control variable simplified mapping table in parallel, and the update of the mapping table is based on the power loss gradient field output from the nonlinear power flow analysis unit. An online feature extraction algorithm is integrated in the unit, which continuously analyzes the internal correlation between active control variables and reactive control variables. The power loss gradient field reveals the influence mode of different control variable combinations on system loss. Based on these modes, the unit uses principal component analysis technology to reduce the dimension of the original high-dimensional control variable space and identify the dominant control direction that has the greatest impact on the system state. The reduced correlation rules are recorded in the control variable simplified mapping table, which clearly specifies how to map the original active and reactive control variable set to a new variable space with significantly reduced dimension. For example, the associated control variables of adjacent nodes are aggregated to reduce the number of decision variables of the optimization problem and improve the subsequent calculation efficiency. The Jacobian matrix generation unit, as the final output link of the module, is responsible for combining the power-voltage Jacobian matrix. Its generation process deeply integrates the fully dimensioned state space basis provided by the dimension increasing mapping unit and the reduced dimension correlation rules provided by the control variable compression unit. The generation process is not a simple calculation of the traditional power flow Jacobian matrix, but forms a complete Jacobian structure containing high-order information on the fully dimensioned state space basis. The structure is compressed and approximated using the reduced dimension correlation rules, and the elements that still play a dominant role in the reduced space are calculated. For a large number of elements close to zero in the matrix, the unit will apply sparsification technology for processing, and calculate the sparsification correction coefficient to selectively ignore these small interactions. A power-voltage Jacobian matrix is generated, which can reflect the nonlinear characteristics of the system and is convenient for numerical calculation, and the structure is relatively sparse.

[0067] The security domain analysis module then performs boundary correction on the linearized power flow state space model. The parameter disturbance simulation unit then begins to work, reading uncontrollable parameter fluctuation data. This data typically includes the historical fluctuation statistics of distributed photovoltaic power output and the probability distribution of load changes. This unit uses the Monte Carlo simulation method to generate a large number of representative random disturbance scenarios. Each disturbance scenario contains a random photovoltaic power output fluctuation sequence and a potentially related load mutation increment sequence. The generation of the scenarios considers the temporal correlation and the spatial correlation between disturbances at different nodes to ensure that the simulated disturbance scenarios can cover various uncertainty modes that may occur in actual operation. The boundary search unit bears the critical task of determining the safe operating boundary of the system. It loads a large number of random disturbance scenarios generated by the parameter disturbance simulation unit into the constructed linearized power flow state space model. For each loaded disturbance scenario, the unit adopts the golden section method, a single-variable optimization search algorithm, to perform an efficient one-dimensional search in the power injection space. The search direction is usually along the direction from the current operating point to the power injection limit. The golden section method accurately locates the boundary point of the feasible region by continuously narrowing the search interval. This boundary point is defined as the limit value of power injection that the system can withstand under the condition of satisfying all node voltage constraints and line thermal stability constraints. This process is repeated in all key power injection directions to outline the boundary contour of the safe operating domain of the system.

[0068] The threshold fitting unit then performs mathematical processing on a large number of feasible domain boundary points obtained from the search. These boundary points are scattered in a high-dimensional power-voltage space. The unit associates these boundary points with a preset node voltage safety benchmark, which is usually specified by the operating procedures and represents the upper and lower limits that each node voltage must maintain. The fitting process uses the least squares method, aiming to find a set of linear relationship parameters such that the power limit curve determined by this relationship can optimally fit all the searched boundary points. By minimizing the fitting error, the unit calculates the slope adjustment amount required for the partition power limit curve. This adjustment amount reflects the actual impact of the allowed power injection changes in the partition on the critical node voltage under the current network structure and operating state. At the same time, the unit also calculates the offset compensation value of the node voltage safety threshold. This compensation value is used to fine-tune the theoretical voltage safety upper limit to absorb the errors caused by model linearization and the effects of system uncertainties. Finally, it outputs dynamically updated partition power limit curves and node voltage safety thresholds, which constitute the core content of the dynamic safe operation constraint set and provide accurate boundary conditions for subsequent load-bearing capacity assessment.

[0069] See Figure 3In the integration and visualization of the multi-dimensional state space construction and the security region analysis of the power distribution network, the technical implementation relies on the collaborative processing of dynamic data acquisition and multi-dimensional state mapping. In the specific operation, subgraph (a) is based on the power injection time series curve of nodes A-C, and the quantitative process of non-stationary operation characteristics is displayed through blue, orange and green lines, where the power injection sequence is used as the input dimension for state space construction; subgraph (b) uses a dashed line to present the voltage deviation rate fluctuation, and the heterogeneity of the stability margin is quantified by calculating the spatio-temporal variance of the voltage deviation between nodes, supporting the dynamic correction of the security region boundary; subgraph (c) realizes the mapping of partition power limit and voltage safety threshold through a column-line composite chart, where the safety threshold range is set to 1.06-0.94pu, and the boundary slope is optimized by least squares fitting. In the quantitative process, the power injection sequence and the voltage deviation rate data are mapped into a linearized power flow state space model through complete dimensionality lifting, and the sparsification correction coefficient of the Jacobian matrix is dynamically calculated based on the node coupling degree; the golden section method is used to locate the feasible region in the power injection space for the security region boundary search, and the fitting error between the boundary point and the voltage reference is evaluated by the root mean square index.

[0070] Example 2: see Figure 4 The start of the partition coupling degree calculation module depends on the power-voltage Jacobian matrix generated by the multi-dimensional state space construction module. The sensitivity matrix construction unit analyzes this matrix, and its core task is to extract parameters from this global matrix that can reflect the electrical interaction strength between different distribution network partitions. The unit has a preliminary partitioning scheme for the distribution network preset, which is usually initially defined based on the physical topology of the power grid (such as feeders, substation power supply range). The extraction process focuses on the boundary nodes connecting different partitions, specifically identifying and extracting mutual admittance parameters from the complete Jacobian matrix whose row index belongs to one partition and whose column index belongs to another partition. These parameters quantitatively describe the direct impact on the voltage of nodes in another partition when the power of nodes in one partition changes. The extracted mutual admittance parameters are used to construct the block diagonal submatrix of the inter-partition power sensitivity matrix. This submatrix is not a matrix that only contains the main diagonal blocks in the traditional sense, but a macroscopic model that considers the entire network as composed of multiple partition subsystems. Each sub-block corresponds to the sensitivity relationship of a partition itself, and the block diagonal submatrix constructed by the module specifically contains these sub-blocks located on the "off-diagonal line" that represent the coupling relationship between partitions, thereby forming a simplified model that clearly shows the interaction relationship between partitions.

[0071] The spectral analysis unit then deeply analyzes the block-diagonal submatrix of the constructed inter-partition power sensitivity matrix. The unit calculates the maximum eigenvalue modulus of the submatrix using numerical methods such as the QR algorithm to ensure stability and accuracy. The size of the maximum eigenvalue modulus directly reflects the overall strength of all inter-partition coupling, which is a global quantitative indicator. The modulus value is compared with a pre-set electrical decoupling reference value, which is a threshold value set based on a large amount of historical operation data, network structure characteristics, and stability requirements. The comparison result is quantified as an inter-partition coupling strength level, such as "high coupling", "medium coupling", and "low coupling". If the maximum eigenvalue modulus is much higher than the reference value, it is determined that the system partition coupling is tight, and the mutual influence is significant. If the modulus value is close to or lower than the reference value, it indicates that the electrical independence between partitions is relatively strong. The independent partition identification unit performs partition identification operations based on the coupling strength level output by the spectral analysis unit. It has a dynamic decoupling threshold that can be fine-tuned according to the overall system operation risk level. When the coupling strength level of a partition and all other associated partitions is calculated to be lower than the dynamic decoupling threshold, the unit marks this partition as an electrically independent partition. After marking is completed, the unit generates a partition division topology structure diagram that clearly indicates the range of all marked electrically independent partitions, their connection relationships, and the location of boundary nodes in a graphical manner. This topology diagram is an important basis for subsequent partition autonomous control and load carrying capacity independent assessment.

[0072] The load carrying capacity assessment engine is the core component of the system's final output assessment result. The uncertainty modeling unit probabilistically processes the uncontrollable parameter fluctuation data input into the system. It fits the historical fluctuation data of distributed photovoltaic output into a continuous probability density function, usually using a mixture Gaussian model to better characterize its randomness and intermittency. For load changes, the unit uses a discrete state Markov chain for modeling. By analyzing the state transition law of historical load sequences, it constructs a Markov transition matrix that describes the probability of load transitioning from the current state to various possible future states. The segmented analysis unit divides the entire power distribution network operating range into multiple continuous load carrying capacity assessment segments based on the partition power limit curve provided by the safety domain analysis module, such as light load segment, normal load segment, and heavy load segment. Each segment corresponds to a different level of linearized power flow equation simplification. In the light load segment, the system has weak nonlinearity, and a highly simplified linear model can be used to improve calculation speed. In the heavy load segment, the system operates close to the stability limit, and the nonlinearity is significant, so a refined model that retains more high-order terms is used to ensure assessment accuracy. This segmented processing approach achieves an adaptive balance between calculation efficiency and model accuracy.

[0073] The optimization solving unit is responsible for solving the core chance-constrained optimization model, which takes the maximum accessible capacity of photovoltaic as the optimization objective, and requires the probability of events such as node voltage out-of-limit and line overload to be lower than a given confidence level in the form of probability (i.e. chance constraint); the unit embeds the probability density function generated by the uncertainty modeling unit and the Markov transition matrix into the constraint conditions, so that the model can fully consider the inherent uncertainty of photovoltaic output and load demand; solving such a complex optimization problem containing probability constraints usually adopts the branch and bound method, which gradually narrows down the search range by continuously dividing the feasible region into smaller subsets (branching), and calculating the upper and lower bounds of the objective function of each subset (bounding), and finally finds the global optimal solution or approximate optimal solution of the maximum accessible capacity of photovoltaic in each partition that meets all the chance constraints. The strategy generation unit finally superimposes and analyzes the calculation results of the optimization solving unit and the partition division topological structure diagram generated by the partition coupling degree calculation module, generates a sequence of differentiated control instructions for each electrically independent partition based on the degree of electrical independence of each partition, the maximum accessible capacity of photovoltaic, and the specific equipment conditions within the partition (such as the presence or absence of reactive power compensation devices, the importance of load, etc.); for partitions with high electrical independence and strong internal resource regulation capacity, the control strategy may focus more on using local reactive power resources for voltage support; for partitions with high coupling degree or concentrated key loads, the control strategy will contain more conservative load shedding and photovoltaic output limiting measures, and these instruction sequences constitute a set of specific executable safety and supply guarantee control strategies.

[0074] The role of the real-time correction module is to enable the entire evaluation system to evolve continuously, which periodically collects actual operation data from the distribution network monitoring system, the core is to obtain the actual photovoltaic access capacity of each partition and the theoretical maximum accessible capacity calculated by the carrying capacity evaluation engine, the system compares the two sets of data to generate a deviation data sequence, which reflects the difference between model prediction and actual consumption capacity; the deviation data is sent to an adaptive filter for smoothing to eliminate short-term disturbances of measurement noise, and then a correction coefficient for adjusting the weight of the internal parameters of the chance-constrained optimization model is generated based on the processed deviation sequence, which serves as a feedback signal to narrow the gap between the model and reality. The deviation analysis unit inside the module is responsible for handling the uncertainty of the data, which calculates the root mean square error between the actual photovoltaic consumption rate and the model predicted consumption rate within a statistical period, the actual photovoltaic consumption rate is calculated from the actual power generated by the meter and the load power, and the predicted consumption rate comes from the results of the last optimization solving, the formula for calculating the root mean square error is:

[0075]

[0076] where: N represents the total number of sampling points in an evaluation period, k is the sampling point index, is the actual PV consumption rate at the kth sampling moment, is the predicted consumption rate at the corresponding moment; the error value The size of directly determines the direction and width of the confidence interval of the probability density function in the uncertainty modeling unit. When the error is large and persistent, the unit will appropriately relax the confidence interval of the probability density function, acknowledging the existence of insufficiently recognized uncertainty in the model, and vice versa, narrowing the interval to obtain more accurate optimization solutions.

[0077] The model iteration unit is responsible for converting the correction coefficient into model structure updates. It receives the weight correction coefficient from the bias analysis unit, which is a multi-dimensional vector. Different components correspond to the importance weight of different partitions or different constraint conditions in the chance-constrained optimization model. The unit feeds this coefficient back to the segmented analysis unit inside the carrying capacity evaluation engine, triggering its dynamic restructuring process for the simplification level of the power distribution network linearized power flow equation. The restructuring logic depends on the change pattern of the weight coefficient. If the weight correction coefficient of a certain partition increases significantly, it indicates that the model error of this region has a greater impact on the overall evaluation. The segmented analysis unit will divide the evaluation section corresponding to this partition into more detailed sections and use higher-order linear models in this section to improve local accuracy. Conversely, it may merge sections or use simpler models to improve overall calculation speed.

[0078] The topology adaptive module ensures that the system can quickly respond when the physical structure of the power distribution network changes without restarting the complete modeling process. It continuously monitors signals from the power grid dispatch center or the power distribution automation system to detect topology structure change events, which usually include the addition of a distributed power supply access point, network restructuring caused by line switch position changes, or equipment maintenance out of operation. Once such an event is detected, the module immediately starts the recalculation process, focusing on analyzing whether the coupling relationship between the electrical independence index and the dynamic safe operation constraint set after the change has changed in nature. The event response unit in the module acts as a trigger, which identifies topology change flags by analyzing communication messages sent by the power grid energy management system or monitoring the remote signal change of a specific switch. The addition of a distributed power supply access will trigger the "source point increase" flag, while line restructuring operations will trigger the "connection relationship change" flag. Each flag is accompanied by a timestamp and an impact area range identifier. The fast restructuring unit starts working after the flag is triggered. Its design principle is to perform local updates rather than global restructuring to maximize efficiency. The unit locates the areas that need to be modified in the power-voltage Jacobian matrix according to the event impact area identifier, which usually only involves a few nodes directly connected to the changed node. The unit uses the pre-calculated node admittance matrix change to quickly calculate the incremental change of the corresponding mutual admittance parameters in the power-voltage Jacobian matrix using matrix perturbation theory, and only updates these parameters to avoid recalculating the entire large matrix. While updating the electrical parameters, the unit synchronously adjusts the partition division topology structure graph generated by the partition coupling degree calculation module to re-determine the membership of the nodes around the changed point. For example, the closing of a tie switch may cause two previously independent electrical independent partitions to merge into a new partition. The unit will quickly refresh the partition topology graph based on the updated coupling degree calculation to ensure the real-time accuracy of the network model.

[0079] The real-time correction module and the topology adaptive module work together to form the self-correction mechanism of the system. The real-time correction module focuses on soft adjustment of model parameters in response to slow drift or uncertain mode changes in operating parameters, which is a continuous fine-tuning process. The topology adaptive module, on the other hand, deals with hard changes in the network physical connection relationship and performs hard correction of the model structure. Together, the two mechanisms ensure that the carrying capacity assessment system can maintain the reliability and practicality of its assessment results in a dynamically changing real power grid environment. Fine-tuning prevents the model from gradually deviating from reality, while hard correction prevents the model from failing instantly due to structural mutations.

[0080] Embodiment 4: The multi-time scale coordination module is designed to solve the fusion problem caused by the inconsistency of different source data update frequencies in the power distribution network. The core function of this module is to classify and synchronize the execution cycle of the dynamic data acquisition module and the carrying capacity evaluation engine. In a specific application example of a city industrial park power distribution network, the data sources include high-precision synchronous phasor measurement units (PMU) installed at key nodes and widely deployed power distribution automation terminals (FTU). The PMU provides node voltage and current phasor data with an update rate of 100 frames per second, forming a millisecond-level sampling channel, while the FTU collects line power, switch state and other data with an update cycle of typically 2 to 5 minutes, forming a minute-level sampling channel. The fast and slow channel separation unit classifies the input node power injection sequence in real time through the timestamp and source device identifier in the data packet header. The millisecond-level channel data is marked as high priority and directly stored in a high-speed ring buffer for real-time analysis, while the minute-level channel data is marked as regular priority and sent to a regular database for persistent storage. This separation mechanism effectively avoids the problem of low-speed data blocking high-speed data processing flow. The clock alignment unit is the key to ensuring the consistency of data time effectiveness. It maintains a high-precision system logical clock in its internal. The unit stamps a uniform logical timestamp on each arriving data packet. For example, at 14:05:30.800, it may receive a millisecond-level data packet collected by PMU at 14:05:30.800 and a minute-level data packet collected by FTU at 14:05:25.000 (the collection period of this FTU is 5 minutes). The alignment unit does not discard the delayed low-speed data, but uses a prediction interpolation algorithm based on the smoothness of power grid state changes. It uses the trend of recent high-speed data to forward interpolate the low-speed data on the logical time axis, estimates its approximate value at the latest logical timestamp, so that the data of different channels can be fused and calculated under the same time reference, providing a time-consistent dynamic safe operation constraint set snapshot for the carrying capacity evaluation engine.

[0081] The data-driven compensation module serves as a safety guarantee when the accuracy of the theoretical model decreases. Its implementation can also be illustrated by a specific scenario. When the industrial park distribution network is affected by a rapidly moving cloud layer in the afternoon of summer, the distributed photovoltaic output fluctuates dramatically. This fluctuation mode may exceed the typical scenario library based on historical data constructed by the linearized power flow state space model. The residual monitoring unit is continuously working. It compares the actual voltage value measured by the PMU at the key node with the voltage value predicted by the model based on the current node power injection, calculates the absolute deviation between the two, and sets a dynamically adjusted residual tolerance threshold for each node. This threshold is usually associated with the historical fluctuation standard deviation of the voltage at the node. The residual tolerance refers to the maximum allowed absolute deviation value between the actual measured voltage value at each node of the distribution network and the predicted voltage value by the linearized power flow state space model. When the absolute deviation continuously exceeds the residual tolerance, it indicates that the theoretical calculation results of the model deviate too much from the actual operating state of the power grid, and the data-driven compensation mechanism needs to be started to correct the model error and ensure the accuracy of the subsequent carrying capacity evaluation results. The residual tolerance is calculated based on the historical operating data of the node voltage. The actual measurement voltage data of the target node in the past 90 days (the statistical period can be adjusted according to the stability of the power grid operation, and the stable operation of the power grid can be extended to 180 days, and the frequent fluctuation of the power grid can be shortened to 30 days) is extracted (t is the time index, unit: h), a total of M data points, using the current linearized power flow state space model, the node power injection sequence at the corresponding time is back substituted, and the historical predicted voltage data of the node in the past 90 days is calculated , the historical voltage deviation sequence of the node is calculated , and the standard deviation of the sequence is calculated . The residual tolerance is 1.5 times the standard deviation of the historical voltage deviation sequence. This value not only covers the voltage deviation fluctuation range of the normal operation of the power grid, but also effectively identifies the model error beyond the conventional fluctuation, avoiding the false triggering of the compensation mechanism. At the beginning of each month, the residual monitoring unit of the data-driven compensation module automatically repeats the above calculation steps to update the residual tolerance of each node. If a node triggers the data-driven compensation mechanism more than 3 times in the last month (indicating that the node operating state fluctuates greatly, and the original residual tolerance may be too small), the residual tolerance calculation coefficient of the node in the next month is adjusted from 1.5 to 1.8; if a node has not triggered the compensation mechanism in the last month (indicating that the node operating state is stable, and the original residual tolerance can be appropriately tightened), the calculation coefficient is adjusted from 1.5 to 1.2, realizing the dynamic adaptation of the residual tolerance to the actual operating characteristics of the power grid.

[0082] Referring to Table 1, the residual monitoring of three key nodes at consecutive sampling times is described.

[0083] Table 1: Node voltage prediction residual monitoring

[0084] Node number Time stamp Measured voltage (p.u.) Model predicted voltage (p.u.) Absolute deviation Residual tolerance (p.u.) Out of limit N101 T1 1.032 1.028 0.004 0.010 No N101 T2 1.045 1.029 0.016 0.010 Yes N205 T1 0.981 0.985 0.004 0.015 No N205 T2 0.978 0.986 0.008 0.015 No N307 T1 1.058 1.051 0.007 0.012 No N307 T2 1.062 1.050 0.012 0.012 Yes

[0085] The compensation activation unit decides according to the output of the residual monitoring unit. When it monitors that the absolute deviation not only exceeds its individual tolerance, and such over-limit state continues to occur within a preset time window, as nodes N101 and N307 at T2 time, the unit determines that the pure theoretical model has failed to accurately track the actual state of the system. At this time, the module starts the non-complete dimensionality increasing data-driven compensation mechanism. The so-called "non-complete dimensionality increasing" means that this mechanism does not try to rebuild a complex new model, but quickly retrieves the historical scenarios similar to the current operating point from the historical operation database. These historical scenarios contain the corresponding relationship between the actual power injection and voltage measurement of each node under similar total load and similar total photovoltaic output level. The compensation mechanism embeds the "power-voltage" mapping relationship formed by these historical data points into the calculation process of the Jacobian matrix generation unit in a table lookup manner. When the Jacobian matrix elements calculated by the theoretical model are used for state prediction, the results obtained by data-driven table lookup are weighted and averaged. The weight is dynamically adjusted according to the current residual size. The larger the residual, the higher the weight of the data-driven part, so as to use historical experience to quickly correct the instantaneous model error.

[0086] The collaborative work of the multi-time scale coordination module and the data-driven compensation module reflects the strategy of the system to deal with complex operating conditions. The former solves the problem of the unity of data in time, providing high-quality and synchronous input for the evaluation engine. The latter serves as a "safety net" for model accuracy. When the theoretical model is inaccurate due to the deviation of the system operating point from the conventional or encountering unmodeled dynamics, the reliability of the overall evaluation result is maintained by introducing data-driven experience. This design makes the distribution network carrying capacity evaluation system not only rely on accurate physical models, but also has the ability to learn from historical data and adaptively adjust, enhancing its robustness in real uncertain environments.

[0087] The function of the dynamic weight adjustment module is to realize adaptive resource allocation in the power distribution network carrying capacity evaluation process, and the core is to dynamically adjust the weight coefficients of each partition in the opportunity constraint optimization model according to the real-time changing power grid operation condition. The implementation process of the optimization model starts from the real-time perception of the operation condition and the quantitative evaluation of the event severity. The condition recognition unit in the dynamic weight adjustment module continuously monitors the real-time data stream of the power distribution network. When events such as sudden increase in power demand in key load concentration area or abnormal drop in distributed power output are detected, the unit will immediately extract and classify the features of the event. The evaluation of the severity of the event is not a single-dimensional judgment, but a comprehensive consideration of multiple factors such as the proportion of power shortage relative to the load base of the partition, the time gradient of power change, and the social and economic importance level of the affected load, forming a quantitative severity index. The weight optimization unit dynamically adjusts the weight coefficients of each partition in the opportunity constraint optimization model according to the calculated severity index and the node voltage sensitivity parameters obtained from the power-voltage Jacobian matrix. The adjustment process follows the two principles of voltage stability priority and key load protection. For partitions with high voltage sensitivity and important load, their weight coefficients will be significantly improved, which directly affects the reconstruction of the objective function of the optimization model. In the calculation process, the optimization solver will give the constraint conditions of high-weight partitions a larger violation penalty factor, so that the solving process naturally tilts towards the direction of prioritizing the safe operation conditions of these partitions. The optimization solver unit operates in the model embedded with dynamic weights. Its core is to solve a mathematical programming problem that maximizes the global photovoltaic accessible capacity while meeting a series of weighted safety constraints. Due to the introduction of weight coefficients, the landscape of the original objective function changes, and the solver needs to use an iterative algorithm to gradually approach the optimal solution. In each iteration, the solver checks the voltage constraints, line capacity constraints, and other conditions of each partition. However, for high-weight partitions, the tolerance of these constraints is set very low, even zero, while for low-weight partitions, there is a certain possibility of constraint relaxation in extreme cases. This differentiated processing mechanism ensures that under the condition of limited global computing resources, the power supply safety and power quality of the most critical areas are always prioritized. The model solution generates not a fixed photovoltaic access capacity value, but a set of optimal solution sets corresponding to the current weight distribution. The weight optimization unit analyzes the overall system operation state under this solution set. If the weight adjustment leads to excessively low resource utilization in some non-critical partitions or the overall optimization objective function value deteriorates too much, the unit will start the weight coefficient fine-tuning program. By observing the response of the system's overall performance through small-step exploratory weight changes, it finds a balance point that effectively safeguards the safety of critical areas while considering the overall consumption capacity. Finally, it outputs a set of photovoltaic maximum accessible capacity schemes and corresponding safety control strategies that match the dynamic weights.

[0088] Consider a distribution network partitioning case with residential, commercial and an important hospital, on a sunny weekday afternoon, the load is relatively stable, the initial weight of each partition in the optimization model can be set based on its proportion of peak load, the commercial area has a higher weight, followed by the residential area; however, as the evening approaches, the residential load rises sharply, while a cloud covers the photovoltaic power station, causing a sharp drop in output, this simultaneous change in load and power constitutes a typical operating condition mutation event. The operating condition recognition unit continuously monitors such events, which analyzes real-time data streams from the data acquisition system, the unit is equipped with multiple filters and comparators for detecting load mutation events, the principle is to calculate the instantaneous rate of change of the total load of each partition, when the rate of change exceeds the threshold value calculated based on historical data (for example, the change exceeds 15% of the average load of the partition per minute), a load mutation flag is triggered, the flag information includes the mutation partition identification, the change amount and direction (increase or decrease); for distributed power output abnormal events, the detection logic is similar, but focuses on the rate and amplitude of output decline, when the photovoltaic output drops by more than 30% of its maximum output within a few minutes, a power abnormality flag is triggered, the unit can distinguish between planned output adjustment and unexpected abnormal fluctuations.

[0089] The weight optimization unit starts calculating immediately after receiving the event flag from the operating condition recognition unit, the unit is embedded with a weight allocation algorithm, the algorithm evaluates the severity of the event, which is a comprehensive indicator determined by the event type (load mutation or power loss), the impact range (involving several partitions), the power shortage size (in megawatts), and the speed of the event; for example, the load in the partition where the hospital is located increases significantly due to an emergency treatment task, while a main distributed photovoltaic line supplying power to this partition trips due to a fault, this event is considered high severity because it involves critical load and large power shortage. The calculation of weight coefficients follows the principle of voltage sensitivity priority, the unit extracts the voltage sensitivity parameter of each partition center node or key node from the power-voltage Jacobian matrix, which quantifies the impact of power balance changes on voltage stability; for critical load concentration areas such as hospitals, the requirement for voltage stability is extremely high, so the voltage sensitivity value is larger, the weight optimization unit makes the weight coefficient of the partition positively related to the voltage sensitivity, which is realized through a variant of the weighted least squares method, when constructing the objective function of the optimization model, the violation cost of the voltage deviation, power overrun and other constraint conditions of the high-weight partition is set much higher than that of the low-weight partition, which makes the solver prioritize the power supply safety and power quality of high-weight partitions during global optimization, even if it means more strictly limiting the photovoltaic access capacity of other non-critical partitions or taking more aggressive load shedding measures.

[0090] The implementation process of dynamic weight adjustment is a closed-loop feedback. The weight coefficient is not set once and for all, but the module will continue to monitor the system response after weight adjustment. If the voltage stability of the key partition is improved after adjustment, but the overall photovoltaic consumption capacity is too low, the weight optimization unit will fine-tune the weight coefficient to find a balance point. On the contrary, if the overall system operation is still unstable, the weight of the key partition will be further increased. This dynamic adjustment ensures that when the power distribution network is dealing with unexpected situations, limited control resources (such as reactive power compensation devices, energy storage systems, and interruptible loads) can be prioritized to ensure the power supply reliability of the most critical areas, thereby achieving the best balance between safety and economy on a global level. The carrying capacity evaluation result is not only a static number, but also a dynamic decision-making basis that can adapt to changes in the grid state.

[0091] It should be noted that the relational terms herein, such as first and second, are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0092] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A power distribution network carrying capacity assessment system based on dynamic correction, characterized in that, The method comprises the following steps: A dynamic data acquisition module is used to acquire a real-time operation parameter set of each subarea of a power distribution network, wherein the real-time operation parameter set comprises a node power injection sequence, a voltage deviation rate sequence, and uncontrollable parameter fluctuation data; A multi-dimensional state space construction module is used to perform complete dimensionality elevation mapping processing on the node power injection sequence and the voltage deviation rate sequence, and generate a linearized power flow state space model of the power distribution network, wherein the linearized power flow state space model comprises a power-voltage Jacobian matrix and a control variable simplified mapping table; A security domain analysis module is used to perform boundary correction processing on the linearized power flow state space model according to the uncontrollable parameter fluctuation data, and generate a dynamic security operation constraint set, wherein the dynamic security operation constraint set comprises a subarea power limit curve and a node voltage safety threshold; A subarea coupling degree calculation module is used to calculate an electrical independence index between subareas based on the power-voltage Jacobian matrix, wherein the electrical independence index is quantified by a spectral radius of a subarea-to-subarea power sensitivity matrix; A carrying capacity evaluation engine is used to fuse the dynamic security operation constraint set and the electrical independence index, construct an opportunity constraint optimization model at a subarea level, and output a maximum accessible capacity of photovoltaic and a safety supply control strategy set of each subarea; The subarea coupling degree calculation module comprises: A sensitivity matrix construction unit is used to extract mutual admittance parameters of a cross-subarea node pair from the power-voltage Jacobian matrix, and construct a block diagonal submatrix of a subarea-to-subarea power sensitivity matrix; A spectral analysis unit is used to calculate a maximum eigenvalue modulus of the block diagonal submatrix, compare the maximum eigenvalue modulus with a preset electrical decoupling reference value, and output a coupling strength level between subareas; An independent subarea identification unit is used to mark a corresponding subarea as an electrically independent subarea when the coupling strength level is lower than a dynamic decoupling threshold, and generate a subarea division topology structure diagram.

2. The dynamic correction based power distribution network carrying capacity assessment system as claimed in claim 1, wherein, The multi-dimensional state space construction module comprises: A nonlinear power flow analysis unit is used to input the node power injection sequence into a preset nonlinear power flow equation, generate an initial voltage distribution field and a power loss gradient field; A dimensionality elevation mapping unit is used to perform Taylor series expansion processing on the initial voltage distribution field, and extract a second-order and above term to constitute a complete dimensionality elevation state space basis; A control variable compression unit is used to dynamically update the control variable simplified mapping table according to the power loss gradient field, wherein the control variable simplified mapping table records a dimensionality reduction correlation rule of active-reactive control variables; A Jacobian matrix generation unit is used to calculate a sparsification correction coefficient of the power-voltage Jacobian matrix based on the complete dimensionality elevation state space basis and the dimensionality reduction correlation rule.

3. The power distribution network carrying capacity assessment system based on dynamic correction of claim 2, wherein, The security domain analysis module comprises: A parameter perturbation simulation unit is used to generate a plurality of groups of random perturbation scenarios according to the uncontrollable parameter fluctuation data, wherein each group of perturbation scenarios comprises a distributed photovoltaic output fluctuation and a load mutation increment; A boundary search unit is used to load the random perturbation scenarios in the linearized power flow state space model one by one, and search for a feasible region boundary point of a power injection space by using a golden section method. A threshold fitting unit is configured to perform least square fitting on the feasible region boundary points and a preset node voltage safety reference to generate a slope adjustment amount of a partition power limit curve and an offset compensation value of a node voltage safety threshold.

4. The dynamic correction based power distribution network carrying capacity assessment system as claimed in claim 1, wherein, The carrying capacity evaluation engine comprises: An uncertainty modeling unit is configured to convert the uncontrollable parameter fluctuation data into a probability density function of photovoltaic output and a Markov transition matrix of load variation; A segmented analysis unit is configured to divide the power distribution network into a plurality of carrying capacity evaluation segments according to the partition power limit curve, each segment corresponding to a different level of linearized power flow equation simplification; An optimization solving unit is configured to embed the probability density function and the Markov transition matrix in the chance-constrained optimization model, and solve the maximum accessible capacity of each partition by using a branch and bound method; A strategy generating unit is configured to superimpose the safety supply control strategy set and the partition division topology graph to output a differentiated control instruction sequence for each electrically independent partition.

5. The power distribution network carrying capacity assessment system based on dynamic correction of claim 4, wherein, Further comprising: A real-time correction module is configured to periodically collect deviation data of actual photovoltaic access capacity and theoretical maximum accessible capacity of the power distribution network, and generate a weight correction coefficient of the chance-constrained optimization model; The real-time correction module comprises: A deviation analysis unit is configured to calculate a root mean square error of actual photovoltaic consumption rate and predicted consumption rate, and adjust a confidence interval width of the probability density function according to the root mean square error; A model iteration unit is configured to feed back the weight correction coefficient to the segmented analysis unit to trigger dynamic reorganization of the level of linearized power flow equation simplification.

6. The dynamic correction based power distribution network carrying capacity assessment system as claimed in claim 5, wherein, Further comprising: a topology adaptive module is configured to re-calculate the coupling relationship between the electrical independence index and the dynamic safety operation constraint set when detecting a topology structure change event of the power distribution network; The topology adaptive module comprises: An event response unit is configured to identify a topology change flag triggered by a newly added distributed power supply access point or a line reconstruction operation; A fast reconstruction unit is configured to locally update mutual admittance parameters related to the changed area in the power-voltage Jacobian matrix, and synchronously adjust the node membership relationship of the partition division topology graph.

7. The dynamic correction based power distribution network carrying capacity assessment system as claimed in claim 6, wherein, Further comprising: A multi-time scale coordination module is configured to grade and synchronize the execution period of the dynamic data acquisition module and the carrying capacity evaluation engine according to the difference in the update frequency of the power distribution network operation state; The multi-time scale coordination module comprises: A fast-slow channel separation unit is configured to divide the node power injection sequence into a millisecond-level sampling channel and a minute-level sampling channel; A clock alignment unit is configured to establish a timestamp mapping relationship of different sampling channels in the chance-constrained optimization model to ensure the timeliness consistency of the dynamic safety operation constraint set.

8. The power distribution network carrying capacity assessment system based on dynamic correction of claim 7, wherein, Further comprising: A data-driven compensation module is configured to start a non-complete dimension-up data-driven compensation mechanism when detecting that the fitting residual of the linearized power flow state space model exceeds a threshold value; The data-driven compensation module comprises: A residual monitoring unit is configured to compare the absolute deviation between the actual measured voltage value and the model predicted voltage value; A compensation activation unit is configured to embed the power-voltage mapping relationship in the historical operation data in a table form into the Jacobian matrix generation unit when the absolute deviation continuously exceeds the residual tolerance.

9. The dynamic correction based power distribution network carrying capacity assessment system as claimed in claim 1, wherein, Further comprising: A dynamic weight adjustment module is configured to adjust the weight distribution of each sub-zone in the chance-constrained optimization model in real time according to the change of the operation condition of the power distribution network. The dynamic weight adjustment module comprises: An operation condition identification unit is configured to detect a load mutation event or a distributed power output abnormal event of the current power distribution network. A weight optimization unit is configured to dynamically calculate the weight coefficient of each sub-zone in the chance-constrained optimization model based on the event severity, wherein the weight coefficient of the key load concentration sub-zone is positively correlated with the voltage sensitivity.

Images