Intelligent power grid adaptive closed-loop power flow optimization control method

Abstract

The invention discloses a self-adaptive closed-loop power flow optimization control method for a smart power grid, and relates to the technical field of power grid optimization control, and the method comprises the following steps: S1, collecting data, and building a dynamic power flow optimization scheme library; s2, extracting characteristic parameters, and mapping the characteristic parameters into a power grid power flow characteristic mapping matrix; s3, setting a dynamic weighting rule in combination with a real-time working condition, and constructing a BP neural network model iterative optimization matrix; s4, performing similarity matching on the optimization matrix and a scheme library according to a preset rule, and performing screening and sorting to determine an optimal power flow optimization scheme; s5, in combination with historical parameter deviation compensation, debugging the power flow control equipment in stages, and collecting a power flow change value after debugging; and S6, presetting a power flow safety threshold value for safety verification, and dynamically adjusting the scheme according to a result to form closed-loop optimization. By acquiring the multi-source core data of the smart power grid, the problems of one-sided data acquisition and poor adaptability to the real-time working condition of the power grid are avoided.

Description

Technical Field

[0001] This invention relates to the field of power grid optimization control technology, and more particularly to an adaptive closed-loop power flow optimization control method for smart grids. Background Technology

[0002] In the era of deep integration between smart grids and multi-source data fusion technology, power flow fluctuations are a core control challenge in smart grid operation. The demand for dynamic optimization control of power flow fluctuations continues to grow, and grid operators are increasingly demanding higher accuracy, adaptability to operating conditions, and economic efficiency in power flow control. Multi-source data fusion technology, with its ability to integrate heterogeneous data from the generation, transmission, and user sides, demonstrates unique advantages in power flow optimization. By integrating core elements such as power flow fluctuation parameters, real-time operating condition data, and historical control experience from each side, a multi-dimensional dynamic optimization system is formed.

[0003] The core significance of power grid flow optimization driven by multi-source data fusion lies in breaking through the limitations of traditional single-source data, achieving dynamic regulation with higher control accuracy, faster response speed and lower power loss. By utilizing the complementarity of power generation fluctuations, transmission loss characteristics and user load changes, multi-source data fusion technology can simultaneously cover the dynamic needs of each link, solving the problem of lagging regulation caused by single data in traditional optimization.

[0004] However, existing smart grid adaptive closed-loop power flow optimization control methods lack unified multi-source data association rules and operating condition adaptation mechanisms when in use. They fail to establish a precise mapping relationship between power flow fluctuation characteristics, real-time operating conditions, and control parameters, resulting in poor adaptability of optimization schemes to the real-time operating status of the power grid. Furthermore, they lack standardized processes across multiple stages, and the selection of safety thresholds for power flow control parameters is relatively simplistic, making it difficult to comprehensively avoid the risks of power flow exceeding limits and equipment overload.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] To address the problems in related technologies, this invention proposes an adaptive closed-loop power flow optimization control method for smart grids, in order to overcome the aforementioned technical problems existing in existing related technologies.

[0007] To achieve the above objectives, the specific technical solution adopted by the present invention is as follows: The adaptive closed-loop power flow optimization control method for smart grids includes the following steps: S1. Acquire multi-source core data, real-time power grid operating parameters, and historical power flow control parameters from the power generation side, transmission side, and user side of the smart grid, and establish a smart grid dynamic power flow optimization scheme library, wherein the historical optimization schemes in the smart grid dynamic power flow optimization scheme library are associated with the feature matrix in a one-to-one correspondence. S2. Extract the power flow fluctuation feature parameters from the multi-source core data, preset the multi-source data fusion association rules, and map the power flow fluctuation feature parameters into a power grid power flow feature mapping matrix according to the multi-source data fusion association rules. S3. Based on the real-time operating parameters of the power grid, set multi-dimensional dynamic weighting rules, construct an optimization model using the BP neural network algorithm, and iteratively optimize and adjust the power flow feature mapping matrix of the power grid with the dual optimization objectives of minimizing power grid power loss and reducing the probability of power flow exceeding limits, to obtain the optimized power flow feature mapping matrix. S4. Preset dynamic power flow scheme matching rules, and perform feature similarity matching between the optimized power flow feature mapping matrix and the smart grid dynamic power flow optimization scheme library based on the dynamic power flow scheme matching rules, and filter and sort to obtain smart grid dynamic power flow optimization schemes. S5. Combine the aforementioned smart grid dynamic power flow optimization scheme with the deviation compensation logic of historical power flow control parameters, perform phased debugging of the smart grid power flow control equipment, and collect the power flow change value of the grid after debugging in real time. S6. Preset a power flow change safety threshold, perform safety verification on the power flow change value of the power grid based on the power flow change safety threshold, and dynamically adjust the smart grid dynamic power flow optimization scheme according to the safety verification result to form a closed-loop optimization.

[0008] Preferably, in step S1, acquiring multi-source core data includes: collecting output fluctuation rate and power stability coefficient from the power generation side through the new energy monitoring system and unit monitoring platform; collecting line load rate and voltage deviation rate from the transmission side through the SCADA system and WAMS system; and collecting load mutation rate and electricity consumption period distribution coefficient from the user side through smart meters and load monitoring terminals.

[0009] Preferably, in step S2, the power flow fluctuation characteristic parameters include output fluctuation rate, power stability coefficient, line load rate, voltage deviation rate, load mutation rate, and electricity consumption period distribution coefficient.

[0010] Preferably, in step S2, the preset multi-source data fusion association rules include: setting the weight of power generation side data to 40%, the weight of transmission side data to 30%, and the weight of user side data to 30%, and configuring initial weights for the power flow fluctuation characteristic parameters of each side, wherein the power generation side output fluctuation rate weight is 0.25, the power stability coefficient weight is 0.15, the transmission side line load rate weight is 0.2, the voltage deviation rate weight is 0.1, the user side load mutation rate weight is 0.15, the electricity consumption period distribution coefficient weight is 0.15, and the sum of the initial weights of all characteristic parameters is 1.

[0011] Preferably, in step S3, the optimization model constructed using the BP neural network algorithm includes: constructing a three-layer structure of input layer-hidden layer-output layer, wherein the number of nodes in the input layer is the total number of elements in the power flow characteristic mapping matrix, the number of nodes in the output layer is 2 and corresponds to the predicted power loss value and the predicted power flow over-limit probability value, respectively, and the number of nodes in the hidden layer is determined by empirical formula based on the number of nodes in the input layer and the number of nodes in the output layer.

[0012] Preferably, the empirical formula is: ; Where n is the number of hidden layer nodes, m is the number of input layer nodes, k is the number of output layer nodes, and a is an adjustment constant between 10 and 20.

[0013] Preferably, in step S4, the dynamic power flow scheme matching rule is a cosine similarity algorithm, and the formula for the cosine similarity algorithm is: , Where A is the optimized power flow feature mapping matrix, and B is the feature matrix of historical optimized schemes in the scheme library.

[0014] Preferably, the preset feature similarity screening threshold is 0.85. When the calculated feature similarity is greater than or equal to 0.85, the corresponding historical optimization scheme is included in the candidate optimization scheme set.

[0015] Preferably, step S5, which involves phased commissioning of the smart grid power flow control equipment, includes: a pre-commissioning phase, a fine-tuning phase, and a stability verification phase.

[0016] Preferably, in step S6, the preset power flow change safety threshold is set based on the power grid safe operation standard and historical stable operating condition data. The power flow change safety threshold includes a power loss safety range, a power flow over-limit probability safety range, and a voltage frequency fluctuation safety range.

[0017] The beneficial effects of this invention are as follows: 1. This invention acquires multi-source core data of the smart grid, real-time operating parameters of the grid, and historical power flow control parameters, and pre-sets a dynamic power flow optimization scheme library for the smart grid. This avoids the problems of one-sided data acquisition and poor adaptability of control schemes to real-time grid operating conditions in traditional power flow optimization, and reduces the risk of excessive power loss or power flow exceeding limits due to fixed parameter matching. At the same time, by clarifying the corresponding relationship between power flow fluctuation characteristic parameters and core power flow control parameters on the generation side, transmission side, and user side, a dynamic weight fusion rule for multi-source data is established to ensure that the adaptation logic of power flow characteristics and control parameters conforms to the operating principle of multi-source heterogeneous data of the smart grid, avoids interference from invalid parameter combinations, and improves the accuracy of power flow characteristic mapping matrix construction.

[0018] 2. This invention constructs a three-level optimization system consisting of dynamic weight adjustment, matrix weighting correction, and abnormal data removal. This system eliminates risk parameters that exceed the safe operating range of the power grid, addresses the pain point of ignoring the differences in data fluctuations and individual operating conditions on each side, improves the safety and personalized adaptability of the power flow optimization scheme, and reduces equipment debugging costs and time by reusing historical parameters. This reduces power grid operating losses and improves the stability of power flow operation, making the scheme more valuable for power system implementation and large-scale promotion.

[0019] 3. This invention achieves hierarchical matching of optimized schemes by pre-setting dynamic power flow scheme matching rules, combined with multi-dimensional weighted calculations and feature similarity screening. Based on the safety priority rule, a set of safe candidate schemes is selected from the optimized feature matrix to clarify the safety bottom line of power flow control. Then, through the optimal treatment rule and historical adaptation rule, the set of safe candidate schemes is accurately matched with the scheme library. An iterative optimization verification mechanism ensures the reliability of the matching results and avoids optimization deviations caused by a single rule. At the same time, the matched optimized schemes are combined with historical power flow control parameters to debug the power flow control equipment, realizing a closed-loop process and improving the operability of the optimized schemes. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a smart grid adaptive closed-loop power flow optimization control method according to an embodiment of the present invention. Detailed Implementation

[0022] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0023] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0024] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, the smart grid adaptive closed-loop power flow optimization control method according to an embodiment of the present invention includes the following steps: S1. Acquire multi-source core data from the power generation side, transmission side, and user side of the smart grid, real-time operating parameters of the power grid, and historical power flow control parameters, and establish a smart grid dynamic power flow optimization scheme library, with historical optimization schemes in the smart grid dynamic power flow optimization scheme library corresponding one-to-one with the feature matrix; Specifically, the power generation side collects data such as output fluctuation rate and power stability coefficient through the new energy monitoring system and unit monitoring platform; the transmission side relies on the SCADA system and WAMS system to obtain parameters such as line load rate and voltage deviation rate; and the user side collects information such as load mutation rate and electricity consumption period distribution coefficient through smart meters and load monitoring terminals.

[0025] Real-time power grid operating parameters (voltage, frequency, load factor) are collected in real time through a global sensor network. Historical power flow control parameters are retrieved from the power grid dispatch historical database, including records of equipment debugging deviations and control response delays. The establishment of the solution library requires integrating previously effective optimization solutions, extracting the power flow feature mapping matrix (generated by multi-source data fusion) corresponding to each solution, and establishing a one-to-one correspondence between "historical optimization solutions and feature matrices" through an indexing mechanism. This ensures that the solution library can quickly match and adapt solutions based on the current operating condition feature matrix.

[0026] S2. Extract the power flow fluctuation feature parameters from the multi-source core data, preset the multi-source data fusion association rules, and the core of the multi-source data fusion association rules is to match the weight allocation of each side of the data to their actual impact on the power flow fluctuation, and map the power flow fluctuation feature parameters into the power grid power flow feature mapping matrix according to the association rules. In this embodiment of the application, the steps of extracting power flow fluctuation feature parameters from the multi-source core data, pre-setting multi-source data fusion association rules, and the core of the multi-source data fusion association rules being to match the weight allocation of each side of the data to their actual impact on power flow fluctuations, and mapping the power flow fluctuation feature parameters to a power grid power flow feature mapping matrix according to the association rules include the following steps: S21. Classify and preprocess the multi-source core data, remove outliers and fill in missing values ​​to obtain power generation side data, transmission side data and user side data; Specifically, the data is first divided into three categories according to its source: generation side, transmission side, and user side. Generation side data includes new energy output and conventional unit operating parameters, transmission side data covers active / reactive power flow and voltage deviation, and user side data includes industrial and residential load and electricity consumption time distribution. Then, the data is standardized in format and time-aligned to unify the data units and sampling frequency.

[0027] Outliers were removed using the 3σ principle, filtering out data that exceeded the mean ± 3 standard deviations. For missing values, short-term missing values ​​were filled using linear interpolation, while long-term missing values ​​were filled using the KNN algorithm based on historical data of similar operating conditions. Data consistency checks were performed to ensure that there were no logical conflicts among the three sides of the data, forming a clean and well-organized dataset for the generation, transmission, and user sides, laying the foundation for the subsequent extraction of power flow fluctuation characteristic parameters.

[0028] S22. Extract the power flow fluctuation characteristic parameters corresponding to the power generation side data, transmission side data and user side data. The power flow fluctuation characteristic parameters include output fluctuation rate, line load rate, load change rate, power stability coefficient, voltage deviation rate and electricity consumption period distribution coefficient. Specifically, based on the clean data from the generation, transmission, and user sides after classification and preprocessing, six types of power flow fluctuation characteristic parameters are extracted side-by-side: For the generation side, the power output fluctuation rate is obtained by calculating the output change amplitude per unit time using renewable energy output and conventional unit operation data; the power stability coefficient is obtained based on the ratio of the standard deviation and mean of the output data. For the transmission side, the line load rate is determined by the ratio of real-time active / reactive power flow data to the thermal stability limit capacity; the voltage deviation rate is extracted by the ratio of the difference between the actual voltage and the rated voltage. For the user side, the load mutation rate is obtained by calculating the ratio of the load mutation amplitude to the baseline load using real-time load monitoring data; and the power consumption period distribution coefficient is generated by statistically analyzing the power load ratio at different times. During the extraction process, the rationality of the parameters is simultaneously verified, and outliers exceeding the normal operating range of the power grid are removed to ensure that the characteristic parameters accurately reflect the power flow fluctuation characteristics of each side, laying the foundation for subsequent multi-source data fusion and mapping matrix construction.

[0029] S23. Set the weight of power generation side data to 40%, transmission side data to 30%, and user side data to 30% to form a multi-source data fusion association rule. Based on the multi-source data fusion association rule, use a weighted summation algorithm to calculate the comprehensive value of power flow fluctuations of each dimension. In this embodiment of the application, the step of setting the weight of power generation side data to 40%, transmission side data to 30%, and user side data to 30% to form a multi-source data fusion association rule, and calculating the comprehensive value of power flow fluctuation of each dimension feature based on the multi-source data fusion association rule using a weighted summation algorithm, includes the following steps: S231. The core of the multi-source data fusion and association rules is that the data weights of the generation side, transmission side, and user side are positively correlated with the degree of influence of each side on the power grid flow fluctuations, with the benchmark weight ratios being 40% for the generation side, 30% for the transmission side, and 30% for the user side. Specifically, based on the core principle that "the weight of each data source is positively correlated with its impact on power flow fluctuations," and considering the direct dominance of power generation output fluctuations on power flow, the intermediate influence of transmission line characteristics on the transmission side, and the relatively moderate impact of load changes on the user side, a baseline weight ratio of 40% for the generation side, 30% for the transmission side, and 30% for the user side is set. Initial weights are assigned to the power flow fluctuation characteristic parameters of each side (e.g., power generation output fluctuation rate of 0.25, transmission line load rate of 0.2, etc.), and fluctuation thresholds are set for each parameter characteristic. The weights are dynamically adjusted according to the actual fluctuation amplitude (weighting increases for values ​​exceeding the threshold and decreasing weights by 50% for values ​​below the threshold). Simultaneously, a summation check ensures that the total weight sum is 1 ± 0.01. This rule both anchors the objective basis for weight allocation and adapts to changes in power grid operating conditions through dynamic adjustment, laying the foundation for accurate fusion of multi-source data and the construction of a power flow characteristic mapping matrix.

[0030] S232. Configure initial weights for the power flow fluctuation characteristic parameters on each side, where the power generation side output fluctuation rate is 0.25, power stability coefficient is 0.15, the transmission side line load rate is 0.2, voltage deviation rate is 0.1, the user side load mutation rate is 0.15, and the electricity consumption period distribution coefficient is 0.15, and the sum of the initial weights of all characteristic parameters is 1. In this embodiment of the application, configuring initial weights for the power flow fluctuation characteristic parameters on each side, wherein the power generation side output fluctuation rate is 0.25 and the power stability coefficient is 0.15, the transmission side line load rate is 0.2 and the voltage deviation rate is 0.1, the user side load mutation rate is 0.15 and the electricity consumption period distribution coefficient is 0.15, and the sum of the initial weights of all characteristic parameters is 1, includes the following steps: S2321. Based on the characteristics of the power flow fluctuation parameters on each side, fluctuation thresholds are set respectively, including power generation side output fluctuation rate ±10%, power stability coefficient ±5%, transmission side line load rate ±8%, voltage deviation rate ±3%, user side load mutation rate ±12%, and electricity consumption period distribution coefficient ±6%. Specifically, based on the impact of each parameter on power grid flow stability, actual operating characteristics, and equipment capacity, differentiated threshold standards are established. For generation-side power output, significantly affected by new energy fluctuations, a power output fluctuation rate of ±10% is set to adapt to dynamic changes in wind and solar power output, and a power stability coefficient of ±5% is set to strictly control the stability of generation-side power output. For the transmission side, directly related to line safety operation, a line load rate threshold of ±8% is set to align with the line's thermal stability limit, and a voltage deviation rate of ±3% is limited to meet grid voltage quality requirements. For the user-side, load variations are diverse and highly random; a load mutation rate of ±12% is set to accommodate sudden fluctuations in industrial and residential loads, and a power consumption time distribution coefficient of ±6% is set to accommodate reasonable deviations in power load at different times. This threshold system not only conforms to the operating patterns of parameters on each side but also provides clear judgment benchmarks for subsequent dynamic weight adjustments and power flow optimization control, ensuring the safety and flexibility of grid operation.

[0031] S2322. Compare the actual fluctuation amplitude of each feature parameter with the corresponding fluctuation threshold side by side. If the actual fluctuation amplitude exceeds the threshold, increase the weight of the feature parameter by a step size of 0.05. If the actual fluctuation amplitude is less than 50% of the threshold, decrease the weight of the feature parameter by a step size of 0.03. Specifically, based on preset thresholds for power flow fluctuation characteristic parameters on each side (e.g., power generation fluctuation rate ±10%, transmission line load rate ±8%, user load mutation rate ±12%), a precise comparison is made between the actual fluctuation amplitude and the corresponding threshold on each side. If the actual fluctuation amplitude of a certain characteristic parameter exceeds the preset threshold, the weight of that parameter is increased by a step of 0.05 to strengthen its impact on power flow optimization; if the actual fluctuation amplitude is less than 50% of the threshold, the corresponding weight is decreased by a step of 0.03 to reduce interference from non-critical parameters.

[0032] After the weights are adjusted, a total verification is required to ensure that the total weights of all feature parameters are 1 ± 0.01. If this is not met, the weights of the user-side parameters are fine-tuned first (adjustment step size 0.01). The dynamic adjustment logic not only fits the operating characteristics of each side parameter, but also adapts to the real-time changes in the power grid operating conditions, thereby improving the pertinence and accuracy of multi-source data fusion.

[0033] S2323. Perform a total check on all adjusted feature parameter weights to ensure that the total weight is 1 ± 0.01. If the condition is not met, prioritize adjusting the user-side feature parameter weights. Specifically, the adjusted weights of all characteristic parameters from the generation, transmission, and user sides are first aggregated, and the sum is calculated and compared with the standard range of 1 ± 0.01. If the sum exceeds this range, the weights of user-side characteristic parameters (load mutation rate, electricity consumption period distribution coefficient) are fine-tuned first, with an adjustment step size of 0.01, to avoid significant changes affecting the rationality of weight allocation on other sides. After fine-tuning, the weight sum is recalculated, and the process is iterated until the requirement of 1 ± 0.01 is met. This verification mechanism ensures the logical consistency of weight allocation in the multi-source data fusion association rules, guarantees the accuracy of subsequent weighted summation algorithms, and provides a reliable weight foundation for the construction of the power flow feature mapping matrix.

[0034] S2324. Store the adjusted weights in the weight parameter library and use them as the initial weight reference for subsequent similar working conditions.

[0035] Specifically, the adjusted weights, after being verified to be within 1 ± 0.01, are bound to corresponding operating condition feature tags (including real-time operating conditions such as generation / transmission / user side fluctuation parameters, grid voltage / frequency / load rate, and peak / valley time information) to form a "weight data - operating condition feature" associated data group. These are categorized and stored in a weight parameter library according to operating condition type (e.g., high load / high penetration of new energy / stable operation), time period, etc., with an index directory for quick retrieval. A data update mechanism is also implemented to overwrite older data or supplement records with the latest weight data adapted to the operating condition. When encountering the same / similar grid operating conditions subsequently, the corresponding weights are retrieved from the parameter library through operating condition feature matching and directly used as initial weight references. This eliminates the need to repeatedly execute the "threshold comparison - weight adjustment" process, reducing computation time, ensuring weight adaptability, and improving the efficiency of multi-source data fusion and power flow optimization.

[0036] S233. The weighted summation algorithm is used to calculate the comprehensive value of power flow fluctuations, where the formula for the weighted summation algorithm is: F=0.4×(F1×W1+F2×W2)+0.3×(F3×W3+F4×W4)+0.3×(F5×W5+F6×W6); Where F is the comprehensive value of tidal fluctuation, F1-F6 are the specific values ​​of each characteristic parameter, and W1-W6 are the initial weights of the corresponding characteristic parameters; S234. Verify the validity of the power flow fluctuation composite value, remove abnormal composite values ​​that exceed the reasonable range of normal power grid operation scenarios, and retain the valid composite values ​​for constructing the initial feature matrix.

[0037] Specifically, based on historical data and industry standards for normal power grid operation scenarios, the reasonable range for the composite value is defined as 0-0.8, which is suitable for the normal fluctuation characteristics of data from the generation, transmission, and user sides. The composite values ​​obtained through a weighted summation algorithm are individually validated for their range, directly eliminating outliers outside the 0-0.8 range to avoid extreme data interfering with matrix construction. Simultaneously, multi-source data fusion and association rules are used to assist in verifying the consistency between the composite value and the corresponding power grid operating conditions. If the composite value logically conflicts with real-time operating parameters (e.g., high fluctuation composite values ​​appearing during low-load periods), it is also judged as an anomaly and eliminated. Finally, all valid composite values ​​that meet the range and are logically consistent are collected, and the correspondence between power flow fluctuation characteristic parameter types and power grid operating time nodes is established to provide clean and reliable data support for the initial feature matrix construction.

[0038] S24. Construct an initial feature matrix with power flow fluctuation characteristic parameter types as rows and power grid operation time nodes as columns. Combine the multi-source data fusion association rules to perform weighted correction on each element in the initial feature matrix. After dimension verification and abnormal element removal, the power grid power flow characteristic mapping matrix is ​​obtained.

[0039] In this embodiment of the application, the process of constructing an initial feature matrix with power flow fluctuation characteristic parameter types as rows and power grid operation time nodes as columns, weighting and correcting each element in the initial feature matrix using multi-source data fusion association rules, and obtaining the power grid power flow feature mapping matrix after dimension verification and outlier removal includes the following steps: S241. Using six types of power flow fluctuation characteristic parameters as row vectors and 24-hour continuous operation time nodes of the power grid as column vectors, fill the effective power flow fluctuation comprehensive value according to the correspondence between characteristic parameters and time nodes to construct an initial feature matrix. Specifically, using six types of power flow fluctuation characteristic parameters—output fluctuation rate, line load rate, load mutation rate, power stability coefficient, voltage deviation rate, and electricity consumption period distribution coefficient—as the row vectors of a matrix, and each time point of the grid's continuous 24-hour operation as the column vectors, the one-to-one correspondence between "characteristic parameters and time points" is clearly defined. The comprehensive power flow fluctuation values, after validity verification (excluding those exceeding the reasonable range of 0-0.8), are accurately filled into the corresponding positions in the matrix according to their respective characteristic parameter types and collection time points, constructing an initial characteristic matrix with dimensions of 6 rows × 24 columns. During the filling process, data consistency is simultaneously checked to ensure that the comprehensive values ​​of each characteristic parameter at the same time point all come from the same grid operation scenario, and that the data for the same characteristic parameter across the 24 time points are continuous and complete. This lays a well-organized data foundation for subsequent matrix weighting correction and the construction of the grid power flow characteristic mapping matrix.

[0040] S242. Determine the correction weights according to the multi-source data fusion association rules, where the correction weights of the matrix elements corresponding to the characteristic parameters on the power generation side are 1.2, the elements corresponding to the transmission side are 1.0, and the elements corresponding to the user side are 0.8, and the correction weights are consistent with the degree of influence of the data on the power flow of each side. Specifically, the weights are allocated based on the principle that "the correction weights are strongly correlated with the degree of influence of each side's data on power flow fluctuations," taking into account the differences in the actual effects of the data on each side. Power generation fluctuations directly dominate power flow changes and have the strongest impact on power flow stability, so the correction weight for the corresponding matrix element is set to 1.2. The transmission side, responsible for power transmission, has a moderate impact, so the correction weight is set to 1.0. User-side load fluctuations are relatively mild, with a weak indirect impact on power flow, so the correction weight is set to 0.8. During weight allocation, the corresponding correction weights are matched one by one according to the generation, transmission, and user-side categories to which each element in the initial feature matrix belongs. This ensures that the correction strength of the matrix elements for each type of feature parameter matches its actual impact, aligning with the core of multi-source data fusion and association rules, and enhancing the accuracy of power flow feature mapping through differentiated weights, providing a reliable basis for subsequent matrix optimization.

[0041] S243. By multiplying the original value of each element by the corresponding correction weight, all elements in the initial feature matrix are weighted and corrected point by point to obtain the intermediate feature matrix. Specifically, based on the multi-source data fusion and association rules, the weights of elements are first adjusted according to their respective sides (1.2 for generation-side feature parameters, 1.0 for transmission-side, and 0.8 for user-side) to ensure that the adjustment level is consistent with the impact of each side's data on power flow. Then, all elements in the initial feature matrix (6 rows × 24 columns) are traversed, and the calculation of "original value × corresponding adjustment weight" is performed on each element to precisely strengthen the power flow impact weight of key side data. After the calculation is completed, all adjusted elements are integrated according to their corresponding "feature parameter - time node" positions in the original matrix to form a structurally regular intermediate feature matrix. This provides precisely adjusted basic data for subsequent dimensional verification and outlier element removal, improving the reliability of the power flow feature mapping matrix construction.

[0042] S244. Perform dimension verification and outlier removal on the intermediate feature matrix to obtain the power grid flow feature mapping matrix.

[0043] Specifically, the intermediate feature matrix first undergoes dimensionality verification. The core verification checks whether the matrix dimension is 6 rows × 24 columns, ensuring complete correspondence with the six types of power flow fluctuation characteristic parameters and the 24-hour time nodes. Simultaneously, the matrix dimension is verified to be compatible with the current number of operating nodes in the power grid. If the dimension is inconsistent, missing columns are added or redundant columns are removed to ensure the matrix structure is compliant. Next, outlier element removal is performed. A reasonable threshold of ≤1.0 for matrix elements under normal power grid operation is set. Element values ​​are screened one by one, and outliers exceeding the threshold are removed. For any gaps after removal, neighborhood interpolation is used to fill in the gaps with the average of valid elements with the same characteristic parameters at adjacent time nodes, ensuring data continuity. After dimensionality verification and outlier removal, the matrix data undergoes consistency verification to ensure no logical conflicts. Finally, an accurate, complete, and adaptable power flow characteristic mapping matrix is ​​formed, providing reliable data support for subsequent power flow optimization modeling.

[0044] S3. Based on the real-time operating parameters of the power grid, set multi-dimensional dynamic weighting rules, use the BP neural network algorithm to construct an optimization model, and take minimizing power grid power loss and reducing the probability of power flow exceeding limits as the dual optimization objectives, and iteratively optimize and adjust the power flow characteristic mapping matrix. In this embodiment of the application, the step of setting multi-dimensional dynamic weighting rules based on real-time power grid operating parameters, constructing an optimization model using a BP neural network algorithm, and iteratively optimizing and adjusting the power grid power flow feature mapping matrix with the dual optimization objectives of minimizing power grid power loss and reducing the probability of power flow exceeding limits includes the following steps: S31. Normalize the real-time operating parameters of the power grid, remove abnormal parameters that exceed the reasonable operating range, and obtain a standardized set of operating parameters. With voltage, frequency, and load rate as the core dimensions, the weights are dynamically adjusted according to peak and valley periods. Set multi-dimensional dynamic weighting rules. Specifically, real-time power grid operating parameters are first collected, covering voltage, frequency, load factor, and auxiliary operating parameters. Normalization is then performed, using min-max standardization to map all parameters to the [0, 1] interval, eliminating dimensional differences and ensuring parameter comparability. Next, reasonable operating ranges for each parameter are set, and abnormal data such as voltage exceeding limits, frequency exceeding limits, and load factor overload are removed. After verification, a clean, standardized set of operating parameters is obtained. Voltage, frequency, and load factor are selected as the core dimensions for regulation. Based on the peak-valley characteristics of the power grid, multi-dimensional dynamic weighting rules are set. During peak load periods, the weight of load factor is increased; during flat periods, the weights of the three are balanced; and during off-peak periods, voltage stability is emphasized and its weight is increased. This dynamic weight allocation always aligns with the real-time operating needs of the power grid, providing standardized and adaptable operating data support for subsequent BP neural network optimization modeling, ensuring that the optimization direction accurately matches the actual operating state of the power grid.

[0045] S32. A BP neural network optimization model is constructed using the BP neural network algorithm. The BP neural network optimization model is a three-layer structure of input layer, hidden layer, and output layer. The node data of the input layer are the elements of the power flow feature mapping matrix, and the node data of the output layer are the predicted values ​​of power grid power loss and power flow over-limit probability. The number of nodes in the hidden layer is determined by empirical formula based on the number of nodes in the input layer and the output layer. In this embodiment of the application, the step of constructing a BP neural network optimization model using the BP neural network algorithm, wherein the BP neural network optimization model is a three-layer structure of input layer-hidden layer-output layer, the input layer node data is the power flow feature mapping matrix element, the output layer node data is the power grid power loss prediction value and the power flow over-limit probability prediction value, and the number of hidden layer nodes is determined based on the number of input layer and output layer nodes through an empirical formula, including the following steps: S321. The number of input layer nodes is specified as the total number of elements in the power flow feature mapping matrix, and the number of output layer nodes is 2, which correspond to the predicted power loss of the power grid and the predicted power flow over-limit probability, respectively. Specifically, the logic for setting the number of nodes in the network layers is first clarified. The number of input layer nodes corresponds to the total number of elements in the power flow feature mapping matrix, which has a 6-row × 24-column structure. Therefore, the number of input layer nodes is determined to be 144, ensuring that all power flow feature data is completely input, providing comprehensive basic support for model training. The output layer is set with two nodes according to dual optimization objectives. One node specifically corresponds to the predicted value of power loss in the power grid, accurately outputting the model's prediction result of the power loss in the power grid operation. The other node corresponds to the predicted value of the probability of power flow exceeding the limit, intuitively reflecting the probability of the power flow exceeding the limit risk. The dual output nodes not only meet the core optimization requirements of minimizing power loss and reducing the probability of power flow exceeding the limit, but also make the model output objectives clear and explicit, providing precise guidance for subsequent iterative optimization and result judgment, ensuring that the construction of the BP neural network optimization model meets the actual needs of power flow regulation.

[0046] S322. The number of hidden layer nodes is determined using an empirical formula, where the empirical formula is: ; Where n is the number of hidden layer nodes, m is the number of input layer nodes (144), k is the number of output layer nodes (2), and a is an adjustment constant between 10 and 20. The calculated number of hidden layer nodes is 43. S323. Configure the core auxiliary parameters of the BP neural network optimization model, including an initial learning rate of 0.03, a maximum number of iterations of 500, and a convergence condition of loss function value < 0.001. Specifically, key parameters were precisely set by combining the dual objectives of power grid power flow optimization (minimizing power loss and reducing the probability of power flow exceeding limits) with the training characteristics of 144-dimensional input data. The initial learning rate was configured to 0.03, balancing model convergence speed and training stability, avoiding parameter oscillations caused by an excessively high learning rate and excessively long iteration cycles due to an excessively low learning rate. The maximum number of iterations was set to 500, adapting to the dimensionality of the power flow feature mapping matrix and balancing training efficiency and optimization depth. The convergence condition was explicitly defined as a loss function value < 0.001, serving as the core criterion for terminating model training. After parameter configuration, small-sample pre-training was conducted for verification. If slow convergence or oscillations occurred during pre-training, the learning rate was fine-tuned in steps of 0.01 to ensure that the convergence condition was met within 500 iterations. The final parameter combination not only conforms to the training patterns of power grid data but also ensures that the model converges quickly to the optimal state, providing stable and reliable model parameter support for the iterative optimization of the power flow feature matrix.

[0047] S324. Set the initial range of connection weights between the input layer and the hidden layer, and between the hidden layer and the output layer to [-0.5, 0.5], and use the Sigmoid function as the activation function.

[0048] Specifically, when configuring the connection weights and activation function of the BP neural network, the initial range of the inter-layer connection weights is first set. The connection weights between the input layer and the hidden layer, and between the hidden layer and the output layer, are uniformly set to the range of [-0.5, 0.5]. This range is moderate, avoiding neuron saturation due to excessively large initial weights and gradient vanishing due to excessively small initial weights, thus ensuring the stability of the model's initial training and laying the foundation for subsequent iterative optimization. The Sigmoid function is selected as the activation function. This function can realize the nonlinear mapping of the input signal, adapting to the complex nonlinear relationship of power grid power flow optimization. At the same time, it can output results in the 0-1 range, which is highly consistent with the prediction scenarios of power loss and power flow exceedance probability. After configuration, simultaneous verification is performed to ensure that the weights are randomly and uniformly distributed within the set range, and that the gradient of the activation function is within the effective range, avoiding gradient vanishing from affecting convergence. This allows the model to efficiently process power flow feature mapping matrix data, accurately output dual-objective prediction values, and improve the nonlinear fitting ability and prediction accuracy of the optimized model.

[0049] S33. Preset iterative optimization judgment rules. Iteratively optimize the power flow feature mapping matrix according to the iterative optimization judgment rules. When the iterative optimization judgment rules are met, the optimization stops and the matrix is ​​output. When the iterative optimization judgment rules are not met, the weight ratio of each dimension in the dynamic weighting rules is adjusted and the iteration is restarted.

[0050] Specifically, iterative optimization judgment rules are preset, with the core revolving around the dual optimization objectives and convergence requirements. The rules are clearly defined as follows: the predicted power loss value output by the model drops to a preset reasonable threshold, the predicted power flow exceedance probability is lower than the limit value, and the loss function value is less than 0.001 for 10 consecutive iterations. Simultaneously, the change rate of the power flow feature mapping matrix elements between two iterations is less than 0.02. During execution, the matrix is ​​input into the BP neural network model for iterative optimization. After each iteration, the judgment rules are checked. If they are met, optimization stops and the optimal matrix is ​​output; otherwise, the multi-dimensional dynamic weighting rules are adjusted, prioritizing fine-tuning of the weight ratios of the core dimensions of voltage, frequency, and load factor, with adjustments made to suit peak and valley periods. After adjustment, the matrix is ​​re-input into the model for iteration. The rationality of parameters is monitored synchronously during iteration to avoid weight imbalance. This process is repeated until the criteria are met, thereby ensuring that the optimized matrix meets the dual objective requirements and improves the accuracy of power flow control and the stability of grid operation.

[0051] S4. Preset dynamic power flow scheme matching rules. Based on the dynamic power flow scheme matching rules, perform feature similarity matching between the optimized and adjusted power flow feature mapping matrix and the smart grid dynamic power flow optimization scheme library, and filter and sort to obtain the smart grid dynamic power flow optimization scheme. In this embodiment of the application, the preset dynamic power flow scheme matching rule, which matches the optimized and adjusted power flow feature mapping matrix with the smart grid dynamic power flow optimization scheme library based on the dynamic power flow scheme matching rule, and filters and sorts the smart grid dynamic power flow optimization schemes, includes the following steps: S41. The preset dynamic power flow scheme matching rule is the cosine similarity algorithm, and the preset candidate scheme priority ranking rule is similarity priority, power loss reduction priority, and debugging cost priority. Specifically, the core feature vectors of the power flow feature mapping matrix and the feature vectors of historical schemes in the dynamic power flow optimization scheme library are extracted first. The cosine similarity between the two is calculated, and the closer the value is to 1, the higher the matching degree. This accurately selects candidate schemes suitable for the current operating conditions. At the same time, a priority ranking rule for candidate schemes is preset, which prioritizes similarity, power loss reduction, and commissioning cost. First, qualified schemes are selected based on the cosine similarity, then the power loss reduction of the schemes is compared to select the best one, and finally the historical commissioning cost is calculated to determine the optimal scheme, taking into account adaptability, economy, and feasibility. This rule ensures that the scheme is accurately adapted to the real-time operating conditions, maximizes the reduction of losses and saves costs, provides reliable support for subsequent safety verification and closed-loop commissioning, and improves the efficiency and practicality of power flow optimization.

[0052] S42. Standardize the optimized power flow feature mapping matrix to ensure that the dimension of the power flow feature mapping matrix is ​​consistent with the feature matrix of historical optimization schemes in the smart grid dynamic power flow optimization scheme library. Specifically, the optimized power flow feature mapping matrix is ​​first standardized. The core objective is to ensure that its dimensions are completely consistent with the historical optimized feature matrices in the scheme library, guaranteeing matching effectiveness. First, the dimensions of the current matrix are verified, and the unified standard dimensions of the historical feature matrices in the scheme library are clarified. Then, the optimized matrix undergoes dimensional normalization. If the dimensions are insufficient, interpolated data for the corresponding time period and feature parameters are added; if the dimensions are redundant, invalid rows and columns are removed to ensure accurate matching of the number of rows and columns. Subsequently, min-max normalization is used to map all elements of the matrix to the [0, 1] interval, eliminating the influence of dimensional differences and numerical fluctuations, while aligning data precision and format standards to maintain consistency with historical matrices. After processing, consistency is simultaneously verified to confirm the absence of dimensional deviations and numerical standardization, making the standardized power flow feature mapping matrix comparable to historical scheme feature matrices. This lays a unified data foundation for subsequent cosine similarity matching, improving the accuracy of scheme matching.

[0053] S43. Calculate the feature similarity between the optimized and adjusted power flow feature mapping matrix and the power grid dynamic power flow optimization schemes in the smart grid dynamic power flow optimization scheme library based on the cosine similarity algorithm, and preset the feature similarity screening threshold. Based on the feature similarity screening threshold, filter the feature similarity to obtain the candidate optimization scheme set. In this embodiment of the application, the step of calculating the optimized power flow feature mapping matrix based on the cosine similarity algorithm and the feature similarity of the power grid dynamic power flow optimization schemes in the smart grid dynamic power flow optimization scheme library, and preset a feature similarity screening threshold, and filtering feature similarity based on the feature similarity screening threshold to obtain a candidate optimization scheme set includes the following steps: S431. The formula for the cosine similarity algorithm is as follows: ; in, To measure the similarity between the optimized power flow feature mapping matrix and the feature matrix of historical optimized schemes in the scheme library, A is the optimized power flow feature mapping matrix, and B is the feature matrix of historical optimized schemes in the scheme library. S432. The preset feature similarity screening threshold is 0.85. When the calculated feature similarity is greater than or equal to 0.85, the corresponding historical optimization scheme is included in the candidate optimization scheme set. Specifically, a feature similarity screening threshold of 0.85 is initially preset. This threshold is determined by combining historical adaptation cases with the matching requirements of power grid operating conditions, balancing the adaptability of the solution with screening efficiency. After standardization, the power flow feature mapping matrix is ​​aligned with the feature matrix of historical optimization solutions in the smart grid dynamic power flow optimization solution library, and the feature similarity value is calculated one by one using the cosine similarity algorithm. The calculated result is compared with the 0.85 threshold one by one. When the feature similarity is greater than or equal to 0.85, the historical solution is determined to be highly adapted to the current power grid operating conditions and is directly included in the candidate optimization solution set; if it is lower than 0.85, it indicates insufficient adaptability and is eliminated to avoid invalid solutions increasing screening costs. During the screening process, the similarity value and core adaptability points of the qualified solutions are recorded simultaneously, and the corresponding historical operating conditions and optimization results are marked. This provides a clear basis for subsequent priority ranking, ensuring that the candidate solution set has both adaptability and practicality, and laying a solid foundation for accurately selecting the optimal power flow optimization solution.

[0054] S433. If the candidate optimization scheme set is empty, the screening threshold is gradually reduced in increments of 0.05 until at least 3 candidate optimization schemes are obtained. Specifically, a candidate solution set supplementation mechanism needs to be preset. After completing the feature consistency verification, the candidate optimization solution set is first checked to see if it is empty. If it is empty, the threshold reduction process is initiated, and the feature similarity screening threshold is gradually reduced in a fixed step size of 0.05. After each reduction, historical optimization solutions in the solution library are retrieved again, and the cosine similarity algorithm is reused to calculate the matching degree. Feature consistency verification is carried out simultaneously until at least 3 valid candidate solutions are selected. During the reduction process, a threshold baseline of no less than 0.6 is set to avoid excessive threshold reduction that would lead to a significant decrease in solution adaptability. At the same time, after each round of reduction, solutions with relatively higher similarity are prioritized for retention, balancing quantity requirements and basic adaptability. If the threshold is still not met after being lowered to the baseline, historical solutions under similar working conditions are retrieved and the consistency verification standard is fine-tuned to ensure a sufficient number of candidate solutions. This mechanism ensures that the solution selection has sufficient sample support and strictly controls the adaptability baseline, laying a solid foundation for subsequent priority ranking and selection of the optimal solution, and avoiding the stagnation of the power flow optimization process due to the lack of candidate solutions.

[0055] S434. Perform feature consistency verification on the schemes in the candidate optimization scheme set, eliminate schemes with a compatibility of less than 80% with the current power grid operating parameters, and retain valid candidate schemes.

[0056] Specifically, the core parameters of historical operating conditions for each candidate scheme are first extracted and compared with the core dimensions of the current real-time power grid operating conditions (voltage, frequency, load factor). The fit is calculated according to preset weights, and the key optimization directions of the schemes are simultaneously checked to ensure consistency with current power flow control requirements. Each candidate scheme is comprehensively verified. If the fit is below 80%, it indicates insufficient matching with the current operating conditions, which may lead to optimization deviations, and the scheme is directly eliminated. Schemes meeting the fit standard are retained as valid candidates. The verification process considers both feature similarity and operating condition fit, avoiding the limitations of single-dimensional selection. The fit details and core advantages of valid schemes are recorded to support subsequent priority ranking, ensuring that retained schemes not only fit the current power grid operating status but also possess practical optimization effects, thus improving the reliability of the final power flow scheme implementation.

[0057] S44. Sort the candidate optimization scheme set according to the candidate scheme priority ranking rule, and select the candidate optimization scheme with the highest ranking as the smart grid dynamic power flow optimization scheme.

[0058] Specifically, the primary ranking criterion is feature similarity, prioritizing schemes with higher similarity to those better suited to the current operating conditions. Next, for schemes with similar similarity, the reduction in grid power loss is compared, prioritizing those with greater loss reduction to align with the grid's energy-saving operation requirements. Finally, for schemes with comparable performance in the first two categories, historical commissioning costs are calculated, prioritizing the lower-cost scheme to ensure economic viability during implementation. Throughout the ranking process, the consistency of features between the schemes and the current operating conditions is verified to ensure the ranking remains aligned with the core adaptation criteria. After completing the ranking across all levels, the highest-ranked candidate optimization scheme is selected as the final smart grid dynamic power flow optimization scheme. This approach ensures accurate scheme adaptation while minimizing losses and optimizing costs, thereby enhancing the practicality and efficiency of power flow control.

[0059] S5. Combine the smart grid dynamic power flow optimization scheme with the deviation compensation logic of historical power flow control parameters, perform phased debugging of smart grid power flow control equipment, and collect the power flow change value of the grid after debugging in real time. In this embodiment of the application, the step of combining the smart grid dynamic power flow optimization scheme with the deviation compensation logic of historical power flow control parameters to perform phased debugging of the smart grid power flow control equipment and to collect the power flow change values ​​of the grid after debugging in real time includes the following steps: S51. Retrieve and analyze the equipment debugging deviation records, control response delay data and adaptation effects of previous optimization schemes in the historical power flow control parameters, and set the equipment debugging benchmark parameter range in combination with the smart grid dynamic power flow optimization scheme. Specifically, the process begins with a comprehensive retrieval of all historical power flow control data from the power grid. The focus is on extracting three core types of information: equipment commissioning deviation records, control response delay data, and the effectiveness of previous optimization schemes. This information is then categorized and analyzed to identify common equipment commissioning deviation ranges, reasonable response delay thresholds, and key factors influencing the effectiveness of scheme adaptation. Subsequently, based on the core requirements of the selected smart grid dynamic power flow optimization scheme and using historical data, a targeted range of equipment commissioning benchmark parameters is set. This aims to avoid deviations exceeding limits encountered in previous commissioning while matching the scheme's requirements for control response speed. The benchmark range boundaries are also calibrated by referencing historical parameter ranges with good adaptation results. After setting, synchronous verification ensures that the benchmark parameter range not only complies with equipment safety operation specifications but also closely aligns with the dynamic power flow optimization objectives. This provides a clear standard for subsequent on-site commissioning, reduces commissioning deviations and response delays, ensures the accurate implementation of the optimization scheme, and improves the stability and adaptability of power flow control.

[0060] S52. Based on the equipment debugging benchmark parameter range, the power flow control equipment is debugged in three stages: pre-debugging, fine-tuning, and stability verification, until the equipment operating status matches the smart grid dynamic power flow optimization scheme. Specifically, based on preset equipment debugging benchmark parameter ranges, power flow control equipment is debugged in three phases until the equipment's operating status precisely matches the smart grid's dynamic power flow optimization scheme. The first phase involves pre-debugging, setting initial operating parameters according to the benchmark parameter range, completing basic debugging such as wiring and start-up / shutdown, troubleshooting equipment faults, ensuring the equipment can respond normally to control commands, and eliminating invalid operating states. The second phase involves refined debugging, fine-tuning and calibrating core control parameters one by one against the upper and lower limits of the benchmark parameters, correcting deviations according to the requirements of the power flow optimization scheme, reducing parameter fluctuation ranges, and ensuring the equipment's operating accuracy meets the scheme requirements. The third phase involves stability verification, continuously monitoring equipment operating data, verifying parameter stability and response timeliness, and confirming no deviation exceeding limits or response delays. Throughout the process, the equipment status and the suitability of the optimization scheme are compared synchronously. If the standards are not met, refined debugging and verification are repeated; once the standards are met, debugging is terminated, ensuring the equipment stably supports the implementation of the power flow optimization scheme and improves the reliability of grid operation.

[0061] S53. After the equipment is running stably, continuously collect power grid flow change values ​​and store them according to time nodes.

[0062] Specifically, once the power flow control equipment completes its three-stage commissioning and achieves stable operation, a continuous power flow data acquisition mechanism is immediately activated. This mechanism accurately collects key data such as power loss, power flow limit-related parameters, voltage and frequency fluctuations, and power flow changes at core nodes. It ensures that the acquisition frequency aligns with the 24-hour timeframes of the previously constructed feature matrix, guaranteeing data continuity. Real-time verification is performed concurrently during acquisition to eliminate abnormal and distorted data and prevent the retention of invalid information. Data is then strictly categorized and stored according to preset timeframes, with full power flow change data for each hourly period. Simultaneously, the data is linked to the daily power grid operating condition label, equipment operating parameters, and optimization scheme execution status, establishing a structured data ledger. A dedicated database with hierarchical indexes is built for storage, facilitating rapid retrieval and analysis. This provides complete and accurate time-series data support for subsequent scheme adaptation effect review, dynamic weighted rule optimization, and iterative optimization, solidifying the data foundation for closed-loop power flow control.

[0063] S6. Preset a power flow change safety threshold, perform safety verification on the power flow change value of the power grid based on the power flow change safety threshold, and dynamically adjust the smart grid dynamic power flow optimization scheme to form a closed loop optimization according to the safety verification results.

[0064] Specifically, by combining power grid safety operation standards with historical stable operating data, a preset safety threshold for power flow changes is established. This defines the safe ranges for core indicators such as power loss, power flow exceedance probability, and voltage / frequency fluctuations, and delineates threshold red lines to ensure no breaches of power grid operation specifications. After the equipment is operating stably, continuously collected power flow change values ​​are extracted, and safety verification is performed item by item against the preset safety thresholds to determine if each indicator is within the safe range. Deviation data and exceeding nodes are recorded simultaneously. If verification passes, the current smart grid dynamic power flow optimization scheme is maintained. If indicators exceed the limits or approach the threshold, a dynamic adjustment mechanism is immediately activated to specifically optimize the weighting rules, equipment parameters, or control strategies in the scheme. After adjustment, the scheme is re-implemented and data is collected for verification. This iterative process forms a closed-loop optimization system of "collection-verification-adjustment-implementation," continuously adapting to real-time changes in power grid operating conditions, ensuring that power flow remains stable within the safe range, and improving the accuracy and reliability of dynamic control.

[0065] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for adaptive closed-loop power flow optimization control of a smart grid, characterized in that, The method comprises the following steps: S1, acquiring multi-source core data of a power generation side, a power transmission side and a user side, real-time working condition parameters of a power grid and historical power flow control parameters, and establishing a smart grid dynamic power flow optimization scheme library, wherein the historical optimization schemes in the smart grid dynamic power flow optimization scheme library are associated with feature matrices one by one; S2, extracting power flow fluctuation feature parameters in the multi-source core data, presetting multi-source data fusion association rules, and mapping the power flow fluctuation feature parameters into a power grid power flow feature mapping matrix according to the multi-source data fusion association rules; S3, setting multi-dimensional dynamic weighting rules based on the real-time working condition parameters of the power grid, constructing an optimization model by using a BP neural network algorithm, and iteratively optimizing and adjusting the power grid power flow feature mapping matrix to obtain an optimized power flow feature mapping matrix, with the minimum power loss of the power grid and the lowest probability of power flow exceeding limit as double optimization objectives; S4, presetting dynamic power flow scheme matching rules, performing feature similarity matching between the optimized power flow feature mapping matrix and the smart grid dynamic power flow optimization scheme library based on the dynamic power flow scheme matching rules, and screening and sorting to obtain a smart grid dynamic power flow optimization scheme; S5, combining the smart grid dynamic power flow optimization scheme with the deviation compensation logic of the historical power flow control parameters, performing staged debugging on a smart grid power flow control device, and collecting a power grid power flow change value after debugging in real time; S6, presetting a power flow change safety threshold, performing safety verification on the power grid power flow change value based on the power flow change safety threshold, and dynamically adjusting the smart grid dynamic power flow optimization scheme according to the safety verification result to form a closed-loop optimization.

2. The smart grid adaptive closed loop power flow optimization control method of claim 1, wherein, In the step S1, the multi-source core data is acquired as follows: the output fluctuation rate and the power stability coefficient are collected from the power generation side through a new energy monitoring system and a unit monitoring platform, the line load rate and the voltage deviation rate are collected from the power transmission side through a SCADA system and a WAMS system, and the load mutation rate and the electricity time period distribution coefficient are collected from the user side through a smart meter and a load monitoring terminal.

3. The smart grid adaptive closed loop power flow optimization control method of claim 1, wherein, In the step S2, the power flow fluctuation feature parameters include the output fluctuation rate, the power stability coefficient, the line load rate, the voltage deviation rate, the load mutation rate and the electricity time period distribution coefficient.

4. The smart grid adaptive closed loop power flow optimization control method of claim 1, wherein, In the step S2, the preset multi-source data fusion association rules include: setting the data weight of the power generation side as 40%, the data weight of the power transmission side as 30% and the data weight of the user side as 30%, and configuring initial weights for the power flow fluctuation feature parameters of each side, wherein the weight of the output fluctuation rate of the power generation side is 0.25, the weight of the power stability coefficient is 0.15, the weight of the line load rate of the power transmission side is 0.2, the weight of the voltage deviation rate is 0.1, the weight of the load mutation rate of the user side is 0.15, the weight of the electricity time period distribution coefficient is 0.15, and the sum of the initial weights of all the feature parameters is 1.

5. The smart grid adaptive closed loop power flow optimization control method of claim 1, wherein, In the step S3, the BP neural network algorithm is used to construct the optimization model, including: constructing a three-layer structure of input layer-hidden layer-output layer, wherein the number of input layer nodes is the total number of elements of the power grid power flow feature mapping matrix, the number of output layer nodes is 2 and corresponds to the power loss prediction value and the power flow out-of-limit probability prediction value respectively, and the number of hidden layer nodes is determined based on the number of input layer nodes and the number of output layer nodes through an empirical formula.

6. The smart grid adaptive closed loop power flow optimization control method of claim 5, wherein, The empirical formula is: ； wherein n is the number of hidden layer nodes, m is the number of input layer nodes, k is the number of output layer nodes, and a is an adjustment constant between 10 and 20.

7. The smart grid adaptive closed loop power flow optimization control method of claim 1, wherein, In the step S4, the dynamic power flow scheme matching rule is a cosine similarity algorithm, and the formula of the cosine similarity algorithm is: ， wherein A is the optimized power flow feature mapping matrix, and B is the feature matrix of the historical optimization scheme in the scheme library.

8. The smart grid adaptive closed loop power flow optimization control method of claim 7, wherein, The preset feature similarity screening threshold is 0.85, and when the calculated feature similarity is greater than or equal to 0.85, the corresponding historical optimization scheme is included in the candidate optimization scheme set.

9. The smart grid adaptive closed loop power flow optimization control method of claim 1, wherein, In the step S5, the smart grid power flow control device is debugged in stages, including: a pre-debugging stage, a fine-tuning debugging stage, and a stable verification stage.

10. The smart grid adaptive closed loop power flow optimization control method of claim 1, wherein, In the step S6, the preset power flow change safety threshold is set based on the power grid safe operation standard and the historical stable working condition data, and the power flow change safety threshold includes a power loss safety interval, a power flow out-of-limit probability safety interval, and a voltage frequency fluctuation safety interval.

Images