Adaptive load regulation and control optimization method and system for power supply and distribution operation

Abstract

The invention discloses a self-adaptive load regulation and control optimization method and system for power supply and distribution operation, and relates to the technical field of power supply and distribution operation and load regulation and control optimization of a power system, and the method comprises the steps: receiving original operation records from measurement terminals at different positions, and carrying out the preprocessing of the original operation records, and outputting a standardized operation data table; selecting a continuous time period according to a time window required by future prediction, and constructing a multi-dimensional input fragment for prediction and writing into a future data area; constructing an optimization target model according to the sorted power flow feasible region, outputting load adjustment quantity and energy storage charge and discharge quantity of each time step, and writing the load adjustment quantity and the energy storage charge and discharge quantity into a future regulation and control region; and the load adjustment amount and the energy storage power record which are optimally output are read from a future regulation and control area according to the time step sequence, are converted into specific equipment action parameters according to the logic measurement point serial number, are combined into an execution sequence arranged according to the future time sequence, and are sent to an execution end. According to the method, the perspectiveness, refinement and performability of power supply and distribution operation are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of power system power supply and distribution operation and load regulation optimization technology, specifically to an adaptive load regulation optimization method and system for power supply and distribution operation. Background Technology

[0002] With the development of distributed power sources, large-scale electric vehicle integration, and the coexistence of various load types, the operating characteristics of power supply and distribution systems have gradually evolved from the traditional single-phase, relatively stable load pattern to a multi-source, multi-time-varying coupled pattern. To address issues such as increasing load peak-to-valley differences, voltage fluctuations, and repeated reconfiguration of line power flows, academia and engineering have successively proposed technical solutions based on distribution automation, including load management, demand-side response, and energy storage coordinated control. These solutions utilize smart meters, feeder monitoring devices, and distributed measurement devices to collect operational data, combined with short-term load forecasting, optimal power flow calculation, and scheduling optimization, to achieve coordinated control of the distribution network's economic operation and safety constraints. Some studies have begun to introduce data-driven models to predict and cluster load curves, providing a reference for subsequent regulation; however, overall, the focus remains on segmented functional design and offline strategy configuration.

[0003] Existing load control technologies for power supply and distribution operations generally suffer from fragmented data processing chains and difficulties in forming a closed loop between prediction results and optimization models and actual execution. On the one hand, most solutions only reach the level of simple measurement acquisition and coarse-grained statistical analysis, lacking standardized data construction processes such as unified time steps, data quality labeling, and normalization processing for measurement terminals at different locations. This makes it difficult to form a structurally unified operational data table from multi-source heterogeneous measurements that can be directly used for prediction and optimization calculations. On the other hand, existing research usually separates load forecasting, power flow verification, and scheduling optimization. Prediction results are mostly input in the form of load values ​​at single points or in a few time periods, failing to construct an integrated power flow feasible domain with constraints on node voltage, line current, transformer load rate, and adjustable resource capabilities on a unified future time axis. This makes it difficult for optimization models to accurately characterize the feasible operating space for multiple future time steps. In addition, existing load regulation methods tend to focus on a single aspect, such as adjusting time-of-use electricity prices based solely on load curves or controlling a single type of controllable load. They lack support for joint modeling and time-series optimization of controllable loads and energy storage resources, and also lack the process of mapping optimization results to specific equipment action parameters by time steps, generating execution sequences arranged in future time order, and recording them in detail. This makes it difficult to achieve integrated and adaptive load regulation optimization. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that existing power supply and distribution operation load control methods lack unified preprocessing and standardized modeling of multi-source operation data, making it difficult to form a consistent data foundation that can be directly used for prediction and optimization. Load forecasting, power flow constraints, and optimization target models are disconnected from each other, making it difficult to construct a complete power flow feasible domain on a unified future time axis. Controllable load and energy storage control results are difficult to refine into specific equipment actions driven by time steps, lacking a closed-loop mechanism that connects execution and feedback. The problem is how to achieve adaptive load control optimization based on multi-source data in power supply and distribution operation scenarios.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: an adaptive load regulation optimization method for power supply and distribution operation, comprising receiving raw operation records from measurement terminals at different locations, preprocessing them to output a standardized operation data table; selecting continuous time periods according to the time window required for future prediction in the standardized operation data table, constructing multi-dimensional input segments for prediction and writing them into the future data area; applying power flow constraints within the future data area, organizing the power flow feasible domain, constructing an optimization target model, outputting the load adjustment amount and energy storage charging and discharging amount at each time step, and writing them into the future regulation area; sequentially reading the optimized output load adjustment amount and energy storage power records from the future regulation area according to the time steps, converting them into specific equipment action parameters according to the logical measurement point number, combining them into an execution sequence arranged in future time order, and sending them to the execution end.

[0007] As a preferred embodiment of the adaptive load control optimization method for power supply and distribution operation described in this invention, the step of receiving raw operation records from measurement terminals at different locations and outputting a standardized operation data table after preprocessing includes splitting, sorting, and reorganizing the raw operation records sent by measurement terminals at different locations according to equipment address, measurement point number, and timestamp, and aligning the records at a unified time step; performing interpolation on missing points caused by discontinuity, identifying continuous abnormal segments and using median replacement or local smoothing; and simultaneously writing each record into the standardized operation data table according to the logical measurement point number, and attaching data quality labels, acquisition source markers, and corresponding normalization parameters to the records.

[0008] As a preferred embodiment of the adaptive load control optimization method for power supply and distribution operation described in this invention, the step of constructing multi-dimensional input segments for prediction and writing to the future data area includes dividing the historical power, temperature, weather type, date attributes, and contextual logical measurement points corresponding to the future prediction task in the standardized operation data table into segments according to a continuous time series, while maintaining the time order and original data markings within the segments; each segment carries segment source, time index range, and anomaly identification information when input into the prediction program, and the prediction program generates prediction records for future time steps based on the segments, and writes them to the future data area according to logical measurement points and time indices.

[0009] As a preferred embodiment of the adaptive load regulation optimization method for power supply and distribution operation described in this invention, the following steps are included: applying power flow constraints in the future data area, which involves generating node voltage boundaries, line current limits, transformer future load rate constraints, load adjustability limits, and energy storage charging and discharging capacity limits one by one based on the predicted load records of future time steps and line resistance, reactance, current limits, rated voltage, transformer capacity, allowable load rate, and controllable load adjustment range; each constraint is bound to a corresponding future time index and logical measurement point number to form a future power flow feasible domain structure.

[0010] As a preferred embodiment of the adaptive load regulation optimization method for power supply and distribution operation described in this invention, the construction of the optimization target model includes, without changing the time structure and logical measurement point number of the future prediction record, combining the offset between the future load prediction and the reference load and the future line loss value in a linear weighted manner into a target item; and assembling the target item and the corresponding load adjustment variable and energy storage charging and discharging variable in a unified variable sequence according to the time sequence into an optimization target model.

[0011] As a preferred embodiment of the adaptive load regulation optimization method for power supply and distribution operation described in this invention, the load adjustment amount and energy storage charging and discharging amount output at each time step include being written into the future control area in the order of future time steps. Each record retains the original value of the optimization variable and is accompanied by a time index, logical measurement point number, solution mark and association information with the prediction segment. The future control area and the future data area adopt the same row and column structure, and each optimization result corresponds to the prediction record in the time dimension and logical measurement point dimension.

[0012] As a preferred embodiment of the adaptive load regulation optimization method for power supply and distribution operation described in this invention, the following steps are included: when converting the optimized output into equipment action parameters, the load adjustment, energy storage charging and discharging amounts for each future time step are mapped to the corresponding load equipment and energy storage device action commands according to the logical measurement point numbers in the future control area, while keeping the action values ​​consistent with the optimized output; the action commands for each time step are combined into an execution sequence according to the future time order and sent to the execution end; and the actual execution records returned by the execution end are written into the execution record table according to the corresponding time index.

[0013] Another objective of this invention is to provide an adaptive load regulation and optimization system for power supply and distribution operations. This system can construct an optimization target model by applying power flow constraints in the future data area, sorting out the power flow feasible region, outputting the load adjustment amount and energy storage charging and discharging amount at each time step, and writing them into the future regulation area. This solves the problem that current power supply and distribution operation load regulation methods contain the problem of load forecasting and power flow constraints and optimization target models being mutually isolated.

[0014] As a preferred embodiment of the adaptive load regulation and optimization system for power supply and distribution operation described in this invention, the system includes: an operation data construction module, a future state generation module, and a regulation solution and execution module. The operation data construction module receives continuously transmitted raw operation records from measurement terminals at different locations and, through time synchronization, time alignment, missing point interpolation, abnormal segment marking, and normalization processing, organizes all operation records into a standardized operation data table with time index as rows and logical measurement points as columns. The future state generation module extracts continuous time periods from the standardized operation data table according to the prediction cycle, constructs multi-dimensional input segments, performs short-term load prediction, and writes the prediction results into the future data area according to the time index. In the future data area, it applies power flow constraints based on line parameters, equipment rated parameters, and adjustable capacity parameters to form a power flow feasible domain within the future regulation cycle. The regulation solution and execution module constructs an optimization target model based on the future power flow feasible domain, generates load adjustment amounts and energy storage charging and discharging amounts according to future time steps, and writes the optimized output into the future regulation area. Subsequently, based on the logical measurement point numbers, it converts the regulation values ​​into specific regulation action instructions to form an execution sequence, which is then sent to the equipment execution end in chronological order.

[0015] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement an adaptive load control optimization method for power supply and distribution operation.

[0016] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of an adaptive load control optimization method for power supply and distribution operation.

[0017] The beneficial effects of this invention are as follows: The adaptive load regulation optimization method for power supply and distribution operation provided by this invention transforms heterogeneous measurement data into a standardized operation data table with consistent structure and continuous time by performing time synchronization, missing point imputation, abnormal segment identification, and unified normalization on the original operation records from different locations. This provides a reliable and consistent foundation for subsequent prediction and optimization, avoiding the accumulation of prediction bias and optimization model mismatch problems caused by inconsistent data sources in traditional schemes. This invention constructs multi-dimensional input segments within a unified time index structure and writes the prediction results into the future data area step by step, enabling the future load evolution to have a complete timeline expression. This provides a continuous and forward-looking operational status representation for subsequent power flow constraints and feasible domain construction. This invention applies multiple power flow constraints, such as node voltage, line current, transformer load rate, and adjustable resource capacity, to the future data area, forming a power flow feasible domain that reflects the future power grid safety boundary. Based on this, an optimization target model is constructed, enabling controllable loads and energy storage devices to collaboratively participate in regulation across multiple time steps, achieving consistency of decision results under both physical safety and economic constraints. This invention precisely maps optimization results to specific equipment action parameters over time steps, forming a sequence of control commands that can be directly issued to the execution end, thus realizing a complete closed loop from prediction to optimization to execution. This invention significantly improves the foresight, precision, and executability of power supply and distribution operations. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0019] Figure 1 The first embodiment of the present invention provides an overall flowchart of an adaptive load regulation optimization method for power supply and distribution operation. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0021] Example 1, referring to Figure 1 As an embodiment of the present invention, an adaptive load regulation optimization method for power supply and distribution operation is provided, comprising: S1: Receive raw operation records from measurement terminals at different locations, and output standardized operation data tables after preprocessing.

[0022] Furthermore, measurement terminals deployed at different locations in the power supply and distribution network periodically generate raw operating data. Each terminal generates operating records such as three-phase voltage, three-phase current, active power, reactive power, power factor, transformer load rate, line current, distributed power output, energy storage device charging and discharging power, SOC, and temperature according to its own set sampling period. These records are sent to the data access port sequentially according to the sampling time. Each record received by the data access port is decomposed according to a fixed field format, extracting the device address, measurement point number, measured value, and original timestamp, and a unified logical identifier is added to the measured value according to the predetermined logical measurement point coding rules.

[0023] The timestamps of reported data undergo time consistency checks upon arrival at the access port. The access port maintains a unified reference clock internally, synchronized with external time sources. When the deviation between the timestamp of a record and the reference clock exceeds a preset threshold, the record is marked as time-drifted data and replaced with the timestamp of the corresponding moment on the reference clock according to a pre-defined deviation correction method. This ensures that all subsequent measurements strictly correspond to a unique time series position.

[0024] The data access port sorts records from different sampling frequencies by timestamp and writes them to a temporary buffer. Using a uniform 15-minute time step, a target timeline is first constructed, and data slots are created for each 15-minute interval. Then, multiple records from different measurement terminals within each slot are processed in segments: if multiple sampling points exist in a slot, a weighted average is performed according to their chronological order; if a slot contains only one sampling point, that value is directly used as the representative value for that time point; if a slot is completely empty, it is marked as unfillable and proceeds to the next stage of processing.

[0025] It should be noted that, for the locations where missing markers exist, linear interpolation intervals are constructed for the valid measurement points before and after the missing point. The interpolation calculation generates interpolated values ​​for the missing time points after uniformly dividing the intervals according to time, and the interpolation source (before and after the missing time points) and the interpolation formula are recorded in the missing point's additional attributes. If the sampling interval between two consecutive time points exceeds a set range (e.g., more than 30 minutes), interpolation calculation is not performed on the missing point, but the missing marker is retained in the missing point to avoid artificially constructing large segments of unreliable data.

[0026] For measurements exhibiting abrupt jumps, a sliding time window statistical sequence is established. This sequence contains the N most recent valid sampling points (e.g., N=96, corresponding to 24 hours). Upon each new data point, it is compared with the mean and standard deviation of the sliding window. When the deviation of the measurement value from the mean exceeds a preset multiple (e.g., 3 times the standard deviation), the point is marked as an outlier. Outlier values ​​are replaced using median filtering, with an odd-numbered median filter window size (e.g., 5 points). The median value within the window, sorted by numerical value, is used as the replacement value. Consecutive outlier segments are not replaced to preserve fault characteristics.

[0027] After missing data filling and outlier handling are completed, all sampling points are written into a unified time index structure in chronological order. The time index structure stores the processed values ​​in an organization format corresponding to logical measurement points at different times. For each logical measurement point, extended attributes corresponding to that point are also saved, including sampling source type, voltage level, topology location, physical unit, and data quality label (normal, interpolated, outlier, interrupted).

[0028] When preparing for the standardization of measurement values, the historical statistical parameters of the corresponding logical measurement points are first read, including the maximum, minimum, mean, and standard deviation within the past observation window. The measurement values ​​are then transformed according to a preset normalization formula, such as Z-score normalization.

[0029] If the measurement point belongs to a dimension with physical upper and lower limits, then min-max normalization is preferred. The normalized result is stored in a separate field along with the original value and statistical parameters.

[0030] After completing time alignment, missing data handling, anomaly detection, and normalization, a standardized runtime data table is finally constructed, with a unified time step as the index and logical measurement points as the columns.

[0031] S2: In the standardized operating data table, select continuous time periods according to the time window required for future prediction, construct multi-dimensional input segments for prediction and write them into the future data area.

[0032] Furthermore, historical power data, weather and temperature data, time category data, and contextual data related to load changes for each target node are read from the time index structure of the standardized operational data table. During reading, a data subset is constructed according to the time index order, with consecutive time steps as rows and all relevant logical measurement points as columns. Each record point in the data subset contains the formed standardized value, the original value, and a data quality label. Before constructing the prediction input sequence, the data subset is first segmented according to time windows, for example, by using a fixed-length sliding window to sequentially extract records from several consecutive time steps from the beginning of the data subset. Each sliding window constitutes a sample segment, maintaining the time order within the segment, while different segments are shifted forward by a fixed offset according to the time step.

[0033] When constructing sample segments, logical measurement points related to short-term load changes (such as historical load measurement points of the target node, temperature measurement points of the corresponding area, weather-related coded measurement points, status measurement points indicating whether it is a working day, and status markers indicating whether there are major events) are extracted one by one, so that the segment retains only the associated logical measurement points of samples of the same category in the horizontal dimension. To ensure data consistency within the segment, markers belonging to interpolation, abnormal replacement, and long-term missing data are recorded separately. The location of the marked data within the segment is retained in the additional attributes for confidence adjustment or sample screening in subsequent prediction processing.

[0034] After constructing the segments, the normalized values ​​within each segment are arranged in their original order to form a continuous time series structure. If a segment's length is less than a preset value, that segment is not included in the prediction input construction. When the continuous time series structure is written to the prediction input buffer, it includes the segment's source time range, the set of logical measurement point numbers it contains, and internal segment marker information. The prediction input buffer stacks multiple segments in chronological order. For segments that cross date boundaries, a cross-day marker is retained to prevent subsequent processing from mistakenly merging cross-day segments into data from the same calendar day.

[0035] Once the prediction input buffer accumulates to a certain quantity, a sequence standard check is performed. The check includes: whether the time steps within a segment are strictly continuous; whether there are missing markers exceeding the allowed proportion within the segment; whether attribute values ​​such as weather, temperature, or holidays remain consistent with historical data; and whether there are time misalignments or duplicate records within the segment. After the sequence passes the standard check, the segments are organized into a continuous multidimensional sequence structure according to the input order, and written into the prediction execution entry point along with the corresponding target prediction duration.

[0036] It should be noted that after the input is constructed, each time series is sequentially fed into the prediction program according to a preset order. During runtime, the prediction program reads the standardized data from the input sequence sequentially, maintaining the original temporal order within each segment. When processing each segment, the prediction program automatically identifies whether it contains long-term missing data, interpolated segments, or anomalous substitution points. When multiple consecutive anomalous substitution points exist within a segment, the prediction program marks the segment individually before performing time step expansion to facilitate confidence correction after prediction. After processing the time series sequentially, the prediction program outputs predicted data for several future time steps. The output data is stored in the form of logical measurement points corresponding to the predicted time index, including standardized predicted values ​​and corresponding time range labels.

[0037] After outputting the predicted values, the prediction results are aligned with the original time boundaries within the segment. For each future time step, the predicted value is merged with the corresponding time index to form a prediction run data record for the future time. This record contains the predicted value, segment source information, the input sequence number at the time of prediction, and the set of logical measurement points involved in the prediction. If a segment is marked as of poor quality (e.g., with a high proportion of missing points), the prediction program will add a quality flag to the output record.

[0038] After all output processing is complete, the predicted records for future time periods are written to a separate future data area in chronological order. The future data area maintains the same organization as the generated data table, using a uniform time step as the primary index and logical measurement points as column identifiers. Each record in the future data area not only stores the predicted value but also includes the segment number at which the prediction occurred, the original segment's tagging information, and the prediction execution order number.

[0039] S3: Apply power flow constraints within the future data area, organize the feasible power flow domain, construct the optimization target model, output the load adjustment amount and energy storage charging and discharging amount at each time step, and write them into the future control area.

[0040] Furthermore, within the formed future data region, the predicted power values ​​for all future time steps are arranged according to a uniform time step size and form a structurally consistent record with the corresponding logical measurement points. The predicted active power, reactive power, and rated parameters of related equipment for the target node are read from the future data region. These predicted values ​​are extracted one by one according to the time index order to construct the power flow constraints required for subsequent optimization. During the reading process, the time steps, logical measurement point numbers, and quality markers in the prediction records remain unchanged, ensuring that subsequent power flow calculations accurately correspond to each future moment in the prediction sequence.

[0041] For each future time step The node voltage is approximated using the following linearization formula:

[0042] in, Indicating the future In each time step, the time index structure is numbered as follows: The predicted value of the node voltage; Determined by the rated voltage of the equipment or the voltage of a reference node (such as a bus), it serves as a fixed reference quantity when constructing power flow constraints and does not change with future time; This indicates the line from the node corresponding to the logical measurement point. To the node The resistance component is taken from a pre-registered table of topology and line parameters; Indicates the line from node To the node The reactance component, and They are stored in the parameter table in a fixed manner; Indicates the future number Each time step line node To the node The active power; Indicates the future number Each time step line node To the node The reactive power is obtained from the reactive power prediction value or reactive power estimation rule in the future data area.

[0043] The line current in future time steps is approximately constrained by the following constraints:

[0044] in, Represents line nodes To the node The maximum allowable current value is a fixed parameter defined in the equipment nameplate or engineering specifications. This represents the rated voltage value of the line, which is given in the equipment technical file and is used to approximate the active power into a current-constrained expression.

[0045] The future load factor of the transformer is determined by the following constraints:

[0046] in, Indicates the future number The predicted active load of the transformer at each time step comes from the prediction records of the corresponding transformer load logic measurement points in the future data area. Indicates the future number The amount of controllable load power adjustment to be prepared for adjustment at each time step; The maximum allowable operating load rate is a fixed engineering safety parameter; The rated capacity of the transformer is derived from the equipment's basic parameters.

[0047] The future time-step controllable load power adjustment range is determined by the following constraints:

[0048] Energy storage charge and discharge limits are constituted by the following constraints:

[0049] in, This indicates the minimum allowable downward adjustment and the maximum allowable upward adjustment, as determined by the controllable equipment configuration table. This indicates the charging power of the energy storage at a future time step; This indicates the discharge power of the energy storage at a future time step; This indicates the rated power capacity of the energy storage inverter and battery pack.

[0050] It should be noted that when constructing power flow constraints, the predicted values ​​are read in the order of the time index in the future data area for each future time step. The read predicted records are then mapped to the logical measurement point table, ensuring that all future power values ​​strictly correspond to the logical measurement point numbers. Each power flow constraint, when written to the power flow constraint data area, includes a time index, logical measurement point number, the line parameters used, the equipment capacity, and whether it references any anomaly flags.

[0051] All trend constraints in the future time step are stacked in chronological order to form the trend feasible domain within the future regulation cycle.

[0052] Furthermore, based on the established feasible future power flow domain, the load forecast records and processed power flow boundary parameters for each future time step are read sequentially according to the time index order of the future data area. During the reading process, the structure of the future data area remains unchanged; each record, after extraction, retains its logical measurement point number, future time index, prediction source segment number, and data quality marker. These records are then written to the optimized input buffer and arranged in chronological order, without performing any cross-time merging or clustering operations.

[0053] After optimizing the input buffer sequence, the corresponding control decision variables are constructed for each future time step. These decision variables include: the adjustment amount of the controllable load for that time step, the charging power value of the energy storage device, and the discharging power value of the energy storage device. To ensure accurate mapping between variables and records in the future data area during subsequent solution stages, each variable is accompanied by several basic information fields during creation, including: the future time step it belongs to, the associated logical measurement point number, the correspondence between it and the future prediction segment, the reference position of the power flow boundary parameter in the power flow constraint construction process, and the minimum and maximum allowable value range of the variable. All variables are linearly arranged in the order of future time steps to form a variable sequence, maintaining a strictly consistent index order.

[0054] After establishing the variable sequence, the optimization objective function is expressed as follows:

[0055] in, Represents the objective function value during the optimization process; represents the comprehensive evaluation quantity. The weighting coefficients used to measure the deviation between the predicted load and the reference load are provided by a fixed parameter table before the model is built and do not change over time. Weighting coefficients used to measure line loss; This indicates the future data area containing records of the next... Load forecast values ​​for each time step; Indicates the reference load value; This represents the linearized value of line loss calculated based on line parameters and predicted power flow at future time steps, corresponding one-to-one with future prediction records.

[0056] During the writing of the objective function, the generated data is not reprocessed; only the results are referenced. A time index and logical measurement point identifier are added to each term of the objective function, so that the subsequent solver can trace the correspondence between the terms of the objective function after solving the variables.

[0057] After the objective function is constructed, power flow constraints are written textually to the power flow constraint description area. The writing mode is as follows: for each future time step, in the order of nodal voltage linear approximation, line current limit conditions, transformer load rate upper limit conditions, controllable load adjustment capability conditions, and energy storage device charging and discharging capability conditions, the registered parameter ranges, physical boundaries, line parameters, voltage levels, rated values, etc., are bound by variable numbers. The writing of power flow constraints follows these principles: For each power flow constraint, a mapping relationship is established between the power flow constraint entry and the future time index; Record the number of each logical measurement point involved in the power flow constraint and establish a one-to-one binding with the variable sequence; For entries that reference external fixed parameters (such as resistance, reactance, current limit, load factor limit, etc.), no recalculation is performed; All variable references in the power flow constraint entries are written in the predetermined order of the variable sequence, without adjusting or rearranging the variable numbers.

[0058] After completing the writing of the power flow constraint entries, variable entries, and objective entries, a complete optimization model that can be directly processed by the MILP solver is constructed by combining the three types of elements using a unified index. The model structure includes: variable vectors, constraint lists, objective function units, and the mapping relationships between each objective and constraint and future time steps. The final structure of the model maintains consistency with the temporal order of the future data region, ensuring that the model does not produce any cross-time confusion or historical time backtracking issues during the final solution.

[0059] Once the model structure is prepared, the solution process is invoked, and iterative solutions are executed step by step along the future time dimension. The solver attempts to assign values ​​to variables sequentially, while simultaneously checking whether variables have reached their upper or lower bounds based on the power flow boundaries recorded in the power flow constraint description area. If a variable fails to meet its corresponding power flow constraint, the solver automatically backtracks and changes the variable's value range, then re-attempts to satisfy all power flow constraints. After convergence, the solver outputs the adjustment amounts and energy storage charge / discharge values ​​for all future time steps, writing them sequentially into the future control area according to the time index. The structure of the future control area is completely consistent with the future data area: future time steps serve as row indices, logical measurement points as columns, and adjustment values ​​and solution markers as data fields, forming a complete executable control sequence.

[0060] S4: Read the optimized output load adjustment and energy storage power records from the future control zone in time step order, convert them into specific equipment action parameters according to the logical measurement point number, combine them into an execution sequence arranged in future time order, and send them to the execution end.

[0061] Furthermore, after writing the future control zone, the system first reads the controllable load adjustment, energy storage charging power, energy storage discharging power, and solution markers for each future time step, following the time index order of the future control zone. During the reading process, the structure of the future control zone remains unchanged; the content of any field is based on the complete output record without data correction, merging, or reorganization. After reading, the control value for each time step is structurally compared with the predicted load record for the corresponding time step in the future data area. This confirms that the two completely correspond in time index, logical measurement point number, and future segment source marker, ensuring that subsequent technical actions do not involve cross-time execution or logical measurement point mismatches.

[0062] After comparing the correspondence between the control values ​​and the predicted records, the control action generation process begins. This process breaks down the adjustment value for each time step in the future control zone into several sub-items according to the logical measurement point number. For example, when a logical measurement point corresponds to a user-side load, that adjustment value is converted into a load adjustment action; when a logical measurement point corresponds to an energy storage battery, that adjustment value is converted into an energy storage charging action or an energy storage discharging action. The action generation process does not reference any external intelligent rules, nor does it dynamically infer based on load type. Instead, it strictly follows the logical measurement point number and its associated equipment attribute table, mapping each adjustment value to specific equipment action parameters, including load adjustment direction, power adjustment amplitude, charging power command, and discharging power command. All action parameters maintain a one-to-one mapping relationship with the original adjustment values ​​recorded in the future control zone.

[0063] After the action parameters are generated, they are combined into an execution sequence according to the future time index. The execution sequence consists of multiple consecutive time steps, each containing one or more device action parameters. The execution sequence strictly maintains a temporal order to ensure that the execution end can implement actions step-by-step according to the solved future control rhythm. During execution sequence orchestration, the action parameters are not sorted or clustered, nor are cross-time window merging processes performed.

[0064] After the execution sequence is generated, the action issuance and execution monitoring process begins. During action issuance, all action parameters for the corresponding time step are sequentially converted into numerical commands recognizable by the execution end, according to the chronological order of the execution sequence. During this conversion, the values ​​of the action parameters remain unchanged; no compression, scaling, or offset is applied to the action values. For example, if the load adjustment given by the future control zone is a specific power value, this power value will be written into the action command as is; if the action parameter of a certain energy storage device is a fixed charging power, this power value will be written into the corresponding command without adjustment according to the original record.

[0065] After being generated, action commands are sent to the execution end for processing in chronological order. Upon receiving a command, the execution end, based on the logical measurement point number and device attribute fields contained in the command, performs adjustment actions on the associated equipment at the corresponding time step. For example, when the logical measurement point in the command points to a low-voltage load, the execution end directly applies the power adjustment amount of the command to the corresponding load's power setpoint; when the logical measurement point in the command corresponds to an energy storage device, the execution end will switch the energy storage device's operating mode to charging or discharging in the current time step and execute the corresponding function according to the power value specified in the command. During the action implementation process, no modification is made to the original optimized values ​​in the future control zone; only the provided values ​​are used as the basis for the execution end's operation.

[0066] After the action command is sent, the execution status record returned by the execution terminal also needs to be collected, including the current power of the device, the device operating status, the actual adjustment range, and the action response time. All execution records are written into the execution record structure in time index order, and a mapping relationship is established with the future control area record of the corresponding time step. Each record in the execution record structure contains fields such as execution time, device identifier, logical measurement point number, actual execution value, and response status flag, and its time index is consistent with the time index of the action command.

[0067] Example 2, one embodiment of the present invention, provides an adaptive load regulation optimization method for power supply and distribution operation. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0068] First, a 10kV / 0.4kV power distribution system in an industrial park was selected as the test object. This park includes 3 distribution transformers, 12 low-voltage feeders, and approximately 60 key electricity metering points. Load types include production line equipment, air conditioning power, office lighting, and 20 AC slow-charging piles. The test period was three consecutive days, with each day following a typical weekday operating schedule. The morning peak load was concentrated between 8:30 and 10:30, and the evening peak between 18:00 and 21:00. To verify the advantages of the method of this invention compared to existing timed scheduling strategies, the experiment configured two operating modes: a traditional strategy group and the strategy group of this invention, which were executed alternately on different days.

[0069] During the data acquisition and preparation phase, smart meters and measurement terminals are deployed on the low-voltage side outgoing lines of each transformer, the ends of main feeders, and key load branches to collect raw operating records such as three-phase voltage, current, active power, reactive power, and power factor. The sampling period is uniformly set to 15 minutes. Power recording devices are installed at energy storage devices and charging pile junction points to collect charging and discharging power and actual charging curves. At the same time, outdoor temperature and weather type information for the corresponding time period are obtained from the meteorological interface. In the data preprocessing phase, the collected raw records are sorted by timestamp, and obviously erroneous timestamp data is removed. Missing points caused by short-term interruptions are filled in using front-to-back point interpolation. Continuous abnormal jump segments are marked using sliding window statistics and replaced with the median value. Then, all measurement values ​​are resampled according to a uniform time step, and the measurement terminal records at different locations are mapped into a standardized operating data table with time as the row and logical measurement points as the column.

[0070] In the load forecasting and future data generation phase, historical data from the past 30 days were selected from standardized operational data tables as training and validation data for the model. Multidimensional input segments were constructed in 30-time-step windows, each segment containing features such as the target transformer's active load, corresponding temperature, holiday markers, and typical production shift information. After offline training, during three consecutive days of experimental operation, the latest historical segment was read every 15 minutes to generate short-term load forecasts for the next two hours (a total of eight time steps). The forecast results and associated features were written to the future data area according to time indexes. Subsequently, for the next two-hour time axis, parameters such as line current limits, transformer allowable load rate limits, rated charging and discharging power of energy storage, and adjustable power limits of charging piles were uniformly entered into the power flow constraint configuration to form a power flow feasible domain description for each future time step. Based on this, an optimized target model was constructed.

[0071] During the control and execution phase, at the start of each 15-minute control cycle, the predicted load curve and power flow feasible region within the current future data area are read to calculate the load adjustment amount and energy storage charging / discharging power for each time step within the next two hours. The results are then written into the future control area. Subsequently, the optimized output in the future control area is read sequentially according to the time step. Based on the logical measurement point number, the corresponding items are converted into specific equipment action parameters, such as the set temperature fine-tuning range for air conditioning group control, the maximum charging power limit for charging piles, and the charging or discharging power setting for energy storage devices. These action parameters are then used to generate an execution sequence according to the future time order and sent to the field control terminal. Traditional strategy groups use existing fixed-period peak shaving schemes, which uniformly reduce the number of air conditioners in operation and limit the charging pile power during the evening peak, without performing rolling prediction and rolling optimization. After three consecutive days of testing, key operating indicators are extracted from the standardized operating data table and the future control area to generate comparative data to verify the effectiveness of the method of this invention.

[0072] Table 1 Experimental Data

[0073] As shown in Table 1, in the three-day comparative test, the method of this invention outperformed the traditional strategy in several key indicators, including daily maximum load, peak-to-valley difference, line loss rate, voltage qualification rate, and the number of user-perceived complaints. Taking daily maximum load as an example, the maximum load of the traditional strategy on days 1 to 3 was 4850kW, 4720kW, and 4905kW, respectively, while the maximum load of this invention on the corresponding days was reduced to 4300kW, 4210kW, and 4325kW, with a peak reduction of about 10%. This difference stems from the fact that this invention performs rolling predictions based on standardized operating data tables in each control cycle and optimizes the load curves of multiple time steps in the future data area, enabling controllable load and energy storage output to be distributed in advance in the pre-peak, mid-peak, and post-peak stages. In contrast, the traditional strategy only implements peak shaving in a coarse-grained manner during preset periods and cannot make detailed adjustments for load evolution in the next two hours.

[0074] Regarding the peak-valley difference index, the traditional strategy's peak-valley differences are 2600kW, 2450kW, and 2700kW over three days, respectively, while the present invention corresponds to 1800kW, 1750kW, and 1850kW, showing a significant narrowing of the peak-valley difference. This result reflects that the present invention, by applying power flow constraints based on future data areas and constructing a power flow feasible region, smooths the load shape across multiple time steps in the optimization target model. This allows the control actions to not only suppress peak loads but also effectively increase off-peak loads by initiating energy storage charging in advance and staggering some controllable loads during off-peak periods, achieving a more uniform load distribution. Traditional fixed-time-period control cannot precisely set adjustment amounts for every 15-minute time step, nor can it guarantee multi-time-period joint balance within the grid safety boundary.

[0075] This invention also demonstrates significant advantages in terms of daily average line loss rate and voltage qualification rate. The daily average line loss rate of traditional strategies fluctuates between 4.5% and 4.9%, while this invention controls the line loss rate to 3.4% to 3.7%. This indicates that the load adjustment and energy storage charging / discharging power output, under the constraints of the power flow feasible domain, help alleviate the overload of some lines and transformers during high-load periods, reduce reactive power flow, and thus make the distribution network operation closer to the high-efficiency range. Similarly, regarding the voltage qualification rate, the qualification rate of traditional strategies is 95.5% to 96.2% over three days, while this invention maintains a stable rate above 99%. This shows that under the method of this invention, the control scheme fully considers future power flow constraints on voltage-sensitive nodes, avoiding the problem of local voltage fluctuations caused by concentrated peak shaving in traditional schemes.

[0076] The overall effectiveness of this invention can be further illustrated by combining two indicators: daily energy savings (kWh) and the number of user-perceived complaints (times). The traditional strategy resulted in zero energy savings over three days, while this invention achieved energy savings of approximately 520kWh, 510kWh, and 540kWh respectively over those three days. This means that, while maintaining production capacity, load regulation, through refined arrangement of energy storage strategies and flexible loads, makes overall energy utilization more efficient. Simultaneously, the traditional strategy generated 5, 4, and 6 user-perceived complaints over three days, while this invention generated only 1 complaint on each of the corresponding days. This demonstrates that by precisely mapping optimized outputs and logical measurement points in the future control zone to equipment action parameters and generating execution sequences according to time steps, the control actions are more controllable and flexible, avoiding significant impacts on user comfort and production processes caused by large-scale, abrupt load adjustments.

[0077] In summary, the table data objectively reflects that this invention, through the construction of standardized operational data tables, multi-step prediction and power flow feasible domain organization in the future data area, and the optimized output of the future control area and the complete chain generated by the execution sequence according to time steps, significantly outperforms traditional strategies in multiple dimensions, including peak reduction, peak-valley balancing, line loss reduction, voltage quality improvement, and user experience. This adaptive load control optimization method based on a unified data structure and future time axis demonstrates stronger foresight, synergy, and engineering feasibility compared to existing load management methods driven solely by empirical rules or single-step predictions, exhibiting significant creativity and novelty.

[0078] Example 3, one embodiment of the present invention, provides an adaptive load regulation and optimization system for power supply and distribution operation, including an operation data construction module, a future state generation module, and a regulation solution and execution module.

[0079] The operational data construction module receives continuously transmitted raw operational records from measurement terminals at different locations. Through time synchronization, time alignment, missing point imputation, abnormal segment marking, and normalization, it organizes all operational records into a standardized operational data table with time index as rows and logical measurement points as columns. The future state generation module extracts continuous time periods from the standardized operational data table according to the prediction cycle, constructs multi-dimensional input segments, and performs short-term load forecasting. The forecast results are written into the future data area according to the time index. In the future data area, power flow constraints are applied based on line parameters, equipment rated parameters, and adjustable capacity parameters to form the power flow feasible domain within the future control cycle. The control solution and execution module constructs an optimized target model based on the future power flow feasible domain and generates load adjustment and energy storage charging and discharging quantities according to future time steps. The optimized output is written into the future control area. Subsequently, the control values ​​are converted into specific control action instructions according to the logical measurement point numbers to form an execution sequence, which is sent to the equipment execution end in chronological order.

[0080] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0081] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0082] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0083] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An adaptive load regulation optimization method for power supply and distribution operation, characterized in that, The method comprises the following steps: Receiving raw operation records from measurement terminals at different locations, and outputting standardized operation data tables after preprocessing; In the standardized operation data tables, continuous time periods are selected according to the time window required for future prediction, and multi-dimensional input segments are constructed for prediction and writing into the future data area; In the future data area, the power flow constraints are applied, the power flow feasible region is arranged, the optimization objective model is constructed, and the load adjustment amount and the energy storage charging and discharging amount at each time step are outputted and written into the future regulation area; The optimized output load adjustment amount and energy storage power record are read from the future regulation area in sequence according to the time step, converted into specific device action parameters according to the logical measurement point number, combined into an execution sequence arranged in the future time sequence, and sent to the execution end.

2. The adaptive load regulation optimization method for power supply and distribution operation according to claim 1, characterized in that: The raw operation records from the measurement terminals at different locations are split, sorted and reorganized according to the device address, measurement point number and time stamp, and the records are time-aligned under the unified time step; The missing points generated by the discontinuity are interpolated, the continuous abnormal paragraphs are identified and replaced by the median or locally smoothed, and each record is written into the standardized operation data table according to the logical measurement point number, and the data quality label, collection source mark and corresponding normalization parameter are added to the record. The historical power, temperature, weather type, date attribute and context logical measurement point corresponding to the future prediction task in the standardized operation data table are cut into segments according to the continuous time sequence, the time sequence and the original data label in the segment are kept unchanged, the segment source, time index range and abnormal identification information are added to each segment when inputting the prediction program, the prediction program generates the prediction record of the future time step according to the segment, and the prediction record is written into the future data area according to the logical measurement point and the time index.

3. The adaptive load regulation optimization method for power supply and distribution operation according to claim 2, characterized in that: The power flow constraints in the future data area include generating node voltage boundary, line current limit, transformer future load rate constraint, load adjustable capacity limit and energy storage charging and discharging capacity limit information according to the predicted load record of the future time step, line resistance, reactance, current limit, rated voltage, transformer capacity, allowable load rate and controllable load adjustment interval; 4. The adaptive load regulation optimization method for power supply and distribution operation according to claim 3, characterized in that: Each constraint is bound to the corresponding future time index and logical measurement point number to form the future power flow feasible region structure. The optimization objective model is constructed by combining the future load prediction and the offset of the reference load with the future line loss value in a linear weighting manner as the objective term without changing the time structure and logical measurement point number of the future prediction record; 5. The adaptive load regulation optimization method for power supply operation of claim 4, wherein: The objective term, the load adjustment variable and the energy storage charging and discharging variable are sequentially assembled into the optimization objective model according to the unified variable sequence. The load adjustment amount and the energy storage charging and discharging amount at each time step are written into the future regulation area in the order of the future time step, each record retains the original value of the optimization variable, and the time index, logical measurement point number, solution mark and association information with the prediction segment are attached.

6. The adaptive load regulation optimization method for power supply operation of claim 5, wherein: ​ The future control zone and the future data zone adopt the same row and column structure, and each optimization result and prediction record corresponds to the time dimension and logical measurement point dimension.

7. The adaptive load regulation optimization method for power supply operation of claim 6, wherein: The load adjustment amount and energy storage charging and discharging amount of each time step are output and written into the future control area. This includes mapping the load adjustment amount, energy storage charging amount and discharging amount of each future time step to the corresponding load equipment and energy storage device action instructions according to the logical measurement point number in the future control area when converting the optimized output into equipment action parameters, and keeping the action value consistent with the optimized output. The action instructions for each time step are combined into an execution sequence according to the future time order and sent to the execution end. The actual execution records returned by the execution end are written into the execution record table according to the corresponding time index.

8. A system for power supply and distribution operation adaptive load regulation optimization method according to any one of claims 1 to 7, characterized in that: It includes a data construction module, a future state generation module, and a control, solution, and execution module; The operation data construction module is used to receive continuously uploaded raw operation records from measurement terminals at different locations, and through time synchronization, time alignment, missing point interpolation, abnormal paragraph marking and normalization processing, organize all operation records into a standardized operation data table with time index as rows and logical measurement points as columns. The future state generation module is used to extract continuous time periods from the standardized operation data table according to the prediction cycle, construct multi-dimensional input segments and perform short-term load forecasting, write the prediction results into the future data area according to the time index, and apply power flow constraints in the future data area according to line parameters, equipment rated parameters and adjustable capacity parameters to form the power flow feasible domain within the future control cycle. The control solution and execution module is used to construct an optimization target model based on the future power flow feasible domain, and generate load adjustment amount and energy storage charging and discharging amount according to the future time step, and write the optimization output into the future control zone; then, according to the logical measurement point number, the control value is converted into specific control action instructions to form an execution sequence, and sent to the device execution end in time order. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. When the processor executes the computer program, it implements the steps of the adaptive load control optimization method for power supply and distribution operation as described in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the adaptive load control optimization method for power supply and distribution operation as described in any one of claims 1 to 7.

Images