Distributed photovoltaic dynamic output control method and device

Abstract

The application provides a distributed photovoltaic dynamic output control method and device, comprising: obtaining the load change estimation sequence of the transformer area based on the voltage change amount, photovoltaic output change amount and pre-constructed voltage-power sensitivity mapping relationship of the acquired low-voltage bus of the transformer area; fusing the load change estimation sequence with time information and environmental temperature, constructing a user behavior feature vector, clustering the user behavior feature vector, and generating the load image of the transformer area; constructing a multi-objective optimization model based on the load image and real-time operation data of the transformer area, and obtaining the photovoltaic output strategy based on the multi-objective optimization model. The application separates the influence of photovoltaic output fluctuation and load fluctuation from the voltage change, and generates an executable photovoltaic output strategy in combination with the load image, thereby improving the net load identification accuracy and the control accuracy of the distributed photovoltaic output.

Description

Technical Field

[0001] This invention relates to the field of low-voltage power distribution technology, and in particular to a method and device for dynamic output control of distributed photovoltaic power. Background Technology

[0002] In recent years, distributed photovoltaic (PV) installations in low-voltage distribution networks have grown rapidly, forming a widespread pattern of user-side power generation, local consumption, and surplus power fed into the grid. Affected by factors such as rapidly changing weather and cloud formations, uncertain user electricity consumption behavior, and differences in transformer topology and impedance, low-voltage voltage exceeding limits and reverse power flow are becoming more frequent. To improve renewable energy consumption and distribution network operation safety, the grid side typically implements active power limiting, ramp control, or coordinated regulation involving voltage support for distributed PV. Currently, existing distributed PV power limiting schemes usually set fixed output limits and fixed ramp rates based on irradiance prediction results, or passively trigger power limiting after the voltage exceeds a threshold. These schemes struggle to distinguish between voltage changes caused by PV fluctuations and those caused by user load changes, easily leading to excessive power limiting when load is about to increase, or delayed power limiting when surplus power surges into the grid. Another centralized optimization scheme is based on a complete low-voltage topology and full power measurement, but it is highly dependent on topology parameters, communication integrity, and measurement coverage, making it difficult to promote in a large number of low-voltage transformer areas. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a method and apparatus for controlling the dynamic output of distributed photovoltaic power, so as to improve the control accuracy of distributed photovoltaic power output.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a method for dynamic output control of distributed photovoltaic (PV) power, comprising: determining the voltage influence term corresponding to the PV output change based on the obtained voltage changes of the low-voltage busbar and the grid connection point of the distributed PV user, the PV output change, and a pre-constructed voltage-power sensitivity mapping relationship; subtracting the voltage influence term from the voltage change to obtain the load voltage residual caused by the load change, and obtaining the load change estimation sequence of the distribution area based on the load voltage residual and the load power equivalent sensitivity coefficient; fusing the load change estimation sequence with time information and ambient temperature to construct a user behavior feature vector, and clustering the user behavior feature vector to generate a load profile of the distribution area; constructing a multi-objective optimization model based on the load profile, real-time operation data of the distribution area, and the prediction result of the maximum available PV output, and obtaining a PV output strategy based on the multi-objective optimization model; wherein, the PV output strategy includes a time-segmented power upper limit curve, ramp rate, voltage trigger threshold, and backoff rule.

[0005] Optionally, the voltage influence term is subtracted from the voltage change to obtain the load voltage residual caused by the load change. Based on the load voltage residual and the equivalent sensitivity coefficient of the load power, the load change estimation sequence of the transformer area is obtained, including: establishing an initial voltage-power sensitivity mapping relationship based on the actual photovoltaic output, the incoming power of the transformer area, and the low-voltage bus voltage in historical synchronous samples; calculating the voltage influence term corresponding to the photovoltaic output change based on the photovoltaic output sensitivity coefficient of the previous moment and the initial voltage-power sensitivity mapping relationship, and subtracting the voltage influence term from the voltage change to obtain the load voltage residual caused by the load change; updating the equivalent sensitivity coefficient of voltage to load power using a recursive least squares method with a forgetting factor within a sliding window that meets the preset information conditions; and obtaining the load change estimation sequence based on the load voltage residual and the updated equivalent sensitivity coefficient of the load power.

[0006] Optionally, based on the load voltage residual and the updated load power equivalent sensitivity coefficient, a load change estimation sequence is obtained, including: obtaining an initial load change estimation sequence based on the updated load power equivalent sensitivity coefficient and the photovoltaic output change; smoothing the initial load change estimation sequence using a Kalman filter, and limiting the estimated values ​​that exceed the preset physical change boundary to obtain the load change estimation sequence.

[0007] Optionally, a multi-objective optimization model is constructed based on load profiles, real-time operating data of distribution areas, and the predicted maximum available output of photovoltaic power. This includes: determining a net load prediction sub-model based on the profile category of the load profile, and determining the net load prediction distribution based on the net load prediction sub-model and the real-time operating data of the distribution area; constructing a multi-objective optimization model based on the net load prediction distribution, the predicted maximum available output of photovoltaic power, inverter capacity constraints, ramp rate constraints, upper and lower voltage limits of the distribution area, and line current limits; wherein, the multi-objective optimization model takes voltage over-limit risk, curtailment rate, tracking error of the upper-level dispatch power target, and the fairness index of power restriction between distribution areas with the same profile as objective terms, and forms a comprehensive objective function according to preset weights.

[0008] Optionally, a photovoltaic power output strategy is obtained based on a multi-objective optimization model, including: mapping the output of the multi-objective optimization model to a standardized and executable photovoltaic power output strategy; wherein, the backoff rule is: when the measured voltage continuously exceeds the voltage trigger threshold for a preset time, the time-segment power upper limit curve is lowered by a preset percentage. If the voltage still exceeds the voltage trigger threshold after the reduction, the time-segment power upper limit curve is gradually lowered according to the preset backoff step size, the maximum backoff number, and the minimum allowable output limit, until the voltage recovers to the safe range or reaches the minimum allowable output limit.

[0009] Optionally, the photovoltaic output strategy also includes a user-level power limiting object sorting step: when the feeder or phase voltage exceeds the upper limit threshold, identify distributed photovoltaic users that belong to the surplus power grid type profile, calculate the remaining adjustable capacity of each user, and issue power limiting instructions in descending order of remaining adjustable capacity.

[0010] Optionally, the load change estimation sequence can be fused with time information and ambient temperature to construct a user behavior feature vector, including: extracting statistical features of the load change estimation sequence at multiple time scales; and concatenating or weighting the statistical features at each time scale with timestamp encoding, holiday identifiers and / or ambient temperature to obtain the user behavior feature vector.

[0011] Optionally, the photovoltaic power output strategy obtained based on the multi-objective optimization model also includes: based on the current profile category and load change trend of the transformer area, by matching the historical sample library of similar profiles, to make a probability prediction of the net load range for the next period; inputting the probability prediction distribution and confidence interval into the multi-objective optimization model to obtain the photovoltaic power output strategy.

[0012] Secondly, the present invention provides a distributed photovoltaic dynamic output control device, comprising: a voltage influence term determination module, used to determine the voltage influence term corresponding to the photovoltaic output change based on the obtained voltage change of the low-voltage bus of the distribution area and the grid connection point of the distributed photovoltaic user, the photovoltaic output change, and a pre-constructed voltage-power sensitivity mapping relationship; a load change estimation module, used to subtract the voltage influence term from the voltage change to obtain the load voltage residual caused by the load change, and obtain the load change estimation sequence of the distribution area based on the load voltage residual and the load power equivalent sensitivity coefficient; a load profile construction module, used to fuse the load change estimation sequence with time information and ambient temperature to construct a user behavior feature vector, and cluster the user behavior feature vector to generate a load profile of the distribution area; and a strategy generation module, used to construct a multi-objective optimization model based on the load profile, real-time operation data of the distribution area, and the prediction result of the maximum available photovoltaic output, and obtain a photovoltaic output strategy based on the multi-objective optimization model; wherein, the photovoltaic output strategy includes a time-segmented power upper limit curve, ramp rate, voltage trigger threshold, and backoff rule.

[0013] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the steps of the method provided in any of the first aspects above.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs the steps of the method provided in any of the first aspects above.

[0015] This invention brings the following beneficial effects: The distributed photovoltaic dynamic output control method and apparatus provided by this invention firstly determines the voltage influence term corresponding to the photovoltaic output change based on the obtained voltage changes of the low-voltage busbar and the grid connection point of the distributed photovoltaic user, the photovoltaic output change, and the pre-constructed voltage-power sensitivity mapping relationship. Secondly, the voltage influence term is subtracted from the voltage change to obtain the load voltage residual caused by the load change, and the load change estimation sequence of the distribution area is obtained according to the load voltage residual and the load power equivalent sensitivity coefficient. Then, the load change estimation sequence is fused with time information and ambient temperature to construct a user behavior feature vector, and the user behavior feature vector is clustered to generate a load profile of the distribution area. Finally, a multi-objective optimization model is constructed based on the load profile, real-time operation data of the distribution area, and the prediction result of the maximum available photovoltaic output, and a photovoltaic output strategy is obtained based on the multi-objective optimization model. The photovoltaic output strategy includes a time-segmented power upper limit curve, ramp rate, voltage trigger threshold, and backoff rule. In the above method, load changes are inferred by using easily obtainable voltage time-series features and pre-built voltage-power sensitivity mapping relationships, thereby improving the net load identification capability; the load change estimation sequence time information and ambient temperature are fused to construct user behavior feature vectors and generate load profiles for the transformer area. The photovoltaic output strategy is generated using the load profile, which enables the photovoltaic output strategy to dynamically adapt to the profile, thereby improving the control accuracy of distributed photovoltaic output.

[0016] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A flowchart of a distributed photovoltaic dynamic output control method provided in an embodiment of the present invention; Figure 2This is a schematic diagram of the structure of a distributed photovoltaic dynamic output control device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

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

[0021] Currently, the control of distributed photovoltaic power output is usually based on predicting output according to irradiance and setting a fixed upper limit, a fixed ramp rate, or triggering power limitation according to voltage threshold; or by collecting terminal power / voltage data to make state estimation and then centrally optimizing the allocation of photovoltaic power output, which results in insufficient control precision.

[0022] Based on this, the distributed photovoltaic dynamic output control method and device provided in this embodiment of the invention can improve the control accuracy of distributed photovoltaic output.

[0023] To facilitate understanding of this embodiment, a distributed photovoltaic dynamic output control method disclosed in this invention will first be described in detail. This method can be executed by a transformer substation fusion terminal, edge controller, distribution automation terminal, server, or electronic device with data acquisition, calculation, and communication capabilities. See also Figure 1 The flowchart shown illustrates a distributed photovoltaic dynamic output control method, which mainly includes the following steps S101 to S104: Step S101: Based on the obtained voltage changes of the low-voltage busbar in the distribution area and the grid connection point of distributed photovoltaic users, the photovoltaic output changes, and the pre-constructed voltage-power sensitivity mapping relationship, determine the voltage impact item corresponding to the photovoltaic output changes; Step S102: Subtract the voltage influence term from the voltage change to obtain the load voltage residual caused by the load change, and obtain the load change estimation sequence of the transformer area based on the load voltage residual and the load power equivalent sensitivity coefficient.

[0024] In one implementation, voltage sequence V(t) is acquired from the low-voltage busbar / user terminal of the distribution area and sampled according to a preset sampling period. The sampling period can be set to the second or minute level according to the control response requirements. Noise reduction, outlier removal, sliding window interpolation, and timestamp synchronization are performed to form a unified voltage timing segment. Extract multi-scale feature vectors within window W. ,include: Steady-state statistical characteristics: mean, variance, maximum / minimum, quantiles, percentage of time spent exceeding limits; Dynamic characteristics: first-order difference mean, slope distribution, number of peaks, duration of rising / falling edges; Frequency domain / period characteristics: main period energy, harmonic / flicker indicators (if sampling is fast enough); Event trigger characteristics: Event tags such as concentrated rise in the evening, decline at noon, and sharp drop in a short period of time.

[0025] Furthermore, the pre-constructed voltage-power sensitivity mapping relationship is: ΔV≈S PV ·ΔP PV +S LOAD ·ΔP LOAD , of which S PV For the sensitivity of photovoltaic output to voltage, S LOAD Let ΔV be the load change sensitivity to voltage, and ΔP be the voltage change. PV This refers to the real-time output changes uploaded by the photovoltaic inverter. Given the known changes in photovoltaic output, S is updated on a rolling basis. LOAD The load change estimation sequence ΔP was obtained through decoupling calculation. LOAD (t).

[0026] Step S103: Integrate the load change estimation sequence with time information and ambient temperature to construct a user behavior feature vector, and cluster the user behavior feature vector to generate a load profile of the transformer area.

[0027] In one implementation, firstly, statistical features of the load change estimation sequence at multiple time scales are extracted; then, the statistical features at each time scale are concatenated or weighted and fused with timestamp codes, holiday identifiers, and ambient temperature to obtain a user behavior feature vector.

[0028] In practical implementation, statistical characteristics across multiple time scales include: short-term scale: load change rate, peak-to-valley difference, and variance within 1-5 minutes; medium-term scale: main cycle of load fluctuation and center of gravity of energy spectrum within 15-30 minutes; long-term scale: load baseline level, daily cumulative electricity consumption, and maximum continuous rise / fall duration within 2-24 hours. These three scale characteristics are then combined or weighted with one or more of the following: timestamp encoding (hour, weekday), holiday identifiers, and ambient temperature, to form a multi-resolution user behavior feature vector. This user behavior feature vector includes: lifestyle characteristics: morning peak time, evening peak time, peak-to-valley ratio, and nighttime base load level; energy-consuming equipment characteristics: long-term high load caused by electric heating / heat pumps, nighttime step increase caused by charging piles, and stable weekday platform caused by industrial and commercial activities; and response characteristics: response magnitude and lag time to electricity price / demand response events.

[0029] Furthermore, K-means clustering is used to cluster user behavior feature vectors to obtain user profile categories, and the category center parameters and confidence scores are output to form a load profile of the power distribution area. For example: residential evening peak load type, afternoon vacancy type, industrial and commercial stable type, electric heating high fluctuation type, charging pile night charging type, etc.

[0030] Step S104: Construct a multi-objective optimization model based on load profiles, real-time operation data of transformer areas, and prediction results of maximum available photovoltaic output, and obtain photovoltaic output strategies based on the multi-objective optimization model.

[0031] In one implementation, a net load forecasting sub-model is first determined based on the load profile category, and the net load forecasting distribution is determined based on the net load forecasting sub-model and the real-time operation data of the distribution area. Then, a multi-objective optimization model is constructed based on the net load forecasting distribution, the forecast result of the maximum available photovoltaic output, inverter capacity constraints, ramp rate constraints, distribution area voltage upper and lower limits constraints, and line current limit constraints. The multi-objective optimization model takes voltage over-limit risk, curtailment rate, tracking error of the upper-level dispatch power target, and the fairness index of power restriction between distribution areas with the same profile as the objective items, and forms a comprehensive objective function according to preset weights.

[0032] In practical implementation, the corresponding net load prediction sub-model is first selected based on the load profile category. For example, low variance prediction is used for stable profiles, and high confidence intervals / more conservative distributions are used for high-fluctuation profiles. When the confidence level of the load profile is low, a hybrid profile weighted prediction is adopted to avoid strategy mismatch caused by misclassification. Specifically, the ARIMA model is used for stable industrial and commercial profiles, the LSTM model with temporal attention mechanism is used for overnight charging profiles, and a probabilistic prediction model is used to output quantile intervals for high-fluctuation electric heating profiles. When the profile confidence level is <0.7, hybrid profile weighted prediction is enabled, and the weights are determined by the similarity of the profiles of the K adjacent high-confidence areas.

[0033] Then, a multi-objective optimization model is constructed, and a multi-objective optimization strategy for generating photovoltaic power output is adopted. The objective function includes: (1) Voltage compliance: Minimize the risk of voltage exceeding the limit (probability of exceeding the limit / amplitude of exceeding the limit / expected duration of exceeding the limit); (2) Minimum curtailment: Maximize photovoltaic consumption while satisfying voltage constraints, i.e. minimize curtailment rate; (3) Tracking and smoothing: Track the power target of the upper level, suppress rapid fluctuations, and minimize the tracking error of the power target of the upper level; (4) Fairness: Weighting based on capacity or contribution within the same type of portrait platform ensures fair distribution of limited funds, that is, maximizing the fairness of limited funds distribution within the same type of portrait platform.

[0034] The constraints include: (1) Inverter active power constraint: ; (2) Climbing constraint: ; (3) Voltage constraint: To avoid a strong reliance on precise topology and power flow, this embodiment of the invention employs a data-driven correction model, maintaining a lightweight regression model for each transformer area. And updates are triggered by drift detection.

[0035] (4) Communication / Execution Constraints: The strategy is issued in the form of time-segmented power limit curve + ramp rate + triggering conditions, which facilitates inverter execution.

[0036] In one implementation, the output of a multi-objective optimization model is mapped to a standardized, executable photovoltaic (PV) power output strategy. This strategy includes a time-of-use power limit curve, a ramp rate, a voltage trigger threshold, and a rollback rule. The rollback rule is as follows: when the measured voltage continuously exceeds the voltage trigger threshold for a preset time, the time-of-use power limit curve is lowered by a preset percentage. If, after the reduction, the voltage still exceeds the voltage trigger threshold, the time-of-use power limit curve is progressively lowered according to a preset rollback step size, a maximum number of rollbacks, and a minimum allowable output limit, until the voltage recovers to a safe range or reaches the minimum allowable output limit.

[0037] In practice, a photovoltaic power output strategy is issued to the distribution area controller, including: the time-of-use power limit curve Pcap(t), the ramp rate R(t), the voltage trigger threshold, and the backoff rules. This strategy is applicable under certain conditions. The distribution area controller converts the strategy into executable instructions for the inverter (active power setting / power limiting percentage, etc.) and sends back the execution results.

[0038] Furthermore, the power output strategy also includes a user-level power limiting object sorting step: when the feeder or phase voltage exceeds the upper limit threshold, the distributed photovoltaic users belonging to the surplus power grid type profile are identified, the remaining adjustable capacity of each user is calculated, and power limiting instructions are issued in descending order of remaining adjustable capacity.

[0039] The distributed photovoltaic dynamic output control method provided in this embodiment of the invention uses easily obtainable voltage time-series characteristics and pre-built voltage-power sensitivity mapping relationship to infer load changes, thereby improving the net load identification capability; it fuses the load change estimation sequence time information and ambient temperature to construct a user behavior feature vector and generates a load profile of the transformer area, and uses the profile to generate a photovoltaic output strategy, which enables the photovoltaic output strategy to dynamically adapt to the profile, thereby improving the control accuracy of distributed photovoltaic output.

[0040] In one implementation, for the aforementioned step S102, i.e., when subtracting the voltage influence term from the voltage change to obtain the load voltage residual caused by the load change, and obtaining the load change estimation sequence of the transformer area based on the load voltage residual and the load power equivalent sensitivity coefficient, the following methods may be used: First, in the historical synchronous samples, an initial voltage-power sensitivity mapping relationship is established based on the actual photovoltaic output, the incoming power of the transformer area, and the low-voltage bus voltage.

[0041] In practical implementation, the transformer substation is considered as a linearized equivalent circuit, and the relationship between the low-voltage bus voltage V(t) and the total injected power within the substation can be approximated as follows:

[0042] in, This refers to the no-load voltage or the system nominal voltage. R , X The equivalent resistance and reactance of the transformer area; Net reactive power; This represents the net active power of the transformer substation (a positive sign indicates that it is absorbed from the power grid).

[0043] Filter historical data to find those that also indicate actual photovoltaic output. and actual load power The data covers various time periods. Sources of data include: measured active power output uploaded by photovoltaic inverters; total power recorded by smart meters or integrated terminals at the main incoming line of the distribution area, etc. A linear regression method is used to establish the relationship between voltage change and output change: ΔV≈S PV ·ΔP PV +S LOAD ·ΔP LOAD .

[0044] Then, during online operation, the voltage impact term corresponding to the photovoltaic output change is calculated based on the photovoltaic output sensitivity coefficient of the previous moment and the initial voltage-power sensitivity mapping relationship, and the load voltage residual caused by the load change is obtained by subtracting the voltage impact term from the voltage change.

[0045] Next, within a sliding window that meets the preset information conditions, the equivalent sensitivity coefficient of voltage to load power is updated in a rolling manner using a recursive least squares method with a forgetting factor.

[0046] Finally, based on the load voltage residual and the updated load power equivalent sensitivity coefficient, the load change estimation sequence is obtained.

[0047] In practice, for each sampling time k Based on the photovoltaic output sensitivity coefficient of the previous moment Calculate the photovoltaic voltage contribution term The load voltage residual is obtained by subtracting this contribution term from the voltage change ΔV(k). Based on the equivalent sensitivity coefficient of the load power at the previous moment. Calculate the initial load change estimate Within the sliding window, combining historical labeled samples or trusted payload proxy values, [the system] analyzes... S V-L A recursive correction is performed to obtain the updated equivalent sensitivity coefficient of the load power.

[0048] For the forgetting factor, when the detected voltage fluctuation rate of the distribution area exceeds the first threshold (e.g., 0.5% / min) or a change in the load profile category occurs, the forgetting factor is set to a lower value (e.g., 0.95~0.98) to accelerate tracking; when the voltage is stable and the load profile category is stable (e.g., voltage fluctuation <0.1% / min and the profile category is stable for more than 30 minutes), the forgetting factor is set to a higher value (e.g., 0.995~0.999) to suppress measurement noise. In normal mode, the forgetting factor is set to 0.99.

[0049] In one implementation, when obtaining the load change estimation sequence based on the load voltage residual and the updated load power equivalent sensitivity coefficient, the following methods can be used, including but not limited to: First, an initial load change estimation sequence is obtained based on the updated load power equivalent sensitivity coefficient and the change in photovoltaic output; then, a Kalman filter is used to smooth the initial load change estimation sequence, and the estimated values ​​exceeding the preset physical change boundary are amplitude-limited to obtain the final load change estimation sequence. Specifically, the initial load change estimate is used as the observation of the Kalman filter, and the load change is used as the state variable. The smoothed load change estimation sequence is obtained through prediction updates and measurement updates; wherein, the state update process is constrained by a preset maximum load change rate.

[0050] In practice, after obtaining the sensitivity estimate at the current moment, the load change estimate is calculated. The original load change sequence obtained by decoupling is regarded as an observation, and a Kalman filter is used to smooth it. The known fluctuation term of photovoltaic power is removed to obtain the load change estimate sequence.

[0051] In one implementation, the photovoltaic power output strategy obtained based on the multi-objective optimization model further includes: based on the current profile category and load change trend of the transformer area, by matching a historical sample library of similar profiles, to make a probability prediction of the net load range for the next period; and inputting the probability prediction distribution and confidence interval into the multi-objective optimization model to obtain the photovoltaic power output strategy.

[0052] In practice, all historical moments t are selected from the historical database. i satisfy:C ( t i )= C ( t );in C ( t (For the area at any time) t The image categories are recorded, and the current net load for each sample is recorded. P net ( t i ), future Δ T Net load during the period P net ( t i +Δ T ) and trend vector T( t i ).

[0053] For the current moment t Calculate its trend vector T( t The similarity between the trend vector of a sample and each sample in the sample database is calculated using weighted Euclidean distance, and the sample with the smallest distance is selected. K Each sample is designated as a similar neighbor set; for each sample in the similar neighbor set, its future net load value is extracted. The probability density function is established using kernel density estimation:

[0054] Where K is the kernel function (such as the Gaussian kernel). For bandwidth (selected through cross-validation).

[0055] Given confidence level 1 α The prediction interval is: ,in It is the inverse function of the cumulative distribution function. The final output is the probability prediction result of the net load for the next time period: The probability prediction result is input into a multi-objective optimization model to obtain the photovoltaic power output strategy.

[0056] The method provided in this invention links voltage, load, and behavior: it infers load changes through voltage characteristics and further forms a load profile, enabling the control strategy to no longer blindly and passively limit power generation based on voltage changes, but instead to identify net load changes caused by user behavior, thereby significantly improving the accuracy and stability of the output setpoint. Using voltage as the core observation and profile parameters as low-dimensional data transmission targets reduces reliance on low-voltage topologies and full power measurements, making it easier to implement across a wide range of transformer substations. Simultaneously, profile uncertainty is optimized, allowing the strategy to minimize curtailment while ensuring voltage compliance, achieving a quantifiable trade-off for safe power consumption.

[0057] In one implementation, a voltage responsibility period pre-allocation method based on user electricity consumption behavior profiles can be used to achieve photovoltaic power output control, including: (1) Collect the electricity load data, photovoltaic power generation data and daily voltage curve of the grid connection point of each distributed photovoltaic user in the area for thirty consecutive days to form the time-series feature vector of each user; (2) Based on the changing characteristics of the historical intraday load curve, photovoltaic output curve and voltage curve, the daily operation process is divided into the nighttime low period, the morning ramp-up period, the midday photovoltaic peak period, the afternoon power consumption transition period, the evening peak period and the late night period. The time period boundaries can be determined by preset time boundaries or dynamically based on the clustering results of historical curves.

[0058] (3) Based on the ratio of the median electricity load to the median photovoltaic power generation in each time period, users are divided into three categories: self-consumption type, surplus power grid connection type, and balance type.

[0059] (4) When the voltage of a feeder in the distribution area exceeds the upper limit, the users on that feeder who are surplus power connected to the grid are identified first, and the remaining adjustable capacity of these users in the next fifteen minutes is predicted. The remaining adjustable capacity is calculated as the difference between the current photovoltaic output of the user and the median of the historical electricity load during that period.

[0060] Specifically, when the voltage amplitude of a feeder or phase within the distribution area exceeds the set upper limit threshold (i.e., enters an over-limit state), the control device performs the following operations: Based on the current time interval, determine which of the six time periods mentioned above it belongs to; retrieve the user profile type of all distributed photovoltaic users on that feeder within the current time period. Only users marked as having surplus power fed into the grid are selected as the priority control candidate set. For each user in the candidate set, calculate its remaining adjustable capacity at the current moment. The physical meaning of this capacity is: the portion of the power that the user is currently feeding back into the distribution area that exceeds its historical typical load absorption capacity.

[0061] The methods for calculating remaining adjustable capacity include: First, obtain the typical load power value for the current time period from the user's historical data, and record it as the baseline load. The baseline load can be the historical median of the load power during this time period or a smoothed typical value.

[0062] Then, obtain the user's actual photovoltaic power generation at the current moment.

[0063] Finally, the remaining adjustable capacity is calculated according to the following relationship: the remaining adjustable capacity of the i-th user at time t. Determine as follows: ,in For the current photovoltaic power output of the i-th user, This represents the typical load power of the user during the current time period s.

[0064] (5) Select the restricted output targets in descending order of remaining adjustable capacity, and gradually reduce the target output according to the preset step size β; the target output of the i-th user is β is increased stepwise from 0.05, 0.10, and 0.15 until the voltage returns to the safe range, the user's remaining adjustable capacity is exhausted, or the preset maximum limit ratio is reached.

[0065] Specifically, first, the remaining adjustable capacity of all users in the candidate set is sorted in descending order. Users with larger remaining adjustable capacity contribute more to the voltage of the distribution area during the current period, and the impact of reducing their capacity on their own power consumption is smaller. Then, an initial power limiting step size factor is set, which is a positive number less than 1 (e.g., 0.05, representing a 5% reduction in the current remaining adjustable capacity each time). A voltage recovery judgment window is also set, typically one or two sampling periods. Next, power limiting commands are issued sequentially according to the sorted user order. For the currently selected user, the target output value is calculated as follows: Target output = Current photovoltaic power β × Remaining adjustable capacity Here, β is the power limiting step size factor for this execution. In actual execution, β can start from a small initial value and gradually increase.

[0066] After executing a power limiting command for each user, a short delay window (e.g., 5 to 10 seconds to avoid communication delays and inverter response time) is established, followed by a re-acquisition of the feeder's voltage amplitude. It is then determined whether the current voltage has fallen back to a safe range (e.g., below 98% of the upper threshold). If it has fallen back to a safe range, the current control process terminates, and no further commands are issued to subsequent users. If the voltage has not yet fallen back to a safe range, the following rules continue: If the current user still has adjustable capacity remaining after executing this step (i.e., the target output is still greater than the baseline load), then increase the β value for the same user (e.g., from 0.05 to 0.10), recalculate the new target output, and issue it.

[0067] If the remaining adjustable capacity of the current user has been completely reduced (i.e., the target output is equal to or less than the baseline load), then move to the next user in sorted order and start execution from the initial step size.

[0068] To prevent user electricity consumption and photovoltaic (PV) power generation behaviors from becoming ineffective due to seasonal variations or long-term drift, the control device needs to update the user profiles regularly: Every day at midnight (when PV output is zero), the previous day's measured data is added to the historical dataset, while the earliest data day is removed, maintaining a total sliding window length of thirty days. The time period boundaries, average load power, and average PV power for each user are recalculated, and the user profile classification is updated. Alternatively, the time period boundaries can be re-clustered monthly to adapt to changes in sunshine duration and load patterns caused by seasonal variations.

[0069] The method provided in this embodiment of the invention, through user electricity consumption behavior profiling and remaining adjustable capacity calculation, achieves accurate location of voltage over-limit responsibility, accurate selection of power-limiting objects, and accurate calculation of power-limiting amplitude without requiring transformer area topology parameters. While solving the voltage over-limit problem, it significantly reduces photovoltaic power generation losses, improves control fairness, and reduces the number of equipment operations.

[0070] In addition to the distributed photovoltaic dynamic output control method provided in the foregoing embodiments, this invention also provides a distributed photovoltaic dynamic output control device, see [link to relevant documentation]. Figure 2 The schematic diagram shown illustrates the structure of a distributed photovoltaic dynamic output control device, indicating that the device mainly comprises the following parts: The voltage impact term determination module 201 is used to determine the voltage impact term corresponding to the photovoltaic output change based on the obtained voltage change of the low-voltage busbar in the distribution area and the grid connection point of the distributed photovoltaic user, the photovoltaic output change, and the pre-constructed voltage-power sensitivity mapping relationship.

[0071] The load change estimation module 202 is used to subtract the voltage influence term from the voltage change to obtain the load voltage residual caused by the load change, and to obtain the load change estimation sequence of the transformer area based on the load voltage residual and the load power equivalent sensitivity coefficient.

[0072] The load profile construction module 203 is used to fuse the load change estimation sequence with time information and ambient temperature to construct user behavior feature vectors, and to cluster the user behavior feature vectors to generate load profiles for the transformer area.

[0073] The strategy generation module 204 is used to construct a multi-objective optimization model based on load profiles, real-time operation data of the transformer area, and prediction results of the maximum available output of photovoltaic power, and to obtain photovoltaic power output strategies based on the multi-objective optimization model. The photovoltaic power output strategies include time-of-use power upper limit curves, ramp-up rates, voltage trigger thresholds, and rollback rules.

[0074] The distributed photovoltaic dynamic output control device provided in this embodiment of the invention uses easily obtainable voltage timing features and a pre-built voltage-power sensitivity mapping relationship to infer load changes, thereby improving the net load identification capability; it fuses the load change estimation sequence time information and ambient temperature to construct a user behavior feature vector and generates a load profile of the transformer area; it uses the profile to generate a photovoltaic output strategy, which enables the photovoltaic output strategy to dynamically adapt to the profile, thereby improving the control accuracy of distributed photovoltaic output.

[0075] In one embodiment, the load change estimation module 202 is specifically used for: establishing an initial voltage-power sensitivity mapping relationship based on the actual photovoltaic output, the incoming power of the transformer area, and the low-voltage bus voltage in historical synchronous samples; during online operation, calculating the voltage impact term corresponding to the photovoltaic output change based on the photovoltaic output sensitivity coefficient of the previous moment and the initial voltage-power sensitivity mapping relationship, and obtaining the load voltage residual caused by the load change by subtracting the voltage impact term from the voltage change amount; within a sliding window that meets the preset information conditions, using a recursive least squares method with a forgetting factor to continuously update the equivalent sensitivity coefficient of voltage to load power; and obtaining the load change estimation sequence based on the load voltage residual and the updated equivalent sensitivity coefficient of load power.

[0076] In one embodiment, the load change estimation module 202 is specifically used to: obtain an initial load change estimation sequence based on the updated load power equivalent sensitivity coefficient and photovoltaic output change; smooth the initial load change estimation sequence using a Kalman filter, and limit the estimated values ​​that exceed the preset physical change boundary to obtain the load change estimation sequence.

[0077] In one implementation, the strategy generation module 204 is specifically used to: determine the net load forecast sub-model based on the profile category of the load profile, and determine the net load forecast distribution based on the net load forecast sub-model and the real-time operation data of the distribution area; construct a multi-objective optimization model based on the net load forecast distribution, the forecast result of the maximum available output of photovoltaic power, inverter capacity constraints, ramp rate constraints, upper and lower limits of distribution area voltage constraints and line current limits; wherein, the multi-objective optimization model takes voltage over-limit risk, curtailment rate, tracking error of the upper-level dispatch power target and the fairness index of power restriction between distribution areas with the same profile as the objective items, and forms a comprehensive objective function according to preset weights.

[0078] In one implementation, the strategy generation module 204 is specifically used to: map the output of the multi-objective optimization model to a standardized and executable photovoltaic power output strategy; wherein, the photovoltaic power output strategy includes: a time-sharing power upper limit curve, a ramp rate, a voltage trigger threshold, and a backoff rule; the backoff rule is: when the measured voltage continuously exceeds the voltage trigger threshold for a preset time, the time-sharing power upper limit curve is lowered by a preset percentage; if the voltage still exceeds the voltage trigger threshold after the reduction, the time-sharing power upper limit curve is gradually lowered according to the preset backoff step size, the maximum number of backoffs, and the minimum allowable output limit, until the voltage recovers to the safe range or reaches the minimum allowable output limit.

[0079] In one implementation, the photovoltaic output strategy further includes a user-level power limiting object sorting step: when the feeder or phase voltage exceeds the upper limit threshold, the distributed photovoltaic users belonging to the surplus power grid type profile are identified, the remaining adjustable capacity of each user is calculated, and power limiting instructions are issued in descending order of remaining adjustable capacity.

[0080] In one implementation, the load profile construction module 203 is specifically used to: extract statistical features of the load change estimation sequence at multiple time scales; and concatenate or weightedly fuse the statistical features of each time scale with timestamp codes, holiday identifiers, and / or ambient temperature to obtain a user behavior feature vector.

[0081] In one implementation, the strategy generation module 204 is further configured to: based on the current profile category and load change trend of the transformer area, make a probability prediction of the net load range for the next period by matching it with a historical sample library of similar profiles; and input the probability prediction distribution and confidence interval into a multi-objective optimization model to obtain a photovoltaic power output strategy.

[0082] It should be noted that the device provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment. The specific numerical values ​​provided in the implementation of this invention are merely examples and are not intended to limit the scope of the invention.

[0083] This invention also provides an electronic device, specifically, the electronic device includes a processor and a storage device; the storage device stores a computer program, and the computer program, when run by the processor, executes the method described in any of the above embodiments.

[0084] Figure 3This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 100 includes: a processor 30, a memory 31, a bus 32 and a communication interface 33. The processor 30, the communication interface 33 and the memory 31 are connected through the bus 32. The processor 30 is used to execute executable modules, such as computer programs, stored in the memory 31.

[0085] The memory 31 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 33 (which may be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0086] Bus 32 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0087] The memory 31 is used to store programs. After receiving an execution instruction, the processor 30 executes the program. The steps described in any of the foregoing method embodiments can be executed by the processor 30, or implemented by the processor 30 in combination with the program instructions stored in the memory 31.

[0088] Processor 30 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 30 or by instructions in software form. Processor 30 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 31. The processor 30 reads the information in memory 31 and, in conjunction with its hardware, completes the steps of the above method.

[0089] The computer program product of the readable storage medium provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the foregoing method embodiments. For specific implementation, please refer to the foregoing method embodiments, which will not be repeated here.

[0090] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion 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 described in 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.

[0091] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for dynamic output control of distributed photovoltaic power, characterized in that, include: Based on the obtained voltage changes of the low-voltage busbar in the distribution area and the grid connection point of distributed photovoltaic users, the photovoltaic output changes, and the pre-constructed voltage-power sensitivity mapping relationship, the voltage influence item corresponding to the photovoltaic output changes is determined. Subtracting the voltage influence term from the voltage change yields the load voltage residual caused by the load change. Based on the load voltage residual and the load power equivalent sensitivity coefficient, the load change estimation sequence for the transformer area is obtained. The load change estimation sequence is fused with time information and ambient temperature to construct a user behavior feature vector, and the user behavior feature vector is clustered to generate a load profile of the transformer area. A multi-objective optimization model is constructed based on the load profile, real-time operation data of the transformer area, and the prediction results of the maximum available photovoltaic output. A photovoltaic output strategy is obtained based on the multi-objective optimization model. The photovoltaic output strategy includes time-of-use power limit curves, ramp-up rate, voltage trigger threshold, and rollback rules.

2. The method according to claim 1, characterized in that, Subtracting the voltage effect term from the voltage change yields the load voltage residual caused by the load change. Based on the load voltage residual and the load power equivalent sensitivity coefficient, the estimated load change sequence for the transformer area is obtained, including: In the historical synchronous samples, an initial voltage-power sensitivity mapping relationship is established based on the actual photovoltaic output, the incoming power of the transformer area, and the low-voltage bus voltage. During online operation, the voltage impact term corresponding to the photovoltaic output change is calculated based on the photovoltaic output sensitivity coefficient of the previous moment and the initial voltage-power sensitivity mapping relationship, and the load voltage residual caused by the load change is obtained by subtracting the voltage impact term from the voltage change. Within a sliding window that meets the preset information conditions, the equivalent sensitivity coefficient of voltage to load power is updated on a rolling basis using a recursive least squares method with a forgetting factor. Based on the load voltage residual and the updated load power equivalent sensitivity coefficient, a load change estimation sequence is obtained.

3. The method according to claim 2, characterized in that, Based on the load voltage residual and the updated load power equivalent sensitivity coefficient, a load change estimation sequence is obtained, including: Based on the updated load power equivalent sensitivity coefficient and the photovoltaic output change, an initial load change estimation sequence is obtained; The initial load change estimation sequence is smoothed using a Kalman filter, and the estimated values ​​exceeding the preset physical change boundary are amplitude-limited to obtain the load change estimation sequence.

4. The method according to claim 1, characterized in that, Based on the load profile, real-time operating data of the transformer substation, and the predicted maximum available photovoltaic output, a multi-objective optimization model is constructed, including: Based on the image category of the load profile, a net load prediction sub-model is determined, and based on the net load prediction sub-model and the real-time operation data of the transformer area, the net load prediction distribution is determined. Based on the predicted net load distribution, the predicted maximum available photovoltaic output, inverter capacity constraints, ramp rate constraints, voltage upper and lower limits constraints of the distribution area, and line current limits constraints, a multi-objective optimization model is constructed. The multi-objective optimization model takes voltage over-limit risk, curtailment rate, tracking error of the upper-level dispatch power target, and the fairness index of power restriction between similar distribution areas as objective terms, and forms a comprehensive objective function according to preset weights.

5. The method according to claim 1, characterized in that, The photovoltaic power output strategy obtained based on the multi-objective optimization model includes: The output of the multi-objective optimization model is mapped to a standardized and executable photovoltaic power output strategy. The backoff rule is as follows: when the measured voltage continuously exceeds the voltage trigger threshold for a preset time, the time-segmented power upper limit curve is lowered by a preset percentage. If the voltage still exceeds the voltage trigger threshold after the reduction, the time-segmented power upper limit curve is gradually lowered according to the preset backoff step size, maximum backoff number, and minimum allowable output limit until the voltage recovers to the safe range or reaches the minimum allowable output limit.

6. The method according to claim 1, characterized in that, The photovoltaic output strategy also includes a user-level power limiting object sorting step: when the feeder or phase voltage exceeds the upper limit threshold, the distributed photovoltaic users belonging to the surplus power grid type profile are identified, the remaining adjustable capacity of each user is calculated, and power limiting instructions are issued in descending order of remaining adjustable capacity.

7. The method according to claim 1, characterized in that, The load change estimation sequence is fused with time information and ambient temperature to construct a user behavior feature vector, including: Extract statistical features of the estimated load change sequence across multiple time scales; The statistical characteristics of each time scale are concatenated or weighted with timestamp codes, holiday identifiers and / or ambient temperature to obtain the user behavior feature vector.

8. The method according to claim 1, characterized in that, The photovoltaic power output strategy derived from the multi-objective optimization model also includes: Based on the current profile category and load change trend of the transformer area, the net load range for the next period is probabilistically predicted by matching it with a historical database of similar profile samples. The probability prediction distribution and confidence interval are input into the multi-objective optimization model to obtain the photovoltaic power output strategy.

9. A distributed photovoltaic dynamic output control device, characterized in that, include: The voltage impact term determination module is used to determine the voltage impact term corresponding to the photovoltaic output change based on the obtained voltage change of the low-voltage busbar in the distribution area and the grid connection point of the distributed photovoltaic user, the photovoltaic output change, and the pre-constructed voltage-power sensitivity mapping relationship. The load change estimation module is used to subtract the voltage influence term from the voltage change to obtain the load voltage residual caused by the load change, and to obtain the load change estimation sequence of the transformer area based on the load voltage residual and the load power equivalent sensitivity coefficient. The load profile construction module is used to fuse the load change estimation sequence with time information and ambient temperature to construct a user behavior feature vector, and to cluster the user behavior feature vector to generate a load profile of the transformer area. The strategy generation module is used to construct a multi-objective optimization model based on the load profile, real-time operation data of the transformer area, and the prediction results of the maximum available output of photovoltaic power, and to obtain a photovoltaic power output strategy based on the multi-objective optimization model; wherein, the photovoltaic power output strategy includes time-of-use power upper limit curves, ramp rate, voltage trigger threshold, and rollback rules.

10. An electronic device, characterized in that, The method includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the steps of the method according to any one of claims 1 to 8.

Images