Distributed energy optimization scheduling method based on virtual power plant

Abstract

The invention discloses a distributed energy optimization scheduling method based on a virtual power plant, particularly relates to the field of distributed energy aggregation control, and is used for solving the problems that resource grouping execution performance is unstable and grid-connected side electrical constraints are difficult to be synchronously brought into a scheduling decision. Scheduling input is constructed by collecting grid-connected point active power and scheduling target active variable quantity, resource groups are formed according to a resource access position and a resource control mode, and bidirectional detection is implemented to establish response fingerprints; calculating a net response energy following ratio and a voltage sensitive stability score to generate an admission discrimination mark, outputting an effective list and a grouping effective adjustment boundary, and completing electrical admission check based on a power distribution network calculation model to generate an instruction draft and an alternative list; and issuing a final scheduling instruction, and forming a deviation record and response fingerprint update according to a resource acceptance receipt and a grid-connected point measurement result, thereby realizing adjustable resource screening and optimal scheduling landing under the power distribution network security constraint.

Description

Technical Field

[0001] This invention relates to the field of distributed energy aggregation control, and more specifically, to a distributed energy optimization scheduling method based on a virtual power plant. Background Technology

[0002] With the continuous increase in distributed photovoltaic, energy storage, charging facilities, and adjustable loads in power distribution networks, individual resources are small in scale, geographically dispersed, and their operating states change rapidly. Traditional single-site or single-equipment dispatching methods struggle to simultaneously ensure grid security and economic efficiency. Therefore, the industry typically employs virtual power plants to aggregate various distributed resources. By collecting operational and forecasting information on a platform, and combining this with electricity prices and constraints, an optimized dispatching plan is generated. Power or load adjustment commands are then issued to each connected resource to participate in peak shaving, ancillary services, or market transactions. Based on this approach, existing solutions have proposed different aggregation modeling and optimization solution paths, including collaborative optimization frameworks for joint dispatching of multiple virtual power plants, and implementation methods based on information aggregation and optimization models to form dispatching strategies.

[0003] However, existing methods often treat "resource adjustability" as a directly usable input: the platform calculates the optimal plan based on the adjustable range reported by members, historical response performance, and prediction results, assuming that these capabilities will be fulfilled as scheduled during execution. The problem lies in the fact that distributed resources belong to different owners, and there are local control policies, temporary user interventions, communication delays, and equipment state switching on-site. This often leads to inconsistencies between reported capabilities and actual executable capabilities, and even in joint scheduling or competitive scenarios, there may be a motivation to "preferentially report in order to obtain a more favorable scheduling result." When such inconsistencies occur, although the optimization model can calculate seemingly optimal scheduling instructions, these instructions may be unexecutable, partially executed, or immediately offset by local policies on-site. The platform can only continuously track and correct these deviations, affecting the stability of the virtual power plant's external commitments, making accountability difficult to trace, and ultimately weakening the credibility of the virtual power plant as a controllable resource participating in grid operation.

[0004] To address the aforementioned problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies in the prior art, embodiments of the present invention provide a distributed energy optimization scheduling method based on a virtual power plant. This method constructs scheduling inputs by collecting active power data from grid-connected points and changes in active power to the scheduling target. It groups resources according to their access location and control method, implements bidirectional detection to establish response fingerprints, calculates the net response energy following ratio and voltage sensitivity stability score to generate access discrimination markers, and outputs an effective list and effective adjustment boundaries for each group. Based on a distribution network calculation model, it completes electrical access verification, generates draft instructions and a candidate list, issues final scheduling instructions, and updates deviation records and response fingerprints based on resource acceptance receipts and grid-connected point measurement results. This achieves the implementation of adjustable resource screening and optimized scheduling under distribution network safety constraints, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: S1: Collect the active power of the virtual power plant grid connection point, receive the active power change of the scheduling target, and form the scheduling input for this cycle; S2: Issue upward and downward probe power commands according to resource groups, collect active power waveforms and voltage waveforms at grid connection points, and generate response fingerprints for resource groups; S3: Generate admission discrimination markers for resource groups based on response fingerprints, and include resource groups into the effective list or retest list according to the admission discrimination markers. The effective adjustment boundaries of resource groups in the effective list are determined simultaneously. S4: Map the effective list to the distribution network nodes, perform electrical access verification based on power flow constraints and voltage constraints, issue electrical access tags to resource groups that pass the electrical access verification, bind the electrical access tags to the effective adjustment boundary of the group to generate a scheduling plan and instruction draft, and generate a candidate list at the same time; S5: Issue a draft instruction and receive resource acceptance receipts. Resource groups with missing acceptance receipts are removed from the scheduling plan and replaced by the alternative list. The final scheduling instruction is formed and executed. After execution, a deviation record is generated by comparing the active power at the grid connection point with the active power change of the scheduling target. The deviation record is written into the response fingerprint to complete the response fingerprint update.

[0007] Furthermore, the active power measurement sequence of the grid-connected point and the active power change of the scheduling target are collected. The active power measurement sequence of the grid-connected point is processed by unifying the signs of the power injected into the grid as positive and the power absorbed from the grid as negative, so as to obtain the active power measurement sequence of the grid-connected point with consistent signs. The power change rate sequence is calculated based on the power difference between adjacent sampling points and the timestamp interval. The stable segment index set is extracted according to the power change rate sequence. The median value of the power value corresponding to the stable segment index set is taken as the reference power. The reference power and the active power change of the scheduling target constitute the scheduling input.

[0008] Furthermore, resource group identifiers are generated based on resource access location coding and resource control method coding. Resources with the same resource group identifier are merged into resource groups to form a resource group set. The resource group identifier and resource list are registered. The resource group set remains unchanged during the scheduling cycle. The resource group identifier is used throughout the entire process of detection power command issuance and response fingerprint recording.

[0009] Furthermore, each resource group reads the adjustable upper and lower limits and the upper limit of the detection amplitude. The command to increase the detection power is equal to the positive direction of the increase in detection amplitude, and the command to decrease the detection power is equal to the negative direction of the decrease in detection amplitude. The active power waveform and voltage waveform at the grid connection point are extracted according to the start and end time of the detection hold segment and the timestamps are aligned to form a response fingerprint containing the segment boundaries.

[0010] Furthermore, for each resource group identifier, the detection following segment boundary and the detection stable segment boundary are used to form a following window. The power offset value in the direction of the active power waveform at the grid connection point is generated relative to the reference power. The positive energy and the reverse energy are accumulated and the difference is calculated to obtain the net response energy. The net response energy is compared with the command expected energy to obtain the following ratio in the upward adjustment direction and the following ratio in the downward adjustment direction. The bidirectional consistent difference is mapped to the bidirectional consistent deduction factor and the deduction is completed to obtain the net response energy following ratio.

[0011] Furthermore, for each resource group identifier, a sub-window is cut according to the power offset change trend within the following window. The voltage slope sample is calculated based on the difference between the endpoints of the grid connection point voltage waveform and the grid connection point active power waveform to form an upward voltage slope group and a downward voltage slope group. These are merged to obtain the total voltage slope set and the bidirectional stability band is extracted. The voltage sensitivity stability score is obtained based on the hit rate of the bidirectional stability band and the deviation penalty. Based on the threshold comparison, an admission discrimination mark is generated and the effective list, retest list and group effective adjustment boundary are output.

[0012] Furthermore, the effective list is read to obtain the resource grouping identifier and the effective adjustment boundary of the grouping. A distribution network node mapping table is established according to the resource grouping identifier and the distribution network calculation model is assembled. Based on the node load baseline, forward and backward scanning is performed to obtain the baseline node voltage and baseline line current. The node voltage sensitivity coefficient and line current sensitivity coefficient are extracted node by node according to the uniform detection of injected power change, and a lower limit of the sensitivity coefficient is set for the node voltage sensitivity coefficient.

[0013] Furthermore, according to the distribution network node mapping table, the commanded power changes are summarized into node injected power changes, and the node voltage and line current are estimated. Based on the node voltage allowable interval boundary and the upper boundary of the line allowable current interval, the estimation is checked and power flow verification is performed to generate electrical access markers. The voltage limiting power margin is calculated by combining the node voltage margin and the node voltage sensitivity coefficient, and the electrical executable boundary is determined with the effective adjustment boundary of the group. The active power change of the scheduling target is decomposed according to the electrical executable boundary to generate a draft instruction, and a candidate list is generated by sorting according to the electrical executable boundary and the line current sensitivity coefficient.

[0014] Furthermore, based on the draft instruction, resource group identifiers and target power changes are issued. The issuance time and the deadline for receipt are written into the receipt registration form. Resource acceptance receipts are received and a missing receipt label is generated. Resource group identifiers with missing receipt labels are removed from the scheduling plan. Replacement resource group identifiers are selected in the order of the alternative list and the allocation gap is restricted by the effective adjustment boundary of the group. The final scheduling instruction is then issued and executed.

[0015] Furthermore, during execution, the active power measurement sequence of the grid connection point is collected. According to the stable segment discrimination rule, the stable segment before execution and the stable segment after execution are extracted and the mean difference is calculated to obtain the measured active power change. The measured active power change and the active power change of the scheduling target are used to form a deviation record and write the deviation direction and deviation degree level code. The deviation record is written into the response fingerprint and the baseline segment re-estimation and stable band re-estimation are triggered to complete the response fingerprint update.

[0016] The technical effects and advantages of the distributed energy optimization scheduling method based on virtual power plants proposed in this invention are as follows: 1. The effective list uses response fingerprints as the source of evidence and employs the net response energy following ratio and voltage-sensitive stability score to generate admission discrimination markers. It incorporates the continuous following on the execution side and the consistency on the electrical side into the screening process. The effective adjustment boundary of the group is solidified into an executable boundary according to the discrimination result. This allows the resource grouping identifier to complete admission, amplitude limiting, retesting and elimination under the same set of semantics, avoiding misjudgments caused by a single amplitude or voltage change, and improving the accuracy and verifiability of the virtual power plant in identifying the true adjustability of distributed energy.

[0017] 2. Electrical access verification maps the effective list to distribution network nodes and assembles the distribution network calculation model. It obtains node voltage and line current through forward and backward scanning, and then completes the prediction verification and power flow verification by combining the node voltage sensitivity coefficient and the line current sensitivity coefficient. This keeps the electrical access mark and the effective adjustment boundary of the group bound together. The dispatch plan decomposes into a draft instruction and outputs a candidate list within the electrical executable boundary, realizing the coordinated constraint of resource side credibility and grid side security, and reducing the probability of distribution network over-limit risks being exposed in the dispatch implementation stage.

[0018] 3. During the instruction implementation phase, a missing acceptance receipt is marked and a replacement list is triggered based on the resource acceptance receipt. Finally, the scheduling instruction is issued all at once, and the measured active power change is calculated by combining the active power measurement sequence of the grid connection point to generate a deviation record. The deviation record is written into the response fingerprint and the baseline segment re-estimation and stable band re-estimation are performed, so that the response fingerprint is updated automatically with the operation facts. The admission discrimination mark and the effective adjustment boundary of the group converge with the periodic iteration, forming a collaborative mechanism from detection and identification to grid-side verification and then to execution closed-loop verification. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a distributed energy optimization scheduling method based on a virtual power plant according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: Figure 1 This invention presents a distributed energy optimization scheduling method based on a virtual power plant, comprising: S1: Collect the active power of the virtual power plant grid connection point, receive the active power change of the scheduling target, and form the scheduling input for this cycle.

[0022] S2: Issue upward and downward probe power commands according to resource groups, collect active power waveforms and voltage waveforms at grid connection points, and generate response fingerprints for resource groups.

[0023] S3: Generate admission discrimination markers for resource groups based on response fingerprints, and include resource groups into the effective list or retest list according to the admission discrimination markers. The effective adjustment boundaries of resource groups in the effective list are determined simultaneously.

[0024] S4: Map the effective list to the distribution network nodes, perform electrical access verification based on power flow constraints and voltage constraints, issue electrical access tags to resource groups that pass the electrical access verification, bind the electrical access tags to the effective adjustment boundary of the group to generate a scheduling plan and instruction draft, and generate a candidate list at the same time.

[0025] S5: Issue a draft instruction and receive resource acceptance receipts. Resource groups with missing acceptance receipts are removed from the scheduling plan and replaced by the alternative list. The final scheduling instruction is formed and executed. After execution, a deviation record is generated by comparing the active power at the grid connection point with the active power change of the scheduling target. The deviation record is written into the response fingerprint to complete the response fingerprint update.

[0026] The virtual power plant's grid connection point is located at the intersection of distributed energy aggregation and dispatch. The active power at the grid connection point directly reflects the net injection or absorption of aggregated resources on the grid side, while the target active power change reflects the required power change. Grid connection point measurement sequences often contain both disturbance and stationary segments. If the disturbance segment is used as a reference, it will mix the detected response with the actual fluctuations, causing response fingerprint distortion and deviating the admission judgment from the actual executability. However, the grid connection point active power measurement sequence naturally possesses segmentable characteristics. By first unifying the sign direction and then extracting the stationary segment from the measurement sequence, a representative reference power can be obtained and used to express the dispatch input.

[0027] S101. Acquisition of active power measurement sequence at grid connection point.

[0028] The metering direction at the grid connection point may be negative for injection or positive for absorption in different access scenarios. If the power symbol convention is inconsistent, the directional meaning of the active power change of the scheduling target will be reversed, and the upward and downward adjustment of the probe power command will also lose a consistent reference. The active power measurement sequence of the grid connection point is collected, and the timestamp and power value are recorded point by point in the measurement sequence, while the active power change of the scheduling target is received.

[0029] Based on the grid connection point metering direction verification results, the active power measurement sequence at the grid connection point undergoes sign unification processing. This processing either maintains the original sign or reverses the overall sign, ensuring that power injected into the grid always corresponds to the positive direction and power absorbed from the grid always corresponds to the negative direction. This results in a grid connection point active power measurement sequence with consistent signs. The sign-consistent grid connection point active power measurement sequence and the target active power change are expressed within the same directional system, maintaining the stability of the positive and negative meanings of the power change and avoiding misjudgments caused by the cancellation of command and measurement directions.

[0030] S102. Consistency of active power change reception and direction for scheduling target.

[0031] The active power measurement sequence at the grid connection point contains short-term disturbances and load fluctuations. Directly selecting a reference from the original sequence is easily biased by jump points. Once the reference power deviates from the true stable level, the dispatch input will be passively biased. For active power measurement sequences at the grid connection point with consistent signs, sort them by timestamp, take two adjacent sampling points one by one, calculate the absolute value of the difference between adjacent power and divide it by the interval between adjacent timestamps to obtain the power change rate sequence.

[0032] Each term in the power change rate sequence represents how quickly the power at the grid-connected point changes per unit time. The faster the change, the more likely it is to be in a disturbance phase, and the slower the change, the more likely it is to be in a stable phase. The power change rate sequence makes the slope characteristics of the power waveform explicit, and the extraction of the stable phase no longer depends on the magnitude of a single point, but on the trend of change, thus improving noise resistance.

[0033] S103. Setting rules for identifying stable segments within the sampling period.

[0034] The power change rate series exhibits significant variations in distribution across different transformer areas and under varying load conditions. Fixed thresholds are difficult to apply, and using only the median power change rate may lead to overly broad discrimination when overall fluctuations are large, causing stable segments to mistakenly enter disturbed segments. Therefore, the median power change rate is calculated for the power change rate series, along with the lower and upper quartiles of the power change rate. The interquartile range of the power change rate is formed by subtracting the lower quartile from the upper quartile.

[0035] The median rate of power change and the interquartile range of the rate of power change are combined to form a stability discrimination threshold. This threshold is used to distinguish between normal fluctuations and abnormal fluctuations, and it adaptively adjusts according to the dispersion of the sequence itself. The interquartile range is used because it is not sensitive to extreme jump points, and the threshold better represents the normal change level of most sampling points. The stability segment discrimination still has convergence when the overall fluctuation is large.

[0036] S104. Reference power extraction and reference power solidification.

[0037] The stable segment index set needs to satisfy both continuity and representativeness. Discrete point selection will splice together multiple segments with different operating conditions, and the reference power will lose its physical meaning.

[0038] The power change rate sequence is scanned sequentially over time. Sampling points with power change rates not exceeding the stability threshold are marked as candidate stable points, forming multiple continuous intervals on the time axis. Each continuous interval is then constructed into a candidate stable segment index set, with the longest candidate stable segment index set being selected first. If multiple candidate stable segment index sets of the same length exist, the cumulative value of the power change rate within each candidate stable segment is calculated, and the candidate stable segment index set with the smaller cumulative value is prioritized. Continuity ensures that the reference power comes from the same operating state, and the selection of the best cumulative value avoids mistaking slow but continuous drift segments as stable segments. The stable segment index set is more representative of the true stable level.

[0039] S105. Construction and encapsulation of inputs for this cycle.

[0040] Even within the stable section, a small number of residual noise points or measurement glitches may still exist. Directly averaging these glitches can easily lead to biases in the power reference, and the offset in the reference power will be transmitted to the dispatch input. Based on the stable section index set, the power value set corresponding to the stable section is extracted from the active power measurement sequence of grid-connected points with consistent signs, and the median value is used as the reference power. The median value is not sensitive to individual outliers, and the reference power is closer to the center level of the power distribution in the stable section. After the reference power is given by the median value of the stable section, the true level of the grid-connected point on the injection or absorption side is accurately characterized, and the drift of the power reference caused by occasional fluctuations is significantly reduced.

[0041] S106. Output object delivery and consistency verification.

[0042] The scheduling input needs to simultaneously express the current grid-connected power level and the required power change. The absence of either will cause the scheduling objective to lose its reference at the execution level. The scheduling input is formed by combining the baseline power and the target active power change. The baseline power represents the grid-connected power level that is closer to steady state within the sampling period, while the target active power change represents the required power change and its direction.

[0043] After the scheduling input is completed, the scheduling target direction and the power direction of the grid connection point are expressed in the same symbol system. The process of determining the stable segment index set and the reference power is reproducible and auditable. The direction selection and amplitude expression of the probe power command have clear references. The generation of the response fingerprint is no longer affected by the reference drift.

[0044] After symbol unification, the active power measurement sequence at the grid connection point is formed into a consistent active power measurement sequence. The power change rate sequence is used to extract the stability threshold through the median and interquartile range. The set of stable segment indices is selected based on continuity and cumulative value. The reference power is represented by the median of the stable segment to characterize the steady-state level of the grid connection point. The scheduling input consists of the reference power and the active power change of the scheduling target. The above processing constrains two common problems, inconsistency in direction and fluctuation pollution, at the scheduling input construction stage. The scheduling input has a clear physical meaning and consistent directional semantics, providing a reliable power reference for detection response identification.

[0045] After the grid connection point's baseline power and the active power change of the dispatch target are consistently expressed, the virtual power plant enters the detection phase, shifting its focus from the aggregation level to the resource grouping level. Resource groups include both devices with similar control methods and devices with similar connection locations. Short-term detection is used to simultaneously record the execution-side response and electrical-side impact within the same trajectory. In actual operation, the grid connection point power exhibits natural fluctuations, and relying solely on the grid connection point active power measurement sequence is insufficient to distinguish between fluctuations and responses. Voltage changes are also easily affected by external disturbances. Bidirectional detection and recording of response fingerprints can fix the directionality and consistency of responses. However, if the detection commands are not constrained by boundary pruning and time windows, the response fingerprints will be mixed with out-of-bounds and truncation actions, making the evidentiary basis for subsequent access judgment marking unstable.

[0046] S201. Resource group generation.

[0047] Resource grouping adopts a combination rule of access location and control method. Resources with the same access location but different control methods often exhibit different characteristics in power following speed and limiting strategy. Resources with the same control method but different access locations often exhibit different characteristics in voltage sensitivity. Unclear grouping boundaries will superimpose different patterns into a mixed curve.

[0048] Each resource in the resource set reads the resource access location code and the resource control method code. The resource access location code comes from the topology mapping result of the grid connection point, and the resource control method code comes from the access protocol registration information. The resource access location code and the resource control method code are concatenated in a fixed order to form a resource group identifier. Resources with the same resource group identifier are merged into the same resource group, and all resource groups form a resource group set.

[0049] After the resource group set is established, the resource group identifier is registered together with the resource list. The same resource group identifier is used throughout the entire detection process. The resource group remains stable during the scheduling cycle, and the response fingerprint has a clear attribution relationship and retest consistency.

[0050] S202. Determination and issuance of bidirectional detection power commands.

[0051] The power detection command needs to trigger an observable response without changing the operating state of the grid connection point. If the detection amplitude is too large, it is easy to introduce protection and limiting actions. If the detection amplitude is too small, it is easy to be overwhelmed by the natural fluctuations of the grid connection point. The detection amplitude needs to satisfy both the execution side boundary and the operation side constraint.

[0052] Each resource group reads its control range, which includes both the upper and lower adjustable boundaries. It also reads the upper limit of the detection amplitude, derived from the operational strategy table. The upward adjustment of the detection amplitude is the smaller of the upper adjustable boundary and the upper limit; the downward adjustment is the smaller of the lower adjustable boundary and the upper limit. The upward detection power command is in the positive direction and equal to the upward detection amplitude, while the downward detection power command is in the negative direction and equal to the downward detection amplitude. The start and end times of the detection hold segment are determined by the command issuance time and the preset hold duration, which remains constant within the same scheduling cycle. The upward and downward detection power commands each complete one hold segment. A buffer segment for returning to the reference power is inserted between hold segments. The buffer segment ends when the active power measurement sequence at the grid connection point returns to near the reference power and remains continuous. The detection amplitude is trimmed within the directional boundaries, and the detection hold segment is locked by time definition. The response trajectory of bidirectional detection has a symmetrical reference, and the response differences of resource groups in the two directions can be stably exposed.

[0053] S203. Grid connection point waveform acquisition and response fingerprint construction.

[0054] The response fingerprint requires simultaneous recording of the active power waveform and the voltage waveform at the grid connection point, both aligned on the same time axis. Otherwise, a causal correspondence cannot be established between the power response and the voltage response, resulting in temporal misalignment in the fingerprint content. The active power waveform at the grid connection point is extracted from the active power measurement sequence with consistent signs. The extraction window covers all sampling points within the pre-detection stable segment before the start of the probe-hold segment and within the probe-hold segment. The voltage waveform at the grid connection point is extracted from the voltage measurement sequence at the same timestamp. Timestamp alignment uses a unified sampling time axis. Missing sampling points are filled in using linear interpolation of adjacent timestamps. Interpolated points are marked as interpolated sampling points in the record to avoid confusion with the original sampling points. The boundary of the pre-detection stable segment is taken from the continuous index segment of the stable segment index set before the start of the probe-hold segment. The start of the probe-following segment is taken from the start of the probe-hold segment. The end of the probe-following segment is identified through the power change rate sequence. The identification rule uses the moment when the power change rate is continuously lower than the stability discrimination threshold and remains continuous. The start of the probe-stable segment is taken from the end of the probe-following segment, and the end of the probe-stable segment is taken from the end of the probe-hold segment.

[0055] The response fingerprint consists of resource group identifier, detection direction, detection amplitude, start and end time of detection hold segment, active power waveform at grid connection point, voltage waveform at grid connection point, boundary of the stable segment before detection, boundary of the following segment after detection, and boundary of the stable segment after detection. An upward and downward detection power command each generates a response fingerprint, distinguished by the detection direction field. The response fingerprint simultaneously stores the power trajectory and voltage trajectory on the same time axis. Segment boundaries are derived from consistency rules, allowing subsequent admission decisions to directly reference the detection direction and segment boundaries for consistent calculations, avoiding drift caused by repeated segmentation.

[0056] After bidirectional detection is completed, each resource group within the resource group set forms a response fingerprint. The response fingerprint records the detection amplitude and the start and end times of the holding period for both the upward and downward adjustment directions. The active power waveform at the grid connection point and the voltage waveform at the grid connection point are aligned in time. The boundaries of the pre-detection stable segment, the detection following segment, and the detection stable segment are determined according to the power change rate rule. The resource group identifier is used throughout the entire process of resource group generation, detection power command issuance, waveform acquisition, and response fingerprint construction. The response fingerprint has three elements: stability attribution, stability window, and stability segmentation. The resource group execution-side response and electrical-side impact can be mutually verified in the same record, and the calculation basis of the admission discrimination mark is consistent and verifiable. Step S3 can directly calculate the net response energy following ratio and voltage-sensitive stability score and generate the admission discrimination mark without re-segmentation.

[0057] After the grid connection point completes the scheduling input, the resource group performs bidirectional detection and forms a response fingerprint. The response fingerprint records the active power waveform and voltage waveform of the grid connection point on the same time axis, and simultaneously provides the boundary of the stable segment before detection, the boundary of the following segment after detection, and the boundary of the stable segment after detection. Relying solely on the active power waveform of the grid connection point can easily lead to misjudging instantaneous impacts as execution capabilities, and relying solely on the voltage waveform of the grid connection point can easily lead to misjudging external voltage disturbances as electrical effects. The response fingerprint provides both the execution-side trajectory and the electrical-side trajectory. The two types of trajectories are cross-verified under the same resource group identifier, thus providing a verifiable source of evidence for the admission discrimination mark.

[0058] S301. Calculation of net response energy follow-up ratio.

[0059] The active power waveform at the grid connection point often exhibits surges and oscillations during probe direction switching. The magnitude of the surge is not equivalent to continuous following, and the oscillation offset can mask the actual execution deviation. Energy accumulation over time can separate continuous following from short-term surges. The fingerprint of the upward probe direction response corresponding to the resource group identifier is used to read the boundaries of the probe following segment and the probe stable segment, merging the probe following segment and the probe stable segment into a single following window.

[0060] For each sampling point within the following window, the power value of the active power waveform at the grid-connected point with the same sign is subtracted from the reference power to obtain the power offset value. The power offset value is kept in its original sign when the detection direction is adjusted upwards, and reversed when the detection direction is adjusted downwards, resulting in a direction-consistent power offset value. The timestamp interval between adjacent sampling points is calculated point-by-point along the time sequence of the following window. The positive portion of the direction-consistent power offset value is multiplied by the timestamp interval and accumulated to obtain the following cumulative energy. The absolute value of the negative portion of the direction-consistent power offset value is multiplied by the timestamp interval and accumulated to obtain the oscillation cumulative energy. The following cumulative energy minus the oscillation cumulative energy yields the net response energy. If the net response energy is lower than zero, it is treated as zero.

[0061] The expected energy of the instruction is obtained by multiplying the absolute value of the detection amplitude by the detection hold duration. The net response energy is divided by the expected energy of the instruction to obtain the in-direction following ratio. If the in-direction following ratio is less than zero, it is treated as zero; if it is greater than one, it is treated as one. Adjusting the detection direction upwards yields the adjusted in-direction following ratio, and adjusting the detection direction downwards yields the adjusted in-direction following ratio. The absolute value of the difference between the adjusted and adjusted in-direction following ratios forms the bidirectional consistency difference. The operating strategy table provides low and high difference boundaries. When the bidirectional consistency difference is not higher than the low difference boundary, the bidirectional consistency deduction factor is zero; when the bidirectional consistency difference is not lower than the high difference boundary, the bidirectional consistency deduction factor is one; when the bidirectional consistency difference is between the two boundaries, the bidirectional consistency deduction factor increases proportionally.

[0062] The lower of the upward and downward internal following ratios is used as the base value. This base value is multiplied by one and then subtracted from the bidirectional consistency deduction factor to obtain the net response energy following ratio. A net response energy following ratio below zero is treated as zero, and a ratio above one is treated as one. The net response energy following ratio retains the energy contribution of continuous following while removing the energy contribution of offsetting swings. The bidirectional consistency deduction factor converts the asymmetric risks of upward and downward adjustments to the same scale, thus providing a stable characterization of the execution reliability corresponding to the resource group identifier.

[0063] S302. Voltage sensitivity stability score calculation.

[0064] The voltage at the grid connection point is affected by fluctuations in the upstream power grid and the actions of nearby resources. The voltage change in a single instance lacks comparability. Binding voltage changes with power changes to form a voltage slope sample can link command responses to electrical effects.

[0065] For the resource group identifier, the response fingerprint of the upward detection direction is read along the boundary of the following window. The power offset sequence is formed by subtracting the reference power from the power value of the grid-connected point active power waveform with consistent symbols along the following window. A continuous interval where the power offset sequence remains monotonically changing or remains flat is divided into a sub-window, with the endpoints of the sub-windows all falling within the following window. For each sub-window, the voltage value at the endpoint of the grid-connected point voltage waveform and the power value at the endpoint of the grid-connected point active power waveform are read. The voltage difference between the endpoints forms the voltage change, and the power difference between the endpoints forms the power change. Sub-windows with an absolute power change value lower than the power resolution are directly discarded. The voltage change of the remaining sub-windows is divided by the power change to obtain a voltage slope sample. All voltage slope samples form an upward voltage slope group.

[0066] The downward-adjusted detection direction yields a group of downward-adjusted voltage slopes according to the same rules. The upward-adjusted voltage slope group and the downward-adjusted voltage slope group are combined into a total voltage slope set. This total voltage slope set is sorted from smallest to largest value. A sliding bandwidth scan moves the bandwidth interval across the sorted results, with the bandwidth width taken as the voltage-sensitive bandwidth. The bandwidth interval covering the largest number of voltage slope samples is defined as the bidirectional stable band. The hit rate is obtained by dividing the number of samples in the total voltage slope set that fall into the bidirectional stable band by the total number of samples in the total voltage slope set. For each sample in the total voltage slope set that falls outside the bidirectional stable band, the nearest distance from the sample value to the lower and upper boundaries of the bidirectional stable band is calculated. This nearest distance is divided by the voltage-sensitive bandwidth to obtain the normalized deviation. The normalized deviation is accumulated across all out-of-band samples and then divided by the total number of samples in the total voltage slope set to obtain the deviation penalty. The hit rate minus the deviation penalty yields the voltage-sensitive stability score. A voltage-sensitive stability score below zero is treated as zero, and a score above one is treated as one.

[0067] The voltage sensitivity stability score first sorts the voltage slope samples by value and uses a sliding bandwidth scan to find the interval with the densest slope samples. The hit ratio of the slope samples within this interval is used as the main stability term. Then, the slope samples falling outside the interval are converted into deviation penalties based on the shortest distance to the interval boundary and are cumulatively deducted. This results in resource groups with concentrated voltage response patterns and controlled deviations receiving higher scores, while resource groups with scattered voltage response patterns or obvious deviations receiving lower scores. In this way, electrical consistency is expressed in a stable and verifiable manner.

[0068] S303. Admission discrimination mark generation and list output.

[0069] Resource grouping can lead to both insufficient execution-side tracking and inconsistent electrical-side response. A single threshold can easily confuse these two types of risks. Mutual verification and tiered classification can write the source of risk into the admission judgment flag and form an executable boundary. The operating strategy table provides high-following threshold, low-following threshold, stable high threshold, and stable low threshold. For each resource group identifier, the net response energy tracking ratio and voltage-sensitive stability score are read. If the net response energy follow-up ratio is not lower than the follow-up high threshold and the voltage sensitivity stability score is not lower than the stability high threshold, the admission judgment mark is passed.

[0070] When the net response energy follow-up ratio is not lower than the high follow-up threshold and the voltage sensitivity stability score is not lower than the low stability threshold but lower than the high stability threshold, the admission discrimination flag is set to limit the pass.

[0071] When the net response energy follow-up ratio is not lower than the low follow-up threshold and is lower than the high follow-up threshold, and the voltage sensitivity stability score is not lower than the high stability threshold, the admission discrimination mark is retested.

[0072] For the remaining combinations, the admission criteria flag is set to "reject". If the admission criteria flag is set to "pass", the effective adjustment boundary for the group is the upper and lower adjustable boundaries within the resource group control range.

[0073] When the access discrimination mark is used to determine the amplitude limit, the midpoint between the lower boundary and the upper boundary of the bidirectional stable band is defined as the representative slope. When the absolute value of the representative slope is lower than the lower limit of the representative slope, the representative slope is taken according to the lower limit of the representative slope. The allowable range boundary of the grid connection point voltage is provided by the operation strategy table, and the reference voltage is taken as the average value of the grid connection point voltage waveform within the stable segment before the response fingerprint detection. The representative slope and the upward detection direction are combined to determine the voltage change direction. When the voltage change direction points to the upper boundary of the allowable voltage range, the voltage margin is taken as the upper boundary of the allowable voltage range minus the reference voltage. When the voltage change direction points to the lower boundary of the allowable voltage range, the voltage margin is taken as the reference voltage minus the lower boundary of the allowable voltage range. The upward amplitude limit boundary is obtained by dividing the voltage margin by the absolute value of the representative slope. The effective upward amplitude limit boundary is taken as the minimum value of the upward adjustable boundary and the upward amplitude limit boundary.

[0074] The slope and downward detection direction are used to obtain the downward limiting boundary according to the same rule. The effective downward boundary is the minimum value of the adjustable downward boundary and the downward limiting boundary. The effective adjustment boundary of the group is composed of the effective upward boundary and the effective downward boundary. The admission discrimination mark is the resource group identifier that passes or passes the limiting and is written into the effective list. The effective list records the resource group identifier and the effective adjustment boundary of the group. The admission discrimination mark is the resource group identifier that is retested and written into the retest list. The retest trigger reason code is generated according to the net response energy following ratio falling into the middle range and the voltage sensitivity stability score not lower than the stability high threshold. The admission discrimination mark is the resource group identifier that is rejected and is not included in the scheduling of this cycle. The admission discrimination mark puts the mutual verification results of the execution side and the electrical side into the list structure. The effective adjustment boundary of the group solidifies the limiting result into an upper and lower dual boundary expression. The effective list maintains a unique index in the resource group identifier dimension. The electrical access verification and scheduling plan can directly refer to the list records to carry out calculations.

[0075] The net response energy following ratio extracts the degree of continuous following from the active power waveform at the grid connection point; the voltage sensitivity stability score extracts electrical consistency from the voltage waveform at the grid connection point; the admission discrimination label classifies the two types of results into four categories—pass, limited pass, retest, and rejection—through a mutual verification grading method; the effective list carries the effective adjustment boundary of the group; and the retest list carries the retest trigger reason code. Resource group identification runs through the entire process of response fingerprinting, indicator calculation, discrimination grading, and list output. The window boundary comes from the response fingerprint record, the operation path has consistent input boundaries and consistent output objects, and the expression of the executable capability of resource groups in the scheduling phase has stable semantics and verifiable basis.

[0076] In step S3, the virtual power plant has obtained the effective list, which records the resource group identifier and the effective adjustment boundary of the group. Resource groups outside the effective list will not be included in the current cycle scheduling. Distribution network constraints come from the boundary of the node voltage allowable interval and the upper boundary of the line allowable current interval. The constraints are spatially distributed and coupled with each other. Simply relying on the control range of resource groups cannot guarantee electrical safety. Response fingerprints and access discrimination tags solve the problem of whether resource groups are trustworthy. However, the trustworthiness of resource groups is not equivalent to the allowance of the distribution network. Step S4 needs to map the effective list to the distribution network nodes and perform electrical access verification to make the draft instructions form an executable expression within the voltage constraints and power flow constraints.

[0077] S401. Effective List Mapping and Distribution Network Calculation Model Assembly.

[0078] Resource group identifiers are derived from fixed rules based on resource access location codes and resource control method codes. If a resource group identifier does not fall under a distribution network node number, the node voltage and line current cannot be mapped to a specific resource group. The resource group identifiers are read item by item from the effective list, and a mapping table is established based on the resource access location code and the access point number in the distribution network topology. The mapping table assigns a unique distribution network node number to each resource group identifier.

[0079] The distribution network calculation model is assembled using four types of information: distribution network topology, line impedance parameters, transformer parameters, and node load baselines. The model establishes the connection relationship between nodes and lines using the distribution network topology as its framework. Each line in the topology is bound to its line impedance parameters and the upper boundary of its allowable current range. Each transformer is bound to its transformer parameters and tap position or equivalent turns ratio. Each distribution network node in the topology is bound to its node load baseline, which consists of the node's active load baseline and reactive load baseline, and is defined by the same positive and negative direction as the active power at the grid connection point. The model also writes the grid connection point voltage allowable range boundary and maps it to the node voltage allowable range boundary according to the distribution network node mapping table. The model uses the distribution network node number as a unique index, binds resource group identifiers to the corresponding distribution network node number through the distribution network node mapping table, and converts the effective adjustment boundary of the group into the available range of node injected power changes. This forms a set of calculation objects and a set of constraints for forward and backward scanning output of node voltage and line current.

[0080] The node load baseline is obtained from transformer area measurements and branch allocation. The allocation process uses the power level corresponding to the stable segment of the active power measurement sequence at the grid connection point as the total constraint, and maintains consistency with the reference power direction convention in step S1. After completing the mapping table and distribution network calculation model, a unique correspondence is formed between the resource group identifier and the distribution network node number. The node load baseline remains fixed in this cycle, and subsequent electrical verification will not result in object drift or direction drift.

[0081] S402. Extraction of node voltage sensitivity coefficient and line current sensitivity coefficient.

[0082] Distribution networks exhibit different voltage and current responses near the load baseline at different nodes. Directly relying on fixed empirical coefficients can easily lead to overly strict or overly lenient constraints. First, a power flow calculation is performed at the node load baseline using a forward-backward scan method. In the forward phase, branch currents are recursively calculated from the power source side to the load side along the distribution network topology branch direction. In the backward phase, node voltages are recursively calculated in the opposite direction. This iteration continues until the voltage change amplitude between two adjacent iterations does not exceed the voltage convergence threshold, yielding the baseline node voltage and baseline line current. Then, a probe injection power change is selected. This probe injection power change is the power disturbance applied to a single node, with the disturbance direction consistent with the positive direction of power injected into the grid, and the disturbance duration consistent with the probe hold segment length in step S2. The probe injection power change is applied sequentially to each distribution network node number, and the forward-backward scan is performed again to obtain the probe node voltage and probe line current.

[0083] The node voltage sensitivity coefficient is obtained by dividing the difference between the probe node voltage and the baseline node voltage by the change in probed injected power. The line current sensitivity coefficient is obtained by dividing the difference between the probed line current and the baseline line current by the change in probed injected power. The dimensions of the node voltage sensitivity coefficient are voltage per power, and the dimensions of the line current sensitivity coefficient are current per power. When the absolute value of the node voltage sensitivity coefficient is lower than the lower limit of the sensitivity coefficient, the absolute value of the sensitivity coefficient is taken according to the lower limit of the sensitivity coefficient to avoid unreasonable amplification of the subsequent voltage limiting power margin. The sensitivity coefficient is extracted under the same change in probed injected power, and the node voltage constraint and the line current constraint have a consistent disturbance scale, so there will be no inconsistency between the prediction verification and the verification verification.

[0084] S403. Electrical access verification and electrical access mark generation.

[0085] The effective list provides the effective adjustment boundaries for resource groups. These boundaries represent the permissible power variation range for resource groups on the execution side. Distribution network constraints require that power variations, after being mapped to node voltages and line currents, still fall within the permissible range. The effective list is processed item by item, reading the resource group identifier and its effective adjustment boundary. Using a mapping table, the commanded power changes corresponding to the resource group identifiers are aggregated to the distribution network node number. The node-injected power change is the algebraic sum of the commanded power changes for all resource group identifiers mapped to the same distribution network node number. Node voltage sensitivity coefficients are used to predict node voltages. The predicted node voltage is the baseline node voltage plus the node voltage sensitivity coefficient multiplied by the node-injected power change. Line current sensitivity coefficients are used to predict line currents. Line power flow changes are defined according to the branch direction of the distribution network topology. The algebraic sum of the downstream node-injected power changes is used as the line power flow change. The predicted line current is the baseline line current plus the line current sensitivity coefficient multiplied by the line power flow change.

[0086] The boundary of the allowable range of node voltage is read from the operation strategy table according to the distribution network node number, and the upper boundary of the allowable range of line current is read from the equipment ledger according to the line number. When the estimated node voltage falls between the boundaries of the allowable range of node voltage and the estimated line current does not exceed the upper boundary of the allowable range of line current, the estimated verification is deemed to have passed.

[0087] The resource group set verified by the prediction enters the power flow verification. The power flow verification directly injects the commanded power change into the distribution network calculation model and performs a forward and backward scan again. After the scan iterates until the node voltage change does not exceed the voltage convergence threshold, the boundary of the node voltage allowable interval and the upper boundary of the line allowable current interval are verified. The resource group identifier that passes the power flow verification is written into the electrical access tag. The electrical access tag binds the resource group identifier and the effective adjustment boundary of the group. The prediction verification is responsible for quickly filtering out obviously out-of-limit combinations, the power flow verification is responsible for confirming the real feasibility under critical constraints, and the electrical access tag locks the conclusion and boundary to the same resource group identifier, so that the boundary semantics will not be lost in subsequent replacement and acknowledgment processing.

[0088] S404. Generation of scheduling plan instructions: Draft generation of instructions and generation of alternative lists.

[0089] The active power change of the scheduling target comes from step S1. The electrical access mark provides the set of available resource groups and the effective adjustment boundary of the group. The scheduling plan needs to complete the decomposition of aggregated power change within the constraints of node voltage and line current.

[0090] For each item in the electrical access marker set, the resource group identifier and the effective adjustment boundary of the group are read. Simultaneously, the node voltage sensitivity coefficient and the node voltage allowable interval boundary for the corresponding distribution network node number are read. The node voltage margin is determined by the remaining space between the node voltage allowable interval boundary and the baseline node voltage; the direction of this remaining space is determined by the sign of the active power change of the dispatch target. The voltage limiting power margin is obtained by dividing the node voltage margin by the absolute value of the node voltage sensitivity coefficient; if the absolute value of the node voltage sensitivity coefficient is lower than the lower limit, the lower limit is used. The electrical executable boundary is the one whose value does not exceed the other between the effective adjustment boundary of the group and the voltage limiting power margin.

[0091] Resource group identifiers are sorted from highest to lowest according to their electrical executable boundaries, and commanded power changes are allocated sequentially. The allocation rule is that the commanded power change of a single resource group identifier does not exceed the electrical executable boundary, and the algebraic sum of the allocated commanded power changes has the same sign as the dispatch target active power change. Allocation stops when the algebraic sum of the allocated commanded power changes reaches the dispatch target active power change, forming a draft instruction. Resource group identifiers that are not included in the draft instruction but have electrical access markers are added to the candidate list. The candidate list is sorted using both the electrical executable boundary and the line current sensitivity coefficient, prioritizing resource group identifiers with higher electrical executable boundaries and lower absolute values ​​of line current sensitivity coefficients. The draft instruction and the candidate list are output at the resource group identifier level. The draft instruction has a clear executable boundary, and the candidate list has a clear replacement priority. The decomposition process of the dispatch target active power change will not introduce hidden combinations that exceed the constraints of the distribution network.

[0092] After step S4 is completed, the effective list completes the distribution network node mapping and assembles the distribution network calculation model. The node voltage sensitivity coefficient and line current sensitivity coefficient are extracted near the node load baseline and a lower limit for the sensitivity coefficient is set. The electrical access verification first uses the estimated verification to screen out over-limit combinations and then uses the power flow verification to confirm the critical combinations. The electrical access tag is bound to the resource group identifier and the effective adjustment boundary of the group. The dispatch plan decomposes the active power change of the dispatch target within the electrical executable boundary and forms a draft instruction. The candidate list is sorted and output according to the electrical executable boundary and the line current sensitivity coefficient.

[0093] Step S4 has resulted in a draft instruction and a candidate list. The draft instruction satisfies the limits of the allowable voltage range for nodes and the upper limit of the allowable current range for lines within the electrical access verification constraints. The candidate list provides replacement priorities and effective adjustment boundaries for groups. However, communication jitter and rejection at the resource group control terminal may still occur during field operation, and the set of members in the draft instruction may change after issuance. The active power measurement sequence at the grid connection point fluctuates continuously under load fluctuations, making it difficult to determine whether the active power change target for scheduling has been truly implemented based solely on the issued results. Step S5 translates the draft instruction into the final scheduling instruction. It uses resource acceptance receipts to filter out failed groups and fills gaps using the candidate list. Then, it uses the stable segment average method to obtain the measured active power change and writes it into the deviation record and response fingerprint, ensuring that the execution performance of the same resource group identifier is consistent with the electrical performance.

[0094] S501. Instruction Issuance and Receipt Registration.

[0095] The resource group control terminal only returns a resource acceptance receipt after receiving a complete instruction. A lack of receipt registration will result in a mismatch between the receipt and the instruction. The resource group identifier and target power change are read line by line from the instruction draft, generating a distribution record. This record includes the resource group identifier, target power change, and distribution time identifier. The timeout window length is given by the operation strategy table, and the receipt deadline is obtained by adding the distribution time and the timeout window length. The resource group identifier, distribution time, receipt deadline, and initial receipt status value are written into the receipt registration table, with the initial status value set to "not received." After instruction distribution and receipt registration are synchronized, the receipt registration table establishes a unique anchor point for each instruction, clearly defining the attribution of receipt arrivals, fixing timeout judgment boundaries, and ensuring that missing receipts are not overlooked.

[0096] S502. Resource Acceptance Receipt Judgment and Missing Acceptance Receipt Marking.

[0097] When communication links are congested, delayed receipts may occur. Resource-side actions such as bandwidth limiting and protection may result in rejection. The judgment criteria must cover both delayed receipts and non-receipts. The receipt registration table is monitored line by line according to resource group identifiers. Each resource acceptance receipt includes the resource group identifier and acceptance status, and the receipt arrival time is recorded. If the receipt arrival time is earlier than the receipt deadline and the acceptance status is "accepted," the receipt status is updated to "received." If the receipt arrival time is later than the receipt deadline, the receipt status is updated to "delayed," and a missing acceptance receipt marker is added. If no resource acceptance receipt arrives by the receipt deadline, the receipt status is updated to "missing," and a missing acceptance receipt marker is added. Once the missing acceptance receipt marker is formed, the validity boundaries of the instruction draft member set are clear, delayed receipts will not be mistakenly treated as valid execution confirmations, and replacement logic can be completed within a fixed time limit.

[0098] S503. Remove missing groups and fill in the gaps according to the alternative list.

[0099] Retaining resource group identifiers with missing acceptance receipt annotations in the scheduling plan would prevent the achievement of the scheduled active power change target; therefore, they must be removed and supplemented. Match the resource group identifiers to the acceptance receipt registration table, and delete resource group identifiers with missing acceptance receipt annotations from the scheduling plan and draft instructions, resulting in a set of retained resource groups. Algebraically sum the target power changes in the retained resource group set in both positive and negative directions to obtain the allocated aggregated power change. Subtract the allocated aggregated power change from the scheduled active power change to obtain the gap, with the gap sign following the convention for the direction of the scheduled active power change. Scan resource group identifiers sequentially along the candidate list, with filtering criteria including the presence of electrical access markers and resource group identifiers not appearing in the set of retained resource groups. The system reads the effective adjustment boundary of the selected resource group identifier. When the gap is positive, it allocates the target power change in the positive direction. The allocated value does not exceed the upper adjustment boundary of the group's effective adjustment boundary and does not exceed the gap amplitude. When the gap is negative, it allocates the target power change in the negative direction. The allocated value does not exceed the lower adjustment boundary of the group's effective adjustment boundary and does not exceed the gap amplitude. After one allocation, the gap is updated by subtracting the allocated value from the gap. The scanning stops when the gap reaches zero. The gap filling process ensures that the final resource group set completes the aggregated power change compensation within the boundary. The replacement result is consistent with the electrical access mark, and the impact of execution failure is limited to the gap range.

[0100] S504. Final scheduling instruction generation and execution landing data collection.

[0101] Multiple batches of dispatch introduce new timing disturbances, causing segmented jumps in the active power waveform at the grid connection point, making alignment during execution and verification difficult. The reserved resource group set and the replacement resource group set are merged to form the final resource group set. Each resource group identifier and its corresponding target power change are written into the final scheduling instruction, which includes a list of resource group identifiers and a list of target power changes. The final scheduling instruction is dispatched to the resource group control terminal all at once, recording the start and end times of execution. During execution, the active power measurement sequence at the grid connection point is continuously collected, recording timestamps and power values ​​while maintaining consistency with the direction convention of step S1. Dispatching in one go reduces intermediate states, resulting in a power trajectory closer to a continuous response. The stable segments before and after execution are more easily identified by the stable segment discrimination rules.

[0102] S505. Generation of deviation records for measured active power change and update of response fingerprint.

[0103] The active power measurement sequence at the grid connection point exhibits natural fluctuations. Single-point differences may interpret these fluctuations as execution results. The stable segment average method can suppress short-term disturbances. Following the stable segment discrimination rule in step S1, a stable segment before execution is extracted before the execution start time, and a stable segment after execution is extracted after the execution end time. The power average of the stable segment before and after execution is calculated respectively. The measured active power change is obtained by subtracting the power average of the stable segment before execution from the power average of the stable segment after execution. The deviation is obtained by subtracting the target active power change from the measured active power change. The sign of the deviation determines the direction of the deviation, and the magnitude of the deviation is mapped to a level code through a level table. The level table is given by the operation strategy table and uses magnitude intervals to correspond to level codes.

[0104] The deviation direction and degree level are encoded and written into the deviation record, and then the deviation record is written into the response fingerprint. When the deviation direction points to the direction of power injection into the grid, an upward adjustment update of the baseline segment reassessment and the stability band reassessment is triggered; when the deviation direction points to the direction of power absorption, a downward adjustment update of the baseline segment reassessment and the stability band reassessment is triggered. The baseline segment reassessment replaces the pre-detection stability segment boundary in the response fingerprint with the pre-execution stability segment boundary, and recalculates the baseline power and baseline voltage using the pre-execution stability segment mean. The stability band reassessment appends the voltage slope samples formed during this execution period to the original voltage slope sample set according to the timestamp alignment, and then re-extracts the bidirectional stability band according to the stability band extraction rules.

[0105] After the update is completed, the response fingerprint carries the latest deviation information and the latest benchmark information. The historical performance of the resource group identifier is continuously corrected, and the admission judgment mark in the next cycle is closer to the actual operating state.

[0106] After step S5 is completed, the receipt registration form fixes the draft instruction and resource acceptance receipts within a unified time limit boundary. Missing acceptance receipts are flagged to identify failed resource groups and trigger replacement from the alternative list. Finally, the scheduling instruction is issued in a one-time manner. The active power measurement sequence at the grid connection point obtains the measured active power change using the average of the pre-execution and post-execution stable segments. The deviation record encodes the deviation direction and degree level into the response fingerprint. The baseline reassessment and stable band reassessment incorporate the execution facts into the internal structure of the response fingerprint.

[0107] Specifically, the above are merely preferred embodiments of this application and are not intended to limit this application.

[0108] The following high threshold, following low threshold, stable high threshold, and stable low threshold can be pre-calibrated through offline simulation testing, or set to fixed values ​​according to on-site operating procedures.

[0109] In the description of this specification, references to terms such as "an embodiment," "example," and "specific example" indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0110] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A virtual power plant based distributed energy optimization scheduling method, characterized in that, Including the following steps: S1: Collect the active power of the virtual power plant grid connection point, receive the active power change of the scheduling target, and form the scheduling input for this cycle; S2: Issue upward and downward probe power commands according to resource groups, collect active power waveforms and voltage waveforms at grid connection points, and generate response fingerprints for resource groups; S3: Generate admission discrimination markers for resource groups based on response fingerprints, and include resource groups into the effective list or retest list according to the admission discrimination markers. The effective adjustment boundaries of resource groups in the effective list are determined simultaneously. S4: Map the effective list to the distribution network nodes, perform electrical access verification based on power flow constraints and voltage constraints, issue electrical access tags to resource groups that pass the electrical access verification, bind the electrical access tags to the effective adjustment boundary of the group to generate a scheduling plan and instruction draft, and generate a candidate list at the same time; S5: Issue a draft instruction and receive resource acceptance receipts. Resource groups with missing acceptance receipts are removed from the scheduling plan and replaced by the alternative list. The final scheduling instruction is formed and executed. After execution, a deviation record is generated by comparing the active power at the grid connection point with the active power change of the scheduling target. The deviation record is written into the response fingerprint to complete the response fingerprint update. 2.The method of claim 1, wherein, Step S1 includes the following: The active power measurement sequence of the grid-connected point and the active power change of the scheduling target are collected. The active power measurement sequence of the grid-connected point is processed by unifying the signs of the power injected into the grid as positive and the power absorbed from the grid as negative. The power change rate sequence is calculated based on the power difference between adjacent sampling points and the timestamp interval. The stable segment index set is extracted according to the power change rate sequence. The median value of the power value corresponding to the stable segment index set is taken as the reference power. The reference power and the active power change of the scheduling target constitute the scheduling input. 3.The method of claim 2, wherein, Step S2 includes the following: Resource group identifiers are generated based on resource access location coding and resource control method coding. Resources with the same resource group identifier are grouped into resource groups to form a resource group set. The resource group identifier and resource list are registered. The resource group set remains unchanged during the scheduling cycle. The resource group identifier is used throughout the entire process of detection power command issuance and response fingerprint recording.

4. The method of claim 3, wherein, Step S2 also includes the following: Each resource group reads the upper and lower adjustable boundaries and the upper limit of the detection amplitude. The command to increase the detection power is equal to the positive direction of the upward detection amplitude, and the command to decrease the detection power is equal to the negative direction of the downward detection amplitude. The active power waveform and voltage waveform at the grid connection point are extracted according to the start and end time of the detection hold segment and the timestamps are aligned to form a response fingerprint containing the segment boundaries.

5. The method of claim 4, wherein, Step S3 includes the following: For each resource group identifier, the detection and following segment boundary and the detection and stable segment boundary are used to form a following window. The power offset value in the direction of the grid connection point is generated according to the active power waveform relative to the reference power. The positive energy and the reverse energy are accumulated respectively and the difference is calculated to obtain the net response energy. The net response energy is compared with the command expected energy to obtain the following ratio in the upward adjustment direction and the following ratio in the downward adjustment direction. The bidirectional consistent difference is mapped to the bidirectional consistent deduction factor and the deduction is completed to obtain the net response energy following ratio.

6. The method of claim 5, wherein, Step S3 also includes the following: For each resource group identifier, a sub-window is cut according to the power offset change trend within the following window. The voltage slope sample is calculated based on the difference between the endpoints of the grid connection point voltage waveform and the grid connection point active power waveform to form an upward voltage slope group and a downward voltage slope group. These are merged to obtain the total voltage slope set and the bidirectional stability band is extracted. The voltage sensitivity stability score is obtained based on the hit rate of the bidirectional stability band and the deviation penalty. Based on the threshold comparison, an admission discrimination mark is generated and the effective list, retest list and group effective adjustment boundary are output.

7. The method of claim 6, wherein, Step S4 includes the following: The effective list is read to obtain the resource grouping identifier and the effective adjustment boundary of the grouping. The distribution network node mapping table is established according to the resource grouping identifier and the distribution network calculation model is assembled. Based on the node load baseline, forward and backward scanning is performed to obtain the baseline node voltage and baseline line current. The node voltage sensitivity coefficient and line current sensitivity coefficient are extracted node by node according to the uniform detection of injected power change, and a lower limit of the sensitivity coefficient is set for the node voltage sensitivity coefficient.

8. The method of claim 7, wherein, Step S4 also includes the following: The commanded power changes are summarized into node injected power changes according to the distribution network node mapping table, and the node voltage and line current are estimated. The estimation is checked based on the node voltage allowable interval boundary and the upper boundary of the line allowable current interval, and power flow verification is performed to generate electrical access marks. The voltage limiting power margin is calculated by combining the node voltage margin and the node voltage sensitivity coefficient, and the electrical executable boundary is determined with the effective adjustment boundary of the group. The active power change of the scheduling target is decomposed according to the electrical executable boundary to generate the instruction draft, and the candidate list is generated by sorting according to the electrical executable boundary and the line current sensitivity coefficient.

9. A distributed energy optimization scheduling method based on a virtual power plant according to claim 8, characterized in that, Step S5 includes the following: Based on the draft instruction, resource group identifiers and target power changes are issued. The issuance time and the deadline for receipt are written in the receipt registration form. Resource acceptance receipts are received and a missing receipt label is generated. Resource group identifiers with missing receipt labels are removed from the scheduling plan. Replacement resource group identifiers are selected in the order of the alternative list and the allocation gap is restricted by the effective adjustment boundary of the group. The final scheduling instruction is then issued and executed.

10. The method of claim 9, wherein, Step S5 also includes the following: During execution, the active power measurement sequence of the grid connection point is collected. According to the stable segment discrimination rule, the stable segment before execution and the stable segment after execution are extracted and the mean difference is calculated to obtain the measured active power change. The measured active power change and the active power change of the scheduling target are used to form a deviation record and write the deviation direction and deviation degree level code. The deviation record is written into the response fingerprint and the baseline segment re-estimation and stable band re-estimation are triggered to complete the response fingerprint update.

Images