Millisecond-level prediction suppression method and device for reverse power flow in light storage and charging collaborative scenario

Abstract

The application provides a millisecond-level prediction suppression method and equipment for reverse power flow in a light storage and charging collaborative scenario, and relates to the technical field of smart grids. The method comprises the following steps: collecting in real time through an energy management controller; predicting a grid connection point according to multi-source operation data flow in combination with the light storage and charging collaborative scenario; presetting a reverse power flow threshold value, comparing the reverse power flow threshold value with a grid connection point net power trend; grading according to a reverse power flow risk index, formulating a collaborative suppression decomposition scheme; and respectively issuing power control instructions to perform millisecond-level prediction suppression of the reverse power flow. The application solves the technical problem in the prior art that only real-time power threshold values are relied on for triggering judgment, there is a lack of advanced perception of the grid connection point power change trend, protection actions can only be performed after the reverse power flow occurs, there is control lag, and the safety and stability of power grid operation are affected. The application changes reverse power flow control from post-disposal to proactive prevention and collaborative optimization, and ensures the stability of power grid operation.

Description

Technical Field

[0001] This application relates to the field of smart grid technology, specifically to a millisecond-level prediction and suppression method and device for reverse power flow in a photovoltaic-storage-charging collaborative scenario. Background Technology

[0002] Against the backdrop of dual-carbon and electricity substitution, industrial and commercial users are widely deploying distributed photovoltaic (PV), energy storage, and charging facilities. Due to the influence of weather on PV output and user behavior on charging load, the net load within the grid fluctuates significantly. For substations with zero backfeeding requirements, when PV output exceeds the substation's absorption capacity, reverse power flow can easily form at the PCC (Power Distribution Center), triggering grid connection alarms, protection actions, or performance penalties. Existing methods employ simple timing control, forcing energy storage charging during peak PV generation periods to attempt to avoid power backfeeding. This relies on comparing real-time sampled values ​​with fixed thresholds, and the control logic is intuitive and easy to implement, generally meeting anti-reverse current requirements under stable sunlight and load conditions. However, these existing technologies rely solely on the current measured power value for judgment, lacking proactive awareness of power change trends at the grid connection point. This results in protection actions being executed only after the reverse power flow actually occurs, creating a control lag window of hundreds of milliseconds to several seconds. During this period, electricity has already been fed back to the grid, potentially causing grid connection alarms, malfunctions of protection devices, and voltage exceeding limits in distribution areas, among other grid safety and stability issues.

[0003] In summary, existing technologies suffer from a technical problem: they rely solely on real-time power thresholds for triggering and judgment, lacking proactive perception of power change trends at the grid connection point. This results in protection actions being executed only after reverse power flow occurs, leading to control lag and further impacting the safety and stability of the power grid. Summary of the Invention

[0004] The purpose of this application is to provide a millisecond-level prediction and suppression method and device for reverse power flow in a photovoltaic-storage-charging collaborative scenario, in order to solve the technical problem in the prior art that, due to the reliance on real-time power thresholds for trigger judgment, there is a lack of advance perception of the power change trend at the grid connection point, resulting in protection actions being executed only after reverse power flow occurs, which leads to control lag and further affects the safety and stability of the power grid.

[0005] To achieve the above objectives, this application provides a millisecond-level prediction and suppression method and device for reverse power flow in a photovoltaic-storage-charging coordinated scenario.

[0006] Firstly, this application provides a millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario. This method is implemented using a millisecond-level prediction and suppression device for reverse power flow in a photovoltaic-storage-charging collaborative scenario. The method includes: real-time data acquisition via an energy management controller to obtain multi-source operating data streams; prediction of the grid connection point based on the multi-source operating data streams and the photovoltaic-storage-charging collaborative scenario to obtain the grid connection point's net power trend and prediction uncertainty index; pre-setting a reverse power flow threshold; comparing the grid connection point's net power trend with the reverse power flow threshold to calculate a reverse power flow risk index; classifying the risk level according to the reverse power flow risk index, determining multiple control levels for optimization, and formulating a collaborative suppression decomposition scheme; executing the collaborative suppression decomposition scheme and issuing power control commands to perform millisecond-level prediction and suppression of reverse power flow.

[0007] Optionally, the energy management controller initiates a polling process with the grid-connected metering unit according to a first sampling period to obtain a first polling dataset; the energy management controller initiates a polling process with the photovoltaic inverter unit according to a second sampling period to obtain a second polling dataset; the energy management controller initiates a polling process with the battery management system and energy storage converter of the energy storage unit according to a third sampling period to obtain a third polling dataset; the energy management controller initiates a polling process with the station-level charging controller of the charging pile group according to a fourth sampling period to obtain a fourth polling dataset; the energy management controller timestamps the first, second, third, and fourth polling datasets obtained in the first, second, third, and fourth sampling periods to generate multiple labeled datasets; the data update time is extracted, and the data update time is used as a reference to perform interpolation and alignment processing on the multiple labeled datasets to generate the multi-source operating data stream.

[0008] Optionally, the output power records of the photovoltaic inverter units from the multi-source operation data stream are retrieved to construct a photovoltaic output history sequence; adjacent sampling is performed based on the photovoltaic output history sequence to extract multiple adjacent sampling points, and power difference calculation is performed based on the multiple adjacent sampling points to obtain multiple power change step sizes; the average power change step size is calculated by averaging the multiple power change step sizes to obtain the average power change step size, and the average power change step size is used as a ramp-up change feature; the measured photovoltaic output value is introduced and combined with the ramp-up change feature size to be synchronized to a short-time series prediction model for short-time prediction to generate a photovoltaic output prediction sequence; charging load prediction is performed based on the multi-source operation data stream to construct a charging load prediction sequence; the photovoltaic output prediction sequence and the charging load prediction sequence are superimposed on the time axis to generate the net power trend of the grid connection point.

[0009] Optionally, multiple photovoltaic output prediction values ​​are extracted by traversing the photovoltaic output prediction sequence according to time sampling points, and the multiple photovoltaic output prediction values ​​correspond to the time sampling points; multiple charging load prediction values ​​are extracted by traversing the charging load prediction sequence according to the time sampling points, and the multiple charging load prediction values ​​correspond to the time sampling points; based on the time sampling points, the multiple photovoltaic output prediction values ​​and the multiple charging load prediction values ​​are subtracted to obtain the grid-connected active power prediction value at the time sampling point; the grid-connected active power prediction values ​​are sorted according to time sequence to construct a power prediction sequence; based on the power prediction sequence, data change identification is performed to determine multiple data identification points, and the multiple data identification points are connected according to the time sampling points to construct the grid-connected net power trend.

[0010] Optionally, the grid-connected point is continuously and in real-time collected by the grid-connected metering unit to obtain the measured value of the active power at the grid-connected point; the predicted value of the active power at the grid-connected point is compared with the measured value of the active power at the grid-connected point to calculate multiple prediction deviations; extreme value analysis is performed based on the multiple prediction deviations to extract the maximum and minimum deviation values; boundary analysis is performed based on the maximum and minimum deviation values ​​to set the width of the deviation distribution interval; the rated power at the grid-connected point is obtained, and the ratio of the width of the deviation distribution interval to the rated power at the grid-connected point is calculated as the baseline uncertainty; fluctuation analysis is performed based on the photovoltaic-storage-charging collaborative scenario to obtain multiple fluctuation factors; the baseline uncertainty is dynamically corrected according to the multiple fluctuation factors to construct a prediction uncertainty index.

[0011] Optionally, the reverse power transmission allowable limit is read, and the reverse power transmission allowable limit is converted according to the dimension of the grid-connected active power prediction value to determine the reverse power flow threshold; based on the grid-connected point active power prediction value and the reverse power flow threshold, reverse over-limit judgment is performed one by one, and multiple over-limit amplitudes are recorded; the maximum value of the multiple over-limit amplitudes is extracted as the peak over-limit amount, and the arithmetic mean of the multiple over-limit amplitudes is calculated as the average over-limit amount; the peak over-limit amount and the average over-limit amount are used for reverse power flow risk analysis to construct a first risk component; based on the grid-connected point active power sequence, adjacent point difference operation is performed to obtain the instantaneous change rate of grid-connected active power, and the instantaneous change rate is used for power impact intensity analysis to construct a second risk component; based on the prediction uncertainty index, prediction credibility calculation is performed, and risk assessment impact analysis is performed according to the prediction credibility to construct a third risk component; the first risk component, the second risk component, and the third risk component are weighted and summed to obtain the reverse power flow risk index.

[0012] Optionally, based on the energy management controller storing a first risk threshold and a second risk threshold, where the first risk threshold is less than the second risk threshold; the reverse current risk index is compared with the first risk threshold. If the reverse current risk index is less than the first risk threshold, it is determined to be a mild risk level, and a first control level is activated; if the reverse current risk index is greater than or equal to the first risk threshold, the reverse current risk index is compared with the second risk threshold. If the reverse current risk index is less than the second risk threshold, it is determined to be a moderate risk level, and a second control level is activated; if the reverse current risk index is greater than or equal to the second risk threshold, it is determined to be a severe risk level, and a third control level is activated.

[0013] Optionally, when the first control level is activated, the equipment combination is determined to be only calling the energy storage unit for energy storage regulation optimization, obtaining the energy storage regulation power component; constraints are solved based on the energy storage regulation power component to generate a reverse power flow suppression task; when the second control level is activated, the equipment combination is determined to be simultaneously calling the energy storage unit and the charging pile group for group energy storage regulation optimization, obtaining the charging pile group regulation power component; constraints are solved based on the charging pile group regulation power component to generate a charging pile group load component regulation task; when the third control level is activated, the equipment combination is determined to be simultaneously calling the energy storage unit, the charging pile group, and the photovoltaic inverter unit for photovoltaic derating regulation optimization, obtaining the photovoltaic derating power component; constraints are solved based on the photovoltaic derating power component to generate an output derating task; the reverse power flow suppression task, the charging pile group load component regulation task, and the output derating task are integrated to formulate the collaborative suppression decomposition scheme.

[0014] Optionally, the energy storage regulation power component of the reverse power flow suppression task is read and algebraically summed with the actual charging and discharging power setpoint of the energy storage converter to generate a first power control command; the pile group regulation power component of the pile group load component regulation task is read and decomposed into charging power reduction to obtain a first power limit value; a verification is performed based on the first power limit value, and a second power control command is generated when the verification passes; the photovoltaic derating power component of the output derating task is read, and a maximum available output power is extracted based on the photovoltaic inverter unit; the difference between the maximum available output power and the photovoltaic derating power component is calculated to generate a second power limit value; the operating point is tracked according to the second power limit value to generate a third power control command; the first power control command, the second power control command, and the third power control command are issued in parallel for millisecond-level response collaborative prediction and suppression.

[0015] Secondly, this application also provides a millisecond-level prediction and suppression device for reverse power flow in a photovoltaic-storage-charging collaborative scenario, used to execute the millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario as described in the first aspect. The millisecond-level prediction and suppression device for reverse power flow in a photovoltaic-storage-charging collaborative scenario includes: a data acquisition module for real-time acquisition via an energy management controller to obtain multi-source operating data streams; a trend prediction module for predicting the grid connection point based on the multi-source operating data streams and the photovoltaic-storage-charging collaborative scenario, obtaining the grid connection point net power trend and prediction uncertainty index; a risk judgment module for presetting a reverse power flow threshold, comparing the grid connection point net power trend with the reverse power flow threshold, and calculating a reverse power flow risk index; a level classification module for classifying levels according to the reverse power flow risk index, determining multiple control levels for optimization, and formulating a collaborative suppression decomposition scheme; and a millisecond-level parallel execution module for executing the collaborative suppression decomposition scheme and issuing power control commands to perform millisecond-level prediction and suppression of reverse power flow.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: by introducing net power trend prediction and prediction uncertainty index at the grid connection point to calculate the reverse power flow risk index, and classifying the risk index into levels and optimizing the solution of the collaborative suppression scheme, and finally issuing millisecond-level power control commands in parallel, the control point is moved from post-remediation to pre-judgment, eliminating the reverse flow time window, realizing active prediction and millisecond-level suppression of reverse power flow, and ensuring the stability and safety of grid operation.

[0017] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the millisecond-level prediction and suppression method for reverse power flow in the photovoltaic-storage-charging collaborative scenario of this application.

[0020] Figure 2 This is a schematic diagram of the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario of this application.

[0021] Figure labeling: Data acquisition module 11, trend prediction module 12, risk assessment module 13, level classification module 14, millisecond-level parallel execution module 15. Detailed Implementation

[0022] This application provides a millisecond-level prediction and suppression method and device for reverse power flow in a photovoltaic-storage-charging collaborative scenario. It addresses the technical problem in existing technologies where triggering decisions rely solely on real-time power thresholds, lacking proactive awareness of power change trends at the grid connection point. This results in protection actions being executed only after reverse power flow occurs, leading to control lag and further impacting grid safety and stability. By introducing a net power trend prediction and prediction uncertainty index at the grid connection point to calculate a reverse power flow risk index, and then classifying risk levels based on this index and optimizing collaborative suppression schemes, millisecond-level power control commands are issued in parallel. This shifts the control point from reactive remediation to proactive prediction, eliminating the reverse flow time window and achieving proactive prediction and millisecond-level suppression of reverse power flow, ensuring grid operational stability and safety.

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0024] Example 1, please refer to the appendix. Figure 1 This application provides a millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario. The method is applied to a millisecond-level prediction and suppression device for reverse power flow in this scenario. The specific steps of the method are as follows: Real-time data acquisition is performed through the energy management controller to obtain multi-source operational data streams.

[0025] Furthermore, this application also includes the following steps: The energy management controller initiates a polling process with the grid-connected metering unit according to a first sampling period to obtain a first polling dataset; the energy management controller initiates a polling process with the photovoltaic inverter unit according to a second sampling period to obtain a second polling dataset; the energy management controller initiates a polling process with the battery management system and energy storage converter of the energy storage unit according to a third sampling period to obtain a third polling dataset; the energy management controller initiates a polling process with the station-level charging controller of the charging pile group according to a fourth sampling period to obtain a fourth polling dataset; the energy management controller timestamps the first, second, third, and fourth polling datasets obtained in the first, second, third, and fourth sampling periods to generate multiple labeled datasets; the data update time is extracted, and the data update time is used as a reference to perform interpolation and alignment processing on the multiple labeled datasets to generate the multi-source operating data stream.

[0026] Specifically, the energy management controller obtains the operating data of all key equipment under the same time reference. Since the grid-connected metering unit, photovoltaic inverter unit, energy storage unit and charging pile group have different real-time data requirements, the controller is set with four different sampling periods: the first sampling period, the second sampling period, the third sampling period and the fourth sampling period.

[0027] The energy management controller sends a polling request to the grid-connected metering unit in the first sampling cycle. Each response from the metering unit includes the current active power, reactive power, RMS voltage, and RMS current at the grid connection point. In the second sampling cycle, the energy management controller sends a polling request to the photovoltaic inverter unit to obtain the real-time output power of each photovoltaic string, the inverter's operating status, and the available margin for maximum power point tracking (MPPT), i.e., the difference between the current output and the theoretical maximum output at that moment. In the third sampling cycle, the energy management controller sends polling requests to the battery management system and energy storage converter of the energy storage unit. From the battery management system, it obtains the current state of charge percentage, state of health percentage, and the highest and lowest temperatures of all cells in the battery pack. From the energy storage converter, it obtains the current charging / discharging power values ​​and the adjustable upper and lower limits of the power, determined by the state of charge and temperature constraints. The energy management controller sends a polling request to the station-level charging controller of the charging pile group during the fourth sampling cycle to obtain the occupancy status of each charging gun, the requested current value set by the user through the vehicle or mobile phone, the current actual output current value of the charging gun, the amount of electricity accumulated in this charging session, and the estimated remaining charging time based on the current power and battery demand.

[0028] Because the durations of the four sampling periods may differ, and the network transmission delays for each poll are not entirely consistent, the energy management controller receives the four sets of data at different times. To accurately calculate the future net power trend at the grid connection point, the four sets of data must be aligned to the same point in time. The energy management controller appends a timestamp to each received data set, recording the time when the data arrived at the controller's local clock. The latest arrival time from all data streams is selected as the baseline update time. For devices without measured data at this baseline time, the most recent sampled data before and after the baseline time is identified. Using the time interval between these two data points as the denominator and the time difference between the baseline time and the previous sampling time as the numerator, the data value that the device should have at the baseline time is calculated linearly. For example, if a device samples at 0.3s with a value of 100kW and at 0.5s with a value of 110kW, and the baseline time is 0.4s, then the calculated value at 0.4s is approximately 105kW.

[0029] By performing the above interpolation alignment process for each baseline update time, the energy management controller transforms the discrete, asynchronous raw polling data into a continuous and time-aligned multi-source operating data stream. Each time point simultaneously includes the active power at the grid connection point, the photovoltaic output power, the energy storage state of charge and charging / discharging power, the session status of each charging gun in the charging pile group, and the total power of all charging guns.

[0030] Based on the multi-source operation data stream combined with the photovoltaic-storage-charging collaborative scenario, the grid connection point is predicted to obtain the net power trend and prediction uncertainty index of the grid connection point.

[0031] Furthermore, this application also includes the following steps: retrieving the output power records of the photovoltaic inverter units from the multi-source operation data stream to construct a photovoltaic output history sequence; performing adjacent sampling based on the photovoltaic output history sequence to extract multiple adjacent sampling points, calculating the power difference based on the multiple adjacent sampling points to obtain multiple power change step sizes; calculating the average power change step size by averaging the multiple power change step sizes to obtain the average power change step size, and using the average power change step size as a ramp-up change feature; introducing the measured photovoltaic output value and combining it with the ramp-up change feature size to synchronize with the short-time series prediction model for short-time prediction to generate a photovoltaic output prediction sequence; performing charging load prediction based on the multi-source operation data stream to construct a charging load prediction sequence; and superimposing the photovoltaic output prediction sequence and the charging load prediction sequence along the time axis to generate the grid-connected point net power trend.

[0032] Furthermore, this application also includes the following steps: extracting multiple photovoltaic output prediction values ​​by traversing the photovoltaic output prediction sequence according to time sampling points, wherein the multiple photovoltaic output prediction values ​​correspond to the time sampling points; extracting multiple charging load prediction values ​​by traversing the charging load prediction sequence according to the time sampling points, wherein the multiple charging load prediction values ​​correspond to the time sampling points; based on the time sampling points, subtracting the multiple photovoltaic output prediction values ​​from the multiple charging load prediction values ​​to obtain the grid-connected point active power prediction value at the time sampling points; sorting the grid-connected point active power prediction values ​​according to time sequence to construct a power prediction sequence; identifying data changes based on the power prediction sequence to determine multiple data identification points, and connecting the multiple data identification points according to the time sampling points to construct the grid-connected point net power trend.

[0033] Specifically, the output power records of the photovoltaic inverter units are retrieved from the multi-source operation data stream. In other words, the latest total output power value of the photovoltaic inverter units is extracted from the continuously updated multi-source operation data stream at fixed intervals. The photovoltaic output power values ​​at each sampling moment over a past period are then sorted in chronological order to form a historical sequence of photovoltaic output.

[0034] The process iterates through the historical photovoltaic (PV) output sequence, starting with the two oldest adjacent points. The output value of the latter point is subtracted from the output value of the former point to obtain a power change step size. This process is repeated, moving one sampling point forward and calculating the power difference between the next pair of adjacent points, until the entire sequence has been traversed, resulting in multiple power change step sizes. To eliminate the influence of single random fluctuations, the arithmetic mean of all calculated power change step sizes is taken: the algebraic sum of all step sizes is divided by the number of step sizes. This average power change step size is defined as the ramp-up characteristic quantity. A positive ramp-up characteristic quantity indicates an overall upward trend in PV output; a negative value indicates a downward trend; and a larger absolute value indicates more drastic output changes.

[0035] The latest measured photovoltaic (PV) output value and the previously calculated ramp-up change characteristic are input into a pre-trained short-term series prediction model. Based on the current output level and recent rate of change, the model calculates the PV output forecast for each sampling moment within the next few seconds to one minute, outputting the forecasts in chronological order to form a PV output prediction sequence. The input to the short-term series prediction model includes the measured PV output value and the recent ramp-up change characteristic. Upon initial model startup, two internal state variables need to be initialized: the horizontal baseline value is initialized to the first measured PV output value; the trend baseline value is initialized to the first ramp-up change characteristic. The smoothing parameters use empirical values: a horizontal smoothing parameter of 0.3 and a trend smoothing parameter of 0.1. The horizontal component represents the baseline level of PV output at the current moment, eliminating short-term random fluctuations. The new horizontal smoothing parameter = horizontal smoothing parameter × current measured value + (1 - horizontal smoothing parameter) × (old horizontal baseline value + old trend baseline value). The trend component represents the rate of change of current photovoltaic output. The new trend smoothing parameter = trend smoothing parameter × (new horizontal baseline value - old horizontal baseline value) + (1 - trend smoothing parameter) × old trend baseline value. For the k-th sampling interval in the future, the future predicted value = current horizontal baseline value + k × current trend baseline value. To improve the model's responsiveness to sudden changes, after each trend update, the current ramp change feature is weighted and averaged with the internal trend baseline value of the model, with each weight set to 0.5. That is, the corrected trend = 0.5 × model trend + 0.5 × recent ramp feature. One week of historical photovoltaic output data was collected and divided into training and validation sets. A grid search method was used to try various combinations of horizontal smoothing parameters from 0.1 to 0.9 and trend smoothing parameters from 0.01 to 0.3. The evaluation metrics were mean absolute percentage error and root mean square error. After verification, when the horizontal smoothing parameter was 0.3 and the trend smoothing parameter was 0.1, the mean absolute percentage error within the 3-second prediction window was the lowest, approximately 4.7%. Therefore, this set of parameters was fixed. The prediction window length is set to 3 seconds. Reverse power flow suppression requires millisecond-level response, but the power regulation of energy storage and charging piles requires a certain execution time: approximately 50 to 100 ms for energy storage converters and approximately 100 to 200 ms for charging pile power limiting. The 3-second window covers the execution delay without introducing excessive uncertainty due to an excessively long window. Furthermore, the change in photovoltaic output within 3 seconds generally does not exceed 10% of the rated capacity, resulting in high prediction reliability. The short-time series prediction model maintains two state variables: a horizontal baseline value and a trend baseline value, both initialized after power-on and continuously updated. The state variables are stored in the controller's non-volatile memory and will not be lost even during brief power outages.

[0036] Simultaneously, charging load prediction is performed based on charging pile group session information from multi-source operational data streams. The current occupancy status, user-requested current, actual output current, charged amount, and estimated remaining charging time for each charging gun are read. For vehicles currently charging, the power change trajectory of that gun over a future period is estimated based on its remaining charging time and current charging power. Based on the current session status of each charging gun, the trend of total charging power change over the next few seconds to minutes is predicted. For each occupied charging gun, the current actual output power, charged amount, remaining charging time, battery rated capacity, and current charging stage identifier are obtained. Future power changes are predicted according to different stages: in the constant current stage, the charging current is constant, and the power remains essentially unchanged until the voltage rises to a limit value; the power remains unchanged for the next three seconds until the remaining time is less than a certain threshold before considering a decrease; in the constant voltage stage, the voltage reaches its upper limit, the current begins to decay exponentially, and the power decreases approximately linearly or exponentially; the average rate of decrease is estimated based on the current power and remaining time; in the trickle stage, the power is very low and stable, typically used for battery equalization, and the power is predicted to remain unchanged for the next 3 seconds. For scenarios where the charging stage cannot be determined, the current power change rate is used to automatically identify the charging stage. If the power fluctuation within the past second is less than 1%, it is determined to be either constant current or trickle charging. If the power continues to decrease and the absolute value of the decrease rate is greater than 0.5 kW / s, it is determined to be a constant voltage stage, and the remaining time is dynamically estimated. For events approaching full charge, when the remaining charging time is less than 10 seconds, it is predicted that the power of the charging gun will rapidly drop to zero when the remaining time ends. In the prediction sequence, starting from the moment corresponding to the remaining time, the power is set to 0, and a linear transition of 1 is added within 0.5 seconds before and after that moment to avoid abrupt changes impacting the optimization solution. Although it is impossible to predict when the user will unplug the charging gun, the charging station's gun position sensor and vehicle communication status can be used to determine signals that the user may unplug the gun, such as a vehicle stop request or door unlocking. Once such a signal is detected, the model immediately linearly reduces the predicted power of the charging gun to 0 within the next 0.5 seconds for feedforward control. The predicted power curves of all occupied charging guns are summed point by point on the same time axis to obtain the total charging power prediction sequence. Meanwhile, the contribution of idle charging guns is zero. The system takes into account the occupancy status, current power, remaining duration, charging stage identifier, and event flags of each charging gun. It outputs a predicted total charging power for each sampling moment within a future period, as well as the individual adjustable potential for each gun (the difference between the current power and the minimum sustaining power). In a photovoltaic-storage-charging scenario, charging piles represent the largest controllable load and are the primary recipients of reverse power flow. By predicting which charging guns are about to finish charging, the system can proactively increase energy storage charging or reduce photovoltaic output, preventing reverse power flow at the grid connection point due to load drops. Simultaneously, for guns in the constant voltage stage, their power naturally decreases, allowing the energy management controller to adjust without actively limiting power, thus reducing interference with the charging service.

[0037] The photovoltaic (PV) power output prediction sequence and the charging load prediction sequence have the same prediction time length and the same interval of time sampling points. A unified time axis is set, with the starting point being the current time and the ending point being the end time of the prediction window. The sampling point interval on the time axis is consistent with the data alignment interval. Starting from time 0, each time sampling point is taken sequentially, and the PV power output prediction value corresponding to that time is taken from the PV power output prediction sequence, and the charging load prediction value corresponding to that time is taken from the charging load prediction sequence. For each time sampling point, the PV power output prediction value at that time is subtracted from the charging load prediction value at that time, and the result is the active power prediction value of the grid-connected point at that time. Other fixed loads in the station, except for charging piles, remain constant within the prediction window, and their values ​​can be obtained by back-calculating from the most recent measured power balance, and are added or subtracted in advance before calculation. In actual engineering, the energy management controller will first treat other fixed loads as known constant values, deduct them from the grid-connected point power, and then perform balancing of PV and charging. The predicted active power values ​​at the grid connection point calculated for each time sampling point are stored sequentially in an array or queue from earliest to latest time, forming a power prediction sequence. The length of the power prediction sequence is equal to the number of sampling points within the prediction window.

[0038] Traverse the power prediction sequence to identify key points where the power direction is about to change. Find two adjacent points in the sequence where the power value changes from positive to negative or from negative to positive, mark a zero-crossing point between these two points, and estimate the accurate time of the zero-crossing using linear interpolation. Identify the points in the sequence where the power value is locally maximum or minimum as extreme points, representing turning points in the power trend. Calculate the power difference between adjacent sampling points; when the absolute value of the difference exceeds a set threshold, mark that location as a point of abrupt change in the rate of change. These marked points are collectively referred to as data marker points.

[0039] Arrange all data points in chronological order and connect adjacent points with straight line segments to form a broken line, which is the net power trend chart of the grid connection point. The net power trend chart of the grid connection point is used to visually represent the changing trend of grid connection point power over a future period, including whether there is a risk of an impending shift from power purchase to reverse power transmission. If the net power trend chart of the grid connection point enters the negative region and continues to decline, it indicates that a reverse power flow is about to occur; if the trend line is negative but rising, it indicates that the reverse risk is weakening.

[0040] By overlaying timelines to organically integrate photovoltaic forecasts with charging load forecasts, a net power trend at the grid connection point is generated. This allows for accurate prediction of the timing and severity of risks a few seconds before the actual occurrence of reverse power flow, thus achieving a leap from passive response to proactive prediction.

[0041] Furthermore, this application also includes the following steps: continuously collecting data from the grid connection point in real time through the grid-connected metering unit to obtain the measured value of the active power at the grid connection point; comparing the predicted value of the active power at the grid connection point with the measured value of the active power at the grid connection point to calculate multiple prediction deviations; performing extreme value analysis based on the multiple prediction deviations to extract the maximum and minimum deviation values; performing boundary analysis based on the maximum and minimum deviation values ​​to set the width of the deviation distribution interval; obtaining the rated power at the grid connection point and calculating the ratio of the width of the deviation distribution interval to the rated power at the grid connection point as the baseline uncertainty; performing fluctuation analysis based on the photovoltaic-storage-charging collaborative scenario to obtain multiple fluctuation factors; and dynamically correcting the baseline uncertainty according to the multiple fluctuation factors to construct a prediction uncertainty index.

[0042] Specifically, the measured active power at the grid-connected point is continuously collected through the grid-connected metering unit, with the sampling frequency consistent with the data alignment frequency. The predicted value for each moment in the past period is saved. For each historical moment that has passed, the predicted active power at the grid-connected point at that moment is compared with the measured active power at that moment, and the prediction deviation is calculated.

[0043] Within a defined historical time window, all calculated prediction biases are iterated through, and multiple maxima and minima are identified, denoted as the maximum and minimum bias values, respectively. The maximum bias value reflects the largest possible positive prediction error, while the minimum bias value reflects the largest possible negative prediction error. Together, they characterize the distribution range of the prediction error. Subtracting the minimum bias value from the maximum bias value yields the width of the bias distribution interval. A larger bias distribution interval width indicates more dispersed prediction results and poorer stability.

[0044] The rated power capacity of the grid-connected point is read, and the width of the deviation distribution interval is divided by the rated power of the grid-connected point to obtain a dimensionless proportional value, called the baseline uncertainty, which represents the basic reliability of the prediction model under normal operating conditions. Based on the real-time operating status of the photovoltaic-storage-charging collaborative scenario, several factors reflecting the severity of current fluctuations are extracted, including photovoltaic ramp-up factors, charging load mutation factors, weather fluctuation factors, and plug-in / plug-out event factors. The average value of each factor is calculated to obtain the comprehensive fluctuation factor. The baseline uncertainty is multiplied by the comprehensive fluctuation factor to obtain the final prediction uncertainty index. The prediction uncertainty index automatically increases as scenario fluctuations intensify, and falls back to near the baseline value when the scenario stabilizes.

[0045] For example, a photovoltaic-storage-charging station has a rated power of 500 kW at its grid connection point. The data alignment interval is 50 ms, and the prediction window is 3 s. After each prediction, the controller saves the predicted values ​​for future times and calculates the deviation after the actual measured values ​​arrive at the corresponding times. Historical prediction deviation data from the past 5 minutes is collected, with a total of 6000 sampling points. Statistically, among all signed historical prediction deviation data, the maximum deviation is +15.3 kW, and the minimum deviation is -12.8 kW. The deviation distribution interval width = 15.3 - (-12.8) = 28.1 kW, and the baseline uncertainty = 28.1 kW / 500 kW = 0.0562. The photovoltaic ramp-up characteristic is +3.88 kW / s. With a rated power of 500kW and a baseline rate of change of 50kW / s, the ramp rate of 3.88kW / s is much smaller than 50kW / s, therefore the relative ramp rate is 0.0776, and the photovoltaic ramp factor is set to 1.0. In the most recent second, the total charging power decreased from 80.9kW to 80.8kW, a rate of change of -0.1kW / s, a very small absolute value, so the charging load mutation factor is set to 1. There were no plug-in / plug-out events, so the plug-in / plug-out factor is set to 1. The comprehensive fluctuation factor is 1.0. The prediction uncertainty index = baseline uncertainty × comprehensive fluctuation factor = 0.0562 × 1.0 = 0.0562. Assuming a sudden weather change and rapid cloud movement causing drastic fluctuations in photovoltaic output, the ramp change characteristic reaches +25kW / s, at which point the photovoltaic ramp factor is defined as factor = 1.5. Simultaneously, three charging guns at the charging station enter the constant voltage phase, with the total power decreasing at a rate of 15 kW / s. The charging load mutation factor is set to 1.3, the insertion / removal factor remains at 1, and the comprehensive fluctuation factor is 1.5 × 1.3 = 1.95. The baseline uncertainty remains at 0.0562, therefore the prediction uncertainty index is 0.0562 × 1.95 = 0.1096. The controller, recognizing the significantly increased prediction uncertainty, will correspondingly increase the weight of the third risk component in the subsequent calculation of the reverse power flow risk index, thereby increasing the risk index and initiating suppression measures earlier or more aggressively to address the potential risks arising from unreliable predictions. The prediction uncertainty index is a dimensionless numerical value in the final output, used to quantify the reliability of the current prediction result. The larger the index value, the more unreliable the prediction result, and the more conservative the reverse power flow suppression strategy should be.

[0046] A reverse power flow threshold is preset, and the reverse power flow risk index is calculated by comparing the net power trend at the grid connection point with the reverse power flow threshold.

[0047] Furthermore, this application also includes the following steps: reading the reverse power transmission allowable limit, converting the reverse power transmission allowable limit according to the dimensions of the grid-connected active power prediction value, and determining the reverse power flow threshold; performing reverse limit judgment one by one based on the grid-connected active power prediction value and the reverse power flow threshold, and recording multiple limit exceedance values; extracting the maximum value of the multiple limit exceedance values ​​as the peak limit exceedance value, and calculating the arithmetic mean of the multiple limit exceedance values ​​as the average limit exceedance value; performing reverse power flow risk analysis on the peak limit exceedance value and the average limit exceedance value to construct a first risk component; performing adjacent point difference operation based on the grid-connected active power sequence to obtain the instantaneous change rate of grid-connected active power, and performing power impact intensity analysis on the instantaneous change rate to construct a second risk component; performing prediction reliability calculation based on the prediction uncertainty index, and performing risk assessment impact analysis based on the prediction reliability to construct a third risk component; and obtaining the reverse power flow risk index by weighted summation of the first risk component, the second risk component, and the third risk component.

[0048] Specifically, the reverse power supply allowance limit set by the maintenance personnel according to the grid connection protocol is read. It is 0 when there is no reverse power supply requirement and positive when limited reverse power supply is allowed. Since the unit of the active power prediction value at the grid connection point is kilowatts, and the power purchase is positive and the reverse power supply is negative, the controller converts the reverse power supply allowance limit into a reverse power flow threshold, which is equal to the negative reverse power supply allowance limit.

[0049] Iterate through each predicted value in the active power prediction sequence at the grid connection point. For each predicted value, compare it with the reverse power flow threshold. If the predicted active power at the grid connection point is less than the reverse power flow threshold, it is determined to be a reverse over-limit, and the over-limit amplitude is calculated. The over-limit amplitude is calculated as the reverse power flow threshold minus the predicted active power at the grid connection point. The result is a positive number, representing the number of kilowatts of reverse power exceeding the allowable limit. If the predicted active power at the grid connection point is greater than or equal to the reverse power flow threshold, the over-limit amplitude is not recorded. Record all over-limit amplitudes to form an over-limit amplitude list.

[0050] The maximum value among multiple over-limit amplitudes is identified as the peak over-limit amount. The arithmetic mean of the multiple over-limit amplitudes is calculated as the average over-limit amount. If no over-limit occurs, both the peak and average over-limit amounts are 0. A reverse current risk analysis is performed on the peak and average over-limit amounts to obtain the first risk component, which reflects the severity of the over-limit occurrence in the reverse current: a larger peak over-limit amount indicates a more severe instantaneous over-limit occurrence; a larger average over-limit amount indicates a longer duration or greater overall severity of the over-limit occurrence. The first risk component equals the peak over-limit amount plus half of the average over-limit amount. For example, if the peak over-limit amount = 13.4 kW and the average over-limit amount = 13.4 kW, then the first risk component = peak over-limit amount + 0.5 × average over-limit amount = 20.1.

[0051] The active power prediction sequence at the grid connection point is subjected to adjacent-point differential calculation to obtain the instantaneous rate of change within each time interval. The instantaneous rate of change between adjacent points is obtained by subtracting the predicted value of the previous moment from the predicted value of the next moment and then dividing by the time interval. The maximum absolute value of the instantaneous rate of change is extracted and divided by a baseline rate of change to obtain a dimensionless impact intensity coefficient. The second risk component is this coefficient; a larger coefficient indicates a faster power change and a stronger impact on the power grid. For example, if the maximum absolute value (impact intensity) is 42.0 kW / s; the baseline rate of change is 10% of the rated power, i.e., 50 kW / s; the impact intensity coefficient = 42.0 / 50 = 0.84, then the second risk component = 0.84.

[0052] Based on the prediction uncertainty index, the prediction credibility is calculated as: Prediction Credibility = 1 - Prediction Uncertainty Index. A higher prediction credibility indicates a more reliable prediction result, and the risk of reverse current flow should be correspondingly reduced; conversely, a lower prediction credibility indicates that the prediction may be inaccurate, and the risk index should be appropriately increased to allow for some margin. The third risk component can be defined as 1 - Prediction Credibility, which is directly equal to the prediction uncertainty index. For example, assuming the prediction uncertainty index is 0.1096, then the third risk component = 0.1096.

[0053] Weighting coefficients are assigned to the three risk components, such as a weight of 0.5 for the first risk component, 0.3 for the second, and 0.2 for the third. The reverse current risk index = (weight of first risk component × first risk component) + (weight of second risk component × second risk component) + (weight of third risk component × third risk component). The reverse current risk index is a dimensionless positive number used to determine the risk level of reverse current; a higher value indicates a higher risk. For example, if the first risk component is 20.1, the second risk component is 0.84, and the third risk component is 0.1096, then the reverse current risk index is approximately 10.32.

[0054] The first risk component considers both peak and average over-limit quantities, comprehensively reflecting the severity of over-limit quantities in the reverse current flow; the second risk component assesses the power impact intensity through the instantaneous rate of change; the third risk component incorporates prediction uncertainty into the risk assessment; the three components are multiplied by adjustable weighting coefficients to adjust the risk assessment strategy according to different scenarios.

[0055] Based on the aforementioned reverse current risk index, the risk levels are classified, multiple control levels are determined, and optimization solutions are developed to formulate a collaborative suppression decomposition scheme.

[0056] Furthermore, this application also includes the following steps: storing a first risk threshold and a second risk threshold based on the energy management controller, wherein the first risk threshold is less than the second risk threshold; comparing the reverse current risk index with the first risk threshold; if the reverse current risk index is less than the first risk threshold, it is determined to be a mild risk level, and a first control level is activated; if the reverse current risk index is greater than or equal to the first risk threshold, comparing the reverse current risk index with the second risk threshold; if the reverse current risk index is less than the second risk threshold, it is determined to be a moderate risk level, and a second control level is activated; if the reverse current risk index is greater than or equal to the second risk threshold, it is determined to be a severe risk level, and a third control level is activated.

[0057] Specifically, the energy management controller pre-stores two risk thresholds: a first risk threshold and a second risk threshold, with the first risk threshold being lower than the second. The currently calculated reverse power flow risk index is compared to the first risk threshold. If the risk index is lower than the first risk threshold, it indicates a very low reverse power flow risk, and the current state is classified as a mild risk level, activating the first control level. Under the first control level, the energy management controller only calls upon the energy storage unit for fine-tuning, such as charging at a lower power output to absorb potential slight power surpluses, without involving power limiting at charging stations or throttling of photovoltaic power, thereby minimizing the impact on charging services and photovoltaic power generation.

[0058] If the reverse power flow risk index is greater than or equal to the first risk threshold, it indicates that the risk has reached a moderate or higher level. The reverse power flow risk index is then compared with a second risk threshold. If the reverse power flow risk index is less than the second risk threshold, it is determined to be a moderate risk level, and the second control level is activated. Under the second control level, the energy management controller simultaneously coordinates the energy storage unit and the charging pile network. The energy storage unit absorbs charge at a higher power, while the charging pile network performs power limiting operations according to session urgency and fairness factors, appropriately reducing the charging power to release consumption space, but without forcibly interrupting the ongoing charging session.

[0059] If the reverse power flow risk index is greater than or equal to the second risk threshold, it is classified as a severe risk level, and the third control level is activated. Under the third control level, the energy management controller simultaneously activates the energy storage unit, charging pile cluster, and photovoltaic inverter unit for joint operation. The energy storage unit charges and absorbs power at its maximum permissible power; the charging pile cluster implements more significant power limiting, and if necessary, reduces the power of some charging guns to the minimum maintenance power; the photovoltaic inverter unit implements active power derating, proactively reducing photovoltaic output to eliminate surplus power at the source. The three devices work together to ensure that the power at the grid connection point is strictly controlled within the reverse power flow threshold. The first and second risk thresholds are flexibly adjusted based on factors such as grid sensitivity, energy storage battery life requirements, and charging service quality requirements. This tiered activation mechanism reduces unnecessary equipment actions. Under mild risk, charging pile power limiting and photovoltaic derating are not invoked, protecting the charging user experience and photovoltaic power generation revenue. Under moderate risk, energy storage and moderate power limiting are prioritized to avoid curtailment. Photovoltaic derating is only activated under severe risk to maximize energy utilization.

[0060] Furthermore, this application also includes the following steps: when the first control level is activated, the equipment combination is determined to be only calling the energy storage unit for energy storage regulation optimization, and the energy storage regulation power component is obtained; constraint solution is performed based on the energy storage regulation power component to generate a reverse power flow suppression task; when the second control level is activated, the equipment combination is determined to be simultaneously calling the energy storage unit and the charging pile group for group energy storage regulation optimization, and the charging pile group regulation power component is obtained; constraint solution is performed based on the charging pile group regulation power component to generate a charging pile group load component regulation task; when the third control level is activated, the equipment combination is determined to be simultaneously calling the energy storage unit, the charging pile group, and the photovoltaic inverter unit for photovoltaic derating regulation optimization, and the photovoltaic derating power component is obtained; constraint solution is performed based on the photovoltaic derating power component to generate an output derating task; the reverse power flow suppression task, the charging pile group load component regulation task, and the output derating task are integrated to formulate the collaborative suppression decomposition scheme.

[0061] Specifically, when the first control level is activated, only the energy storage unit is invoked for regulation. The reverse power flow suppression demand is calculated, which equals the portion of the current grid-connected point power prediction exceeding the reverse power flow threshold. Under the premise of satisfying the energy storage unit's own constraints, the energy storage regulation power component is solved. Constraints include ensuring the energy storage's state of charge is within a safe range, limiting the maximum charging power of the energy storage converter, preventing the battery temperature from exceeding a protection threshold, and ensuring the charging power change slope does not exceed a set value. This aims to make the energy storage's absorbed power as close as possible to the suppression demand while avoiding overcharging or overheating. After solving, the controller encapsulates the energy storage regulation power component into a reverse power flow suppression task, including parameters such as the target charging power value and power rise time.

[0062] When the second control level is activated, both the energy storage unit and the charging pile group are simultaneously invoked for coordinated regulation. The total suppression demand is calculated and decomposed into energy storage regulation power components and charging pile group regulation power components based on the principle of minimum action cost or highest priority. Energy storage absorption is typically prioritized; when energy storage absorption capacity is insufficient or costs are too high, the charging pile group reduces the load by limiting power. The charging pile group regulation power component is further decomposed to each charging gun currently in use. During decomposition, a power limit value is calculated for each gun based on session urgency, amount charged, and fairness factor, ensuring that the power of each gun does not fall below the minimum maintenance power, such as 3.5kW, and that the power decrease rate is gradual to avoid charging interruptions. The energy management controller generates a charging pile load component regulation task, including the target power upper limit and regulation rate for each gun.

[0063] When the third control level is activated, energy storage, charging pile clusters, and photovoltaic inverter units are simultaneously invoked for joint suppression. The total suppression demand is calculated and allocated according to cost priority: energy storage, with the lowest cost, is used first; charging pile cluster power limiting is next; and photovoltaic derating, with the highest cost, is used last. The power components for energy storage regulation, charging pile cluster regulation, and photovoltaic derating are solved sequentially. For photovoltaic derating, the controller also needs to consider the inverter's derating rate limit, such as no more than 10% of rated power per second, and the minimum output limit, such as no less than 5% of rated power, to avoid frequent start-stops. After the solution is completed, the energy management controller generates an output derating task, which includes the target maximum output value of the photovoltaic inverter.

[0064] The generated reverse power flow suppression tasks, pile group load component adjustment tasks, and output derating tasks are integrated to form a collaborative suppression decomposition scheme. This scheme clarifies the power regulation targets, timing requirements, and compensation conditions for each device, and then issues these tasks to the corresponding devices in parallel. For mild risks, only energy storage is used, avoiding unnecessary interference with charging services and photovoltaic power generation; for moderate risks, pile group power limiting is added to fully utilize the load-side adjustability; for severe risks, photovoltaic derating is activated again to reduce surplus power at the source. The total suppression demand is decomposed into multiple power components, and corresponding execution tasks are generated for each component, allowing each device to execute independently and in parallel without interference. The energy management controller only needs to issue target values, and the local controllers of each device are responsible for closed-loop tracking, improving system reliability and response speed.

[0065] The cooperative suppression decomposition scheme is executed and power control commands are issued to perform millisecond-level prediction and suppression of reverse power flow.

[0066] Furthermore, this application also includes the following steps: reading the energy storage regulation power component of the reverse power flow suppression task and performing algebraic summation with the actual charging and discharging power setpoint of the energy storage converter to generate a first power control command; reading the pile group regulation power component of the pile group load component regulation task and performing charging power reduction decomposition to obtain a first power limit value; performing verification based on the first power limit value, and generating a second power control command when the verification passes; reading the photovoltaic derating power component of the output derating task, performing available maximum analysis based on the photovoltaic inverter unit, extracting the maximum available output power, and generating a second power limit value by subtracting the maximum available output power from the photovoltaic derating power component; performing operating point tracking according to the second power limit value to generate a third power control command; and issuing the first power control command, the second power control command, and the third power control command in parallel for millisecond-level response collaborative prediction and suppression.

[0067] Specifically, the energy storage regulation power component stored in the reverse power flow suppression task is read. This is a signed value, with positive indicating discharge and negative indicating charging. The actual charge / discharge power setpoint currently being executed by the energy storage converter is read. The energy storage regulation power component is algebraically summed with the actual charge / discharge power setpoint of the energy storage converter to obtain a new target power value, i.e., the first power control command. For example, if the energy storage is currently discharging at 20kW, and the suppression task requires an additional 30kW of charging (i.e., the regulation component is -30kW), then the algebraic sum is -10kW, indicating a need to switch from 20kW discharge to 10kW charging.

[0068] The controller reads the pile group adjustment power component stored in the pile group load component adjustment task, which is the total power value that needs to be reduced. This power component is then distributed to each charging gun according to the pile group load component adjustment task, resulting in a first power limit value for each gun. Before generating instructions, the controller verifies each limit value, checking if it is lower than the gun's minimum sustaining power. If so, the limit value is adjusted to the minimum sustaining power, and the allocation to other guns is adjusted accordingly. The controller also checks if the power reduction rate is too fast; if so, it can add a transition ramp or execute in stages. After successful verification, the controller encapsulates the upper power limit value of each gun into a second power control instruction.

[0069] The system reads the derating power component of the photovoltaic (PV) system stored in the output derating task, which represents the output value that needs to be reduced. It then queries the PV inverter unit for the maximum available output power under current environmental conditions, i.e., the theoretically maximum power it can generate. Subtracting the derating power component from the maximum available output power yields the second power limit, which is the upper limit of the maximum power allowed for the PV inverter to output. This limit is encapsulated as a third power control command, instructing the PV inverter to perform maximum power point tracking (MPPT) according to this limit. Based on the received power limit, the system adjusts the MPPT algorithm or directly limits the output active power to ensure that the actual output does not exceed the limit. Typically, the inverter will preferentially operate at its maximum power point below the limit.

[0070] The first, second, and third power control commands are sent simultaneously to the energy storage converter, the charging pile group controller, and the photovoltaic inverter unit, respectively. Upon receiving the commands, each device immediately adjusts its operating status accordingly. The energy storage converter changes its charging and discharging power within milliseconds; the charging pile group controller lowers the power limit of each charging station within tens of milliseconds; and the photovoltaic inverter reduces its output within tens to hundreds of milliseconds. This coordinated action ensures that the power at the grid connection point is controlled within the threshold range before reverse power flow actually occurs.

[0071] In the next sampling cycle after the instruction is issued, the controller again collects the measured value of the active power at the grid-connected point to determine whether the reverse power flow has been successfully suppressed, i.e., whether the power at the grid-connected point is greater than or equal to the reverse power flow threshold. If the requirement is met, the energy management controller enters the recovery phase, gradually restoring the charging pile power and photovoltaic output, and reducing the energy storage charging power to normal levels. If the reverse power flow is still not completely eliminated, the energy management controller will recalculate the risk index and may upgrade the control level.

[0072] The energy storage regulation power component is algebraically summed with the current actual set value, so that the new instruction is an incremental adjustment based on the current operating point, avoiding sudden changes; the charging pile power limiting instruction is verified and adjusted to ensure the quality of charging service; the photovoltaic derating instruction is calculated based on the maximum available output power, rather than based on the current actual output, avoiding the logical error of mistakenly believing that derating is required when the current output is already below the limit value; the parallel issuance of three instructions replaces serial execution, realizing millisecond-level response and collaborative prediction and suppression of reverse power flow.

[0073] In summary, the millisecond-level prediction and suppression method for reverse power flow in the photovoltaic-storage-charging collaborative scenario provided in this application has the following technical effects: by introducing the net power trend prediction and prediction uncertainty index at the grid connection point to calculate the reverse power flow risk index, and classifying the risk index into levels and optimizing the collaborative suppression scheme, the method finally issues millisecond-level power control commands in parallel, shifting the control point from post-remediation to pre-judgment, eliminating the reverse flow time window, realizing active prediction and millisecond-level suppression of reverse power flow, and ensuring the stability and safety of grid operation.

[0074] Example 2: Based on the same inventive concept as the millisecond-level prediction and suppression method for reverse power flow in the photovoltaic-storage-charging collaborative scenario in Example 1, this application also provides a millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario. Please refer to the appendix. Figure 2 The millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario includes: a data acquisition module 11, used to acquire multi-source operating data streams in real time through an energy management controller; a trend prediction module 12, used to predict the grid connection point based on the multi-source operating data streams and the photovoltaic-storage-charging collaborative scenario, and obtain the net power trend and prediction uncertainty index of the grid connection point; a risk judgment module 13, used to preset a reverse power flow threshold, compare the net power trend of the grid connection point with the reverse power flow threshold, and calculate the reverse power flow risk index; a level classification module 14, used to classify levels according to the reverse power flow risk index, determine multiple control levels for optimization, and formulate a collaborative suppression decomposition scheme; and a millisecond-level parallel execution module 15, used to execute the collaborative suppression decomposition scheme and issue power control commands to perform millisecond-level prediction and suppression of reverse power flow.

[0075] Furthermore, the data acquisition module 11 in the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario is also used for: initiating a polling to the grid-connected metering unit according to a first sampling period through the energy management controller to obtain a first polling dataset; initiating a polling to the photovoltaic inverter unit according to a second sampling period through the energy management controller to obtain a second polling dataset; initiating a polling to the battery management system and energy storage converter of the energy storage unit according to a third sampling period through the energy management controller to obtain a third polling dataset; initiating a polling to the station-level charging controller of the charging pile group according to a fourth sampling period through the energy management controller to obtain a fourth polling dataset; the energy management controller timestamps the first, second, third, and fourth polling datasets obtained in the first, second, third, and fourth sampling periods to generate multiple labeled datasets; extracting the data update time, and using the data update time as a benchmark to perform interpolation and alignment processing on the multiple labeled datasets to generate the multi-source operating data stream.

[0076] Furthermore, the trend prediction module 12 in the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario is also used for: retrieving the output power records of the photovoltaic inverter unit from the multi-source operation data stream to construct a photovoltaic output historical sequence; performing adjacent sampling based on the photovoltaic output historical sequence to extract multiple adjacent sampling points, calculating the power difference based on the multiple adjacent sampling points to obtain multiple power change step sizes; calculating the average power change step size by averaging the multiple power change step sizes to obtain the average power change step size, and using the average power change step size as a ramp-up change feature; introducing the measured photovoltaic output value and combining it with the ramp-up change feature size to synchronize to the short-time series prediction model for short-time prediction to generate a photovoltaic output prediction sequence; performing charging load prediction based on the multi-source operation data stream to construct a charging load prediction sequence; and superimposing the photovoltaic output prediction sequence and the charging load prediction sequence along the time axis to generate the grid-connected point net power trend.

[0077] Furthermore, the trend prediction module 12 in the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario is also used for: extracting multiple photovoltaic output prediction values ​​by traversing the photovoltaic output prediction sequence according to time sampling points, wherein the multiple photovoltaic output prediction values ​​correspond to the time sampling points; extracting multiple charging load prediction values ​​by traversing the charging load prediction sequence according to the time sampling points, wherein the multiple charging load prediction values ​​correspond to the time sampling points; subtracting the multiple photovoltaic output prediction values ​​from the multiple charging load prediction values ​​based on the time sampling points to obtain the grid-connected active power prediction value at the time sampling points; sorting the grid-connected active power prediction values ​​according to time sequence to construct a power prediction sequence; identifying data changes based on the power prediction sequence to determine multiple data identification points; and connecting the multiple data identification points according to the time sampling points to construct the grid-connected net power trend.

[0078] Furthermore, the trend prediction module 12 in the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario is also used for: continuously collecting data from the grid connection point in real time through the grid-connected metering unit to obtain the measured value of the active power at the grid connection point; comparing the predicted value of the active power at the grid connection point with the measured value of the active power at the grid connection point to calculate multiple prediction deviations; performing extreme value analysis based on the multiple prediction deviations to extract the maximum and minimum deviation values; performing boundary analysis based on the maximum and minimum deviation values ​​to set the width of the deviation distribution interval; obtaining the rated power at the grid connection point and calculating the ratio of the width of the deviation distribution interval to the rated power at the grid connection point as the baseline uncertainty; performing fluctuation analysis based on the photovoltaic-storage-charging collaborative scenario to obtain multiple fluctuation factors; and dynamically correcting the baseline uncertainty according to the multiple fluctuation factors to construct a prediction uncertainty index.

[0079] Furthermore, the risk judgment module 13 in the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario is also used for: reading the reverse power transmission allowable limit, converting the reverse power transmission allowable limit according to the dimension of the grid-connected active power prediction value, and determining the reverse power flow threshold; performing reverse limit judgment one by one based on the grid-connected active power prediction value and the reverse power flow threshold, and recording multiple limit exceedance values; extracting the maximum value of the multiple limit exceedance values ​​as the peak limit exceedance value, and calculating the arithmetic mean of the multiple limit exceedance values ​​as the average limit exceedance value; performing reverse power flow risk analysis on the peak limit exceedance value and the average limit exceedance value to construct a first risk component; performing adjacent point difference operation based on the grid-connected active power sequence to obtain the instantaneous change rate of grid-connected active power, performing power impact intensity analysis on the instantaneous change rate, and constructing a second risk component; performing prediction reliability calculation based on the prediction uncertainty index, performing risk assessment impact analysis based on the prediction reliability, and constructing a third risk component; and obtaining the reverse power flow risk index by weighted summation of the first risk component, the second risk component, and the third risk component.

[0080] Furthermore, the level classification module 14 in the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario is also used for: storing a first risk threshold and a second risk threshold based on the energy management controller, wherein the first risk threshold is less than the second risk threshold; comparing the reverse power flow risk index with the first risk threshold; if the reverse power flow risk index is less than the first risk threshold, it is determined to be a mild risk level and the first control level is activated; if the reverse power flow risk index is greater than or equal to the first risk threshold, it is compared with the second risk threshold; if the reverse power flow risk index is less than the second risk threshold, it is determined to be a moderate risk level and the second control level is activated; if the reverse power flow risk index is greater than or equal to the second risk threshold, it is determined to be a severe risk level and the third control level is activated.

[0081] Furthermore, the level division module 14 in the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario is also used for: when the first control level is activated, determining that the device combination is to only call the energy storage unit for energy storage regulation optimization, and obtaining the energy storage regulation power component; performing constraint solving based on the energy storage regulation power component to generate a reverse power flow suppression task; when the second control level is activated, determining that the device combination is to simultaneously call the energy storage unit and the charging pile group for group energy storage regulation optimization, and obtaining the pile group regulation power component; performing constraint solving based on the pile group regulation power component to generate a pile group load component regulation task; when the third control level is activated, determining that the device combination is to simultaneously call the energy storage unit, the charging pile group, and the photovoltaic inverter unit for photovoltaic derating regulation optimization, and obtaining the photovoltaic derating power component; performing constraint solving based on the photovoltaic derating power component to generate an output derating task; and integrating the reverse power flow suppression task, the pile group load component regulation task, and the output derating task to formulate the collaborative suppression decomposition scheme.

[0082] Furthermore, the millisecond-level parallel execution module 15 in the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario is also used for: reading the energy storage regulation power component of the reverse power flow suppression task and performing algebraic summation with the actual charging and discharging power setpoint of the energy storage converter to generate a first power control command; reading the pile group regulation power component of the pile group load component regulation task and performing charging power reduction decomposition to obtain a first power limit value; performing verification based on the first power limit value, and generating a second power control command when the verification passes; reading the photovoltaic derating power component of the output derating task, performing available maximum analysis based on the photovoltaic inverter unit, extracting the maximum available output power, and generating a second power limit value by subtracting the maximum available output power from the photovoltaic derating power component; performing operating point tracking according to the second power limit value to generate a third power control command; and issuing the first power control command, the second power control command, and the third power control command in parallel for millisecond-level response collaborative prediction and suppression.

[0083] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Through the foregoing detailed description of the millisecond-level prediction and suppression method for reverse power flow in the photovoltaic-storage-charging collaborative scenario, those skilled in the art can clearly understand the millisecond-level prediction and suppression device for reverse power flow in the photovoltaic-storage-charging collaborative scenario in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0084] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0085] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario, characterized in that, include: Real-time data acquisition is performed through the energy management controller to obtain multi-source operational data streams; Based on the multi-source operation data stream combined with the photovoltaic-storage-charging collaborative scenario, the grid connection point is predicted to obtain the net power trend and prediction uncertainty index of the grid connection point; A reverse power flow threshold is preset, and the reverse power flow risk index is calculated by comparing the net power trend at the grid connection point with the reverse power flow threshold. Based on the aforementioned reverse current risk index, the risk levels are classified, multiple control levels are determined for optimization, and a collaborative suppression decomposition scheme is formulated. The cooperative suppression decomposition scheme is executed and power control commands are issued to perform millisecond-level prediction and suppression of reverse power flow.

2. The millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario as described in claim 1, characterized in that, Real-time acquisition of multi-source operational data streams is achieved through an energy management controller, including the following methods: The energy management controller initiates a polling process with the grid-connected metering unit according to the first sampling period to obtain the first polling dataset. The energy management controller initiates a polling process with the photovoltaic inverter unit according to the second sampling period to obtain the second polling dataset; The energy management controller initiates a polling process with the battery management system and energy storage converter of the energy storage unit according to the third sampling period to obtain the third polling dataset; The energy management controller initiates a polling of the station-level charging controllers of the charging pile group according to the fourth sampling period to obtain the fourth polling dataset; The energy management controller timestamps the first polling dataset, the second polling dataset, the third polling dataset, and the fourth polling dataset obtained in the first sampling period, the second sampling period, the third sampling period, and the fourth sampling period, generating multiple labeled datasets; Extract the data update time, and use the data update time as a benchmark to perform interpolation and alignment processing on the multiple labeled datasets to generate the multi-source running data stream.

3. The millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario as described in claim 1, characterized in that, Based on the multi-source operation data stream combined with the photovoltaic-storage-charging collaborative scenario, the grid connection point is predicted to obtain the net power trend and prediction uncertainty index of the grid connection point. The method includes: The output power records of the photovoltaic inverter units are retrieved from the multi-source operation data stream to construct a historical sequence of photovoltaic output. Based on the photovoltaic power output history sequence, adjacent sampling is performed to extract multiple adjacent sampling points. Power difference is calculated based on the multiple adjacent sampling points to obtain multiple power change step sizes. The average power change step size is calculated by averaging the multiple power change step sizes, and the average power change step size is used as the climbing change characteristic quantity. The measured photovoltaic output value is introduced and combined with the slope change characteristic quantity, and synchronized to the short-time series prediction model for short-time prediction to generate a photovoltaic output prediction series. Based on the multi-source operation data stream, charging load prediction is performed, and a charging load prediction sequence is constructed. The photovoltaic power output prediction sequence and the charging load prediction sequence are superimposed on the time axis to generate the net power trend at the grid connection point.

4. The millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario as described in claim 3, characterized in that, The method of superimposing the photovoltaic power output prediction sequence and the charging load prediction sequence along a time axis to generate the net power trend at the grid connection point includes: Multiple photovoltaic output prediction values ​​are extracted by traversing the photovoltaic output prediction sequence according to time sampling points, and the multiple photovoltaic output prediction values ​​have a corresponding relationship with the time sampling points; Multiple charging load prediction values ​​are extracted by traversing the charging load prediction sequence according to the time sampling points, and the multiple charging load prediction values ​​have a corresponding relationship with the time sampling points; Based on the time sampling points, the difference between the multiple photovoltaic output prediction values ​​and the multiple charging load prediction values ​​is calculated to obtain the grid-connected active power prediction value at the time sampling points. The predicted active power values ​​at the grid connection point are sorted according to time sequence to construct a power prediction sequence; Based on the power prediction sequence, data change identification is performed, multiple data identification points are determined, and the multiple data identification points are connected according to time sampling points to construct the net power trend of the grid connection point.

5. The millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging coordinated scenario as described in claim 4, characterized in that, The process of constructing the prediction uncertainty index includes the following methods: The grid connection point is continuously and in real time collected by the grid connection metering unit to obtain the measured value of the active power at the grid connection point; The predicted active power at the grid connection point is compared with the measured active power at the grid connection point, and multiple prediction deviations are calculated. Extreme value analysis is performed based on the multiple prediction deviations to extract the maximum and minimum deviation values. Boundary analysis is performed based on the maximum and minimum deviation values ​​to determine the width of the deviation distribution interval. Obtain the rated power at the grid connection point, and calculate the ratio of the deviation distribution interval width to the rated power at the grid connection point as the benchmark uncertainty; Based on the aforementioned photovoltaic-storage-charging collaborative scenario, fluctuation analysis was performed to obtain multiple fluctuation factors; The baseline uncertainty is dynamically corrected based on the aforementioned multiple volatility factors to construct a predictive uncertainty index.

6. The millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario as described in claim 1, characterized in that, A reverse power flow threshold is preset, and a reverse power flow risk index is calculated by comparing the net power trend at the grid connection point with the reverse power flow threshold. The method includes: Read the reverse power supply allowable limit, convert the reverse power supply allowable limit according to the dimension of the grid-connected active power prediction value, and determine the reverse power flow threshold. Based on the predicted active power at the grid connection point and the reverse power flow threshold, reverse over-limit judgments are made one by one, and multiple over-limit amplitudes are recorded. Extract the maximum value of the multiple over-limit amplitude values ​​as the peak over-limit value, and calculate the arithmetic mean of the multiple over-limit amplitude values ​​as the average over-limit value; Perform anti-current risk analysis on the peak exceedance and the average exceedance to construct the first risk component; Based on the active power sequence of the grid connection point, the adjacent point difference operation is performed to obtain the instantaneous change rate of the grid-connected active power. The instantaneous change rate is then used for power impact intensity analysis to construct a second risk component. Based on the aforementioned prediction uncertainty index, prediction reliability calculation is performed, and risk assessment and impact analysis are conducted based on the prediction reliability to construct a third risk component. The reverse current risk index is obtained by weighted summation of the first risk component, the second risk component, and the third risk component.

7. The millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario as described in claim 1, characterized in that, The risk level is classified according to the aforementioned reverse current risk index, and multiple control levels are determined. The method includes: The energy management controller stores a first risk threshold and a second risk threshold, wherein the first risk threshold is less than the second risk threshold; The reverse current risk index is compared with the first risk threshold. If the reverse current risk index is less than the first risk threshold, it is determined to be a mild risk level and the first control level is activated. If the reverse current risk index is greater than or equal to the first risk threshold, the reverse current risk index is compared with the second risk threshold. If the reverse current risk index is less than the second risk threshold, it is determined to be a medium risk level and the second control level is activated. If the reverse current risk index is greater than or equal to the second risk threshold, it is determined to be a severe risk level, and the third control level is activated.

8. The millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario as described in claim 7, characterized in that, Multiple control levels are identified for optimization, and a cooperative inhibition decomposition scheme is formulated. The methods include: When the first control level is activated, the device combination is determined to be only calling the energy storage unit for energy storage regulation optimization, and the energy storage regulation power component is obtained; The constraint solution is performed based on the energy storage regulation power component to generate a reverse power flow suppression task. When the second control level is activated, the equipment combination is determined to be simultaneously calling the energy storage unit and the charging pile group for group energy storage regulation optimization, and the charging pile group regulation power component is obtained. The constraint solution is performed based on the power component of the pile group adjustment to generate the pile group load component adjustment task. When the third control level is activated, the equipment combination is determined to simultaneously call the energy storage unit, the charging pile group and the photovoltaic inverter unit to perform photovoltaic derating adjustment optimization, and obtain the photovoltaic derating power component. The constraint solution is performed based on the photovoltaic derated power component to generate the output derated task; The reverse current suppression task, the pile group load component adjustment task, and the output reduction task are integrated to formulate the collaborative suppression decomposition scheme.

9. The millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging collaborative scenario as described in claim 8, characterized in that, The method includes executing the cooperative suppression decomposition scheme and issuing power control commands to perform millisecond-level prediction and suppression of reverse power flow. The energy storage regulation power component of the reverse power flow suppression task is read and algebraically summed with the actual charging and discharging power set value of the energy storage converter to generate the first power control command. Read the pile group load component adjustment power component of the pile group adjustment task, perform charging power reduction decomposition, and obtain the first power limit value; The first power limit value is used for verification. When the verification passes, a second power control command is generated. Read the photovoltaic derated power component of the output derated task, perform available maximum analysis based on the photovoltaic inverter unit, extract the maximum available output power, and generate a second power limit value by subtracting the maximum available output power from the photovoltaic derated power component. Based on the second power limit value, the operating point is tracked to generate a third power control command; The first power control command, the second power control command, and the third power control command are issued in parallel for millisecond-level response collaborative prediction and suppression.

10. A millisecond-level prediction and suppression device for reverse power flow in a photovoltaic-storage-charging collaborative scenario, characterized in that, The step of implementing the millisecond-level prediction and suppression method for reverse power flow in a photovoltaic-storage-charging coordinated scenario according to any one of claims 1 to 9, wherein the millisecond-level prediction and suppression device for reverse power flow in a photovoltaic-storage-charging coordinated scenario comprises: The data acquisition module is used to acquire multi-source operational data streams in real time through the energy management controller; The trend prediction module is used to predict the grid connection point based on the multi-source operation data stream combined with the photovoltaic-storage-charging collaborative scenario, and to obtain the net power trend of the grid connection point and the prediction uncertainty index. The risk assessment module is used to preset a reverse power flow threshold and calculate a reverse power flow risk index by comparing the net power trend at the grid connection point with the reverse power flow threshold. The classification module is used to classify the risk levels according to the reverse current risk index, determine multiple control levels for optimization, and formulate a collaborative suppression decomposition scheme. A millisecond-level parallel execution module is used to execute the cooperative suppression decomposition scheme and issue power control commands to perform millisecond-level prediction and suppression of reverse power flow.

Images