Intelligent optimal scheduling method and system for optical storage charging station

Abstract

The invention discloses an intelligent optimal scheduling method and system for a light storage charging station, and relates to the technical field of energy storage cooperative control, and the method comprises the steps: inputting a photovoltaic fluctuation feature vector into a cause recognition model to execute cause classification operation, obtaining a fluctuation cause label, intensity and predicted duration, and carrying out the short-time power flow prediction of voltage and current, calculating an active upper limit, a reactive range and an allowable voltage offset to form a dynamic grid-connected margin time sequence; inputting the energy storage working condition data into the life model, calculating the health degree state and the residual life, and obtaining a life constraint set and a combined constraint time sequence in combination with the life attenuation relationship and the dynamic grid-connected margin time sequence; and according to the dynamic grid-connected margin vector and the life constraint set, a candidate scheduling scheme set is obtained by adopting a multi-target scheduling function, rolling solution is performed on the photovoltaic limiting proportion sequence and the photovoltaic output prediction curve, and a scheduling instruction is generated. According to the method, the scheduling instruction which gives consideration to safety, stability and service life prolonging is obtained under the uncertain working condition.

Description

Technical Field

[0001] This invention relates to the field of energy storage collaborative control technology, and in particular to an intelligent optimization scheduling method and system for photovoltaic-energy storage charging stations. Background Technology

[0002] With the deep penetration of distributed photovoltaic (PV) power on the distribution side and the rapid expansion of public charging infrastructure, PV-storage charging stations are gradually becoming important carriers for new energy consumption, grid resilience enhancement, and transportation energy electrification. Currently, PV-storage charging systems generally include three types of power sources: PV arrays, energy storage units, and charging loads. Their operating characteristics are jointly affected by factors such as PV power fluctuations, grid operating conditions, and energy storage lifespan degradation. Due to the significant randomness and uncertainty of PV output, under conditions such as cloud cover, rapid changes in irradiance, or module thermal drift, inverter output and grid connection power flow fluctuations often exhibit dynamic characteristics that are multi-dimensional and coupled across time scales.

[0003] Existing photovoltaic-storage-charging station scheduling still suffers from two technical shortcomings: First, traditional scheduling lacks a structured characterization of the causes, intensity, and persistence of fluctuations, making it difficult for generation limiting strategies and energy storage regulation strategies to respond differently to different types of fluctuations. Second, existing literature typically simplifies energy storage lifetime factors to capacity decay costs or charge / discharge depth penalties, preventing the scheduling process from simultaneously incorporating lifetime decay and grid friendliness into a unified constraint system. This exposes energy storage to the risk of overuse or unreasonable peak-shaving allocation in long-term operation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an intelligent optimization scheduling method for photovoltaic-storage-charging stations to solve the problem that photovoltaic-storage-charging systems are unable to balance grid connection safety and energy storage lifespan under uncertain operating conditions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an intelligent optimization scheduling method for photovoltaic-storage charging stations, comprising: collecting operating status data of photovoltaic-storage charging stations and performing alignment and numerical normalization processing to form photovoltaic fluctuation feature vectors and energy storage operating condition data; inputting the photovoltaic fluctuation feature vectors into a cause identification model to perform cause classification operations to obtain fluctuation cause labels, intensity and expected duration, and performing short-term power flow prediction on voltage and current to calculate active power limit, reactive power range and allowable voltage deviation to form a dynamic grid connection margin time series; inputting the energy storage operating condition data into a lifetime model to calculate health status and remaining lifetime, and combining lifetime decay relationship and dynamic grid connection margin time series to obtain lifetime constraint set and joint constraint time series; and using a multi-objective scheduling function to obtain a set of candidate scheduling schemes based on the dynamic grid connection margin vector and lifetime constraint set, and performing rolling solution on photovoltaic curtailment ratio sequence and photovoltaic output prediction curve to generate scheduling instructions.

[0007] As a preferred embodiment of the intelligent optimization scheduling method for photovoltaic-storage charging stations described in this invention, the operating status data of the photovoltaic-storage charging station includes irradiance, component temperature, inverter active power, inverter reactive power, grid connection point voltage, grid connection point current, energy storage state of charge, energy storage temperature, and energy storage cycle count.

[0008] As a preferred embodiment of the intelligent optimization scheduling method for photovoltaic-energy storage charging stations described in this invention, the specific steps for forming the photovoltaic fluctuation feature vector and energy storage operating condition data are as follows: The operation status data of the photovoltaic and energy storage charging station is time-aligned, and the missing time records are interpolated and multiple records at the same time are merged to obtain the operation status data sequence of the photovoltaic and energy storage charging station. Statistical calculations are performed on the operating status data sequence of photovoltaic and energy storage charging stations to generate candidate data for photovoltaic fluctuation feature vectors and candidate data for energy storage operating conditions. Numerical normalization is performed on the candidate data of photovoltaic fluctuation feature vector and energy storage operating condition data to form photovoltaic fluctuation feature vector and energy storage operating condition data.

[0009] As a preferred embodiment of the intelligent optimization scheduling method for photovoltaic-storage charging stations described in this invention, the specific steps for inputting the photovoltaic fluctuation feature vector into the cause identification model to perform cause classification operations and obtain the fluctuation cause label, intensity, and expected duration are as follows. Based on the operational status data of photovoltaic-storage charging stations, and combined with a multi-layer feature mapping structure based on forward computation, the initial structure of the cause identification model is constructed, and the initial cause identification model is generated. Based on the initial structure of the cause identification model, the historical photovoltaic-storage-charging station operation status data is divided into a training set and a validation set to generate a training sample set. The training sample set is input into the initial cause identification model, the classification error output by the initial cause identification model is calculated, and the parameters of each feature mapping are iteratively updated to obtain the cause identification model. The photovoltaic fluctuation feature vector is input into the cause identification model, and forward calculation is performed to output the fluctuation cause label, fluctuation intensity and expected duration.

[0010] As a preferred embodiment of the intelligent optimization scheduling method for photovoltaic-storage-charging stations described in this invention, the specific steps for forming a dynamic grid connection margin time series are as follows: The grid connection point voltage and grid connection point current are combined to form short-time power flow prediction data. Based on the current power grid topology and node admittance matrix, rolling power flow calculation is performed to generate predicted values ​​for grid connection point voltage and grid connection point current. Based on the fluctuation cause label, fluctuation intensity and expected duration, the predicted values ​​of grid connection point voltage and grid connection point current are corrected according to the scenario classification, and the corresponding active power upper limit, reactive power range and allowable voltage deviation are calculated to form a dynamic grid connection margin vector and a grid connection constraint weight vector. The dynamic grid connection margin vector is arranged according to the scheduling cycle to form a dynamic grid connection margin time series.

[0011] As a preferred embodiment of the intelligent optimization scheduling method for photovoltaic-storage charging stations described in this invention, the specific steps for inputting energy storage operating condition data into the lifespan model to calculate the health status and remaining lifespan are as follows. Based on the energy storage state of charge, energy storage temperature, and energy storage cycle number in the energy storage operating condition data, construct the lifetime model input vector; Input the lifespan model input vector into the lifespan model to calculate the health status and remaining lifespan.

[0012] As a preferred embodiment of the intelligent optimization scheduling method for photovoltaic-storage charging stations described in this invention, the specific steps for obtaining the lifetime constraint set and the joint constraint time series are as follows: Based on health status and remaining lifetime, combined with lifetime decay relationship, the maximum charge / discharge power, power change rate and cycle count limit are obtained and combined into a lifetime constraint set; The lifetime constraint set is expanded according to the scheduling cycle to generate a lifetime constraint time series, which is then aligned with the dynamic grid connection margin time series on a unified time axis to form a joint constraint time series.

[0013] As a preferred embodiment of the intelligent optimization scheduling method for photovoltaic-storage charging stations described in this invention, the specific steps for obtaining the candidate scheduling scheme set are as follows: Based on the lifetime constraint set and joint constraint time series, the photovoltaic curtailment ratio and energy storage charging and discharging power of each scheduling cycle are used as variables to be determined. A multi-objective scheduling function is constructed and the value of the multi-objective scheduling function is calculated. Based on the fluctuation cause label and the grid connection constraint weight vector, the weight coefficients in the multi-objective scheduling function are adjusted to form a set of candidate scheduling schemes.

[0014] As a preferred embodiment of the intelligent optimization scheduling method for photovoltaic-storage charging stations described in this invention, the specific steps for generating scheduling instructions are as follows: The candidate scheduling scheme set is optimized in chronological order to select the photovoltaic power curtailment ratio sequence and the energy storage charging and discharging power sequence. The photovoltaic power curtailment ratio sequence is combined with the photovoltaic power output prediction curve to form a photovoltaic grid-connected power output reference curve. The energy storage charging and discharging power sequence is converted into an energy storage power reference curve, and dispatch instructions are generated.

[0015] Secondly, the present invention provides an intelligent optimization scheduling system for photovoltaic-storage charging stations, including: a data acquisition module for acquiring operating status data of photovoltaic-storage charging stations and performing alignment and numerical normalization processing to form photovoltaic fluctuation feature vectors and energy storage operating condition data; and a cause identification module for inputting photovoltaic fluctuation feature vectors into a cause identification model to perform cause classification calculations, obtaining fluctuation cause labels, intensity and expected duration, and performing short-term power flow prediction on voltage and current, calculating active power limit, reactive power range and allowable voltage deviation, and forming a dynamic grid connection margin time series. The lifetime calculation module is used to input energy storage operating condition data into the lifetime model, calculate the health status and remaining lifetime, and obtain the lifetime constraint set and joint constraint time series by combining the lifetime decay relationship and the dynamic grid connection margin time series. The optimization scheduling module is used to obtain a set of candidate scheduling schemes by using a multi-objective scheduling function based on the dynamic grid connection margin vector and the lifetime constraint set, and to perform rolling solution on the photovoltaic curtailment ratio sequence and photovoltaic output prediction curve to generate scheduling instructions.

[0016] The beneficial effects of this invention are as follows: by constructing a dynamic grid-connection margin time series based on fluctuation cause identification, the scheduling can adaptively adjust the photovoltaic curtailment and energy storage response according to the type of disturbance; and by aligning the lifetime constraint time series derived from health and remaining lifetime with the grid-connection margin to form a joint constraint, the scheduling can simultaneously meet the grid friendliness and lifetime protection requirements in the rolling solution, thereby obtaining a scheduling instruction that takes into account safety, stability and lifetime extension under uncertain operating conditions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart for an intelligent optimization scheduling method for photovoltaic-storage charging stations.

[0019] Figure 2 A schematic diagram of a smart optimization scheduling method system for photovoltaic and energy storage charging stations.

[0020] Figure 3 This is a flowchart for identifying the causes and forming dynamic grid connection margin.

[0021] Figure 4 This is a flowchart of joint constraint-driven optimization scheduling. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides an intelligent optimization scheduling method for photovoltaic-storage charging stations, comprising the following steps: S1: Collect the operating status data of the photovoltaic-storage charging station, and perform alignment and numerical normalization processing to form photovoltaic fluctuation feature vector and energy storage operating condition data; S1.1: The operating status data of the photovoltaic-storage charging station includes irradiance, component temperature, inverter active power, inverter reactive power, grid connection point voltage, grid connection point current, energy storage state of charge, energy storage temperature, and energy storage cycle count. Furthermore, at the photovoltaic-storage charging station site, irradiance acquisition devices, module temperature acquisition devices, power acquisition devices, voltage acquisition devices, current acquisition devices, and monitoring devices for monitoring the energy storage state of charge, energy storage temperature, and energy storage cycle count are set up. The operating parameters of the photovoltaic side and the energy storage side are continuously collected with a unified sampling period. During the acquisition process, time stamps and data source stamps are added to each sampling record. The measurement results of irradiance, module temperature, inverter active power, inverter reactive power, grid connection point voltage, grid connection point current, energy storage state of charge, energy storage temperature, and energy storage cycle count are summarized and stored by the data acquisition devices and uniformly named as photovoltaic-storage charging station operating status data.

[0026] S1.2: Time-align the operating status data of the photovoltaic-storage-charging station, and obtain the operating status data sequence of the photovoltaic-storage-charging station by interpolating missing time records and merging multiple records at the same time. Furthermore, the operation status data of the photovoltaic-storage-charging station is sorted according to time stamps. For each target time point, records with the same time stamp are searched in the operation status data of the photovoltaic-storage-charging station. If there is no corresponding record for a certain target time point, then, provided that there are records at adjacent time points, a linear interpolation value is calculated for the same physical quantity, and the linear interpolation value is written into a new record to complete the interpolation of missing time records. When there are multiple records for a certain target time point, the arithmetic mean of each physical quantity is calculated and written into a single record to complete the merging of multiple records at the same time. Finally, the operation status data sequence of the photovoltaic-storage-charging station is obtained by arranging them in chronological order.

[0027] S1.3: Perform statistical calculations based on the operating status data sequence of photovoltaic and energy storage charging stations to generate candidate data for photovoltaic fluctuation feature vectors and candidate data for energy storage operating conditions; Furthermore, a series of continuous time windows are constructed on the photovoltaic-storage charging station operation status data sequence with a fixed time window length and sliding step size. Within each time window, the mean, extreme values, rate of change, and fluctuation amplitude of statistics such as irradiance, module temperature, inverter active power, inverter reactive power, grid connection point voltage, and grid connection point current are calculated. The statistics related to the change in photovoltaic output are arranged and combined into candidate data for photovoltaic fluctuation feature vectors. Within the same time window, the mean, fluctuation amplitude, and cumulative change of statistics such as energy storage state of charge, energy storage temperature, and energy storage cycle number are calculated. The statistics related to the energy storage operation status are arranged and combined into candidate data for energy storage operating conditions.

[0028] S1.4: Perform numerical normalization on the candidate data of photovoltaic fluctuation feature vector and energy storage operating condition data to form photovoltaic fluctuation feature vector and energy storage operating condition data; Furthermore, the minimum and maximum values ​​of each component in the candidate data of photovoltaic fluctuation feature vector are statistically analyzed over all time windows, and numerical normalization is performed to obtain dimensionless results with values ​​ranging from zero to one. The dimensionless results are then used to form a new photovoltaic fluctuation feature vector and new energy storage operating condition data.

[0029] S2: Input the photovoltaic fluctuation feature vector into the cause identification model to perform cause classification operation, obtain the fluctuation cause label, intensity and expected duration, and perform short-time power flow prediction on voltage and current, calculate the active power limit, reactive power range and allowable voltage deviation, and form a dynamic grid-connection margin time series. S2.1: Based on the operating status data of the photovoltaic-storage charging station, and combined with the multi-layer feature mapping structure based on forward computation, the initial structure of the cause identification model is constructed, and the initial cause identification model is generated. Furthermore, based on the operational status data of the photovoltaic-storage charging station, each numerical component of the photovoltaic fluctuation feature vector is input to the input end of a multi-layer feature mapping structure in a fixed order. Inside the multi-layer feature mapping structure, several layers of forward computation units are sequentially set to extract the multi-level relationships of the photovoltaic fluctuation feature vector. At the output end, a fluctuation cause label output unit for indicating the disturbance source category, a fluctuation intensity output unit for indicating the disturbance amplitude, and an output unit for indicating the expected duration of the disturbance's persistence trend are set. These input, internal, and output structures form the initial structure of the cause identification model, enabling it to perform the cause identification task. When constructing the initial structure of the cause identification model, an input layer with the same dimension as the photovoltaic fluctuation feature vector is set at the input end. Following the input layer, a multi-layer feature mapping structure is sequentially set. Each layer performs weighted summation and nonlinear transformation through forward computation. At the output end, a classification output unit for outputting the fluctuation cause label and a numerical output unit for outputting the fluctuation intensity and expected duration are set. This multi-layer feature mapping structure generates the initial cause identification model.

[0030] S2.2: Based on the initial structure of the cause identification model, the historical photovoltaic-storage-charging station operation status data is divided into a training set and a validation set to generate a training sample set; Furthermore, the historical operating status data of photovoltaic and energy storage charging stations are organized, and the continuous data segments related to photovoltaic output fluctuations are divided into multiple sample segments in chronological order. Photovoltaic fluctuation feature vectors are extracted for each sample segment, and the correspondence between the photovoltaic fluctuation feature vectors and known fluctuation cause categories, known fluctuation intensities, and known durations is established. The sample set with labels of cause category, intensity, and duration is divided into training set and validation set. The photovoltaic fluctuation feature vectors and corresponding labels in the training set are organized into training sample set.

[0031] S2.3: Input the training sample set into the initial cause identification model, calculate the classification error output by the initial cause identification model, and iteratively update each feature mapping parameter to obtain the cause identification model; Furthermore, each photovoltaic fluctuation feature vector in the training sample set is input into the initial cause identification model. Forward computation is performed through a multi-layer feature mapping structure to obtain the predicted fluctuation cause label, fluctuation intensity, and predicted duration. The differences between these predictions and the actual fluctuation cause labels, actual fluctuation intensities, and actual durations in the training sample set are calculated. Based on these differences, the weight and bias parameters in the multi-layer feature mapping structure are iteratively updated, gradually bringing the output of the cause identification model closer to the actual labels. The cause identification model is formed after the iteration process meets the stopping condition.

[0032] S2.4: Input the photovoltaic fluctuation feature vector into the cause identification model, perform forward calculation, and output the fluctuation cause label, fluctuation intensity and expected duration; Furthermore, the photovoltaic fluctuation feature vector is input into the cause identification model, which performs forward calculation through a multi-layer feature mapping structure to output a fluctuation cause label to characterize the source of photovoltaic fluctuations, a fluctuation intensity to characterize the amplitude of photovoltaic fluctuations, and an expected duration to characterize the continued trend of photovoltaic fluctuations.

[0033] S2.5: Combine the grid connection point voltage and grid connection point current into short-time power flow prediction data, perform rolling power flow calculation based on the current power grid topology and node admittance matrix, and generate predicted values ​​for grid connection point voltage and grid connection point current. Furthermore, the grid-connected point voltage and grid-connected point current time series are arranged at fixed time intervals, and the two types of series are concatenated to form short-term power flow prediction data. The short-term power flow prediction data, along with the current grid topology and node admittance matrix, serve as inputs for power flow calculation. Within each rolling time window, the active and reactive power injection quantities from the short-term power flow prediction data are used as the basic values ​​for power flow solution. Through power flow calculation, the estimated node voltage and node current values ​​for several future scheduling cycles are obtained. The predicted grid-connected point voltage and grid-connected point current values ​​are then extracted from these estimated values, expressed as follows: ; ; in, Indicates the first Predicted current at grid connection point for each scheduling cycle Indicates the first scheduling nodes The predicted value of the grid connection point voltage. Indicates the first Predicted active power at grid connection points for each scheduling cycle Indicates the first Predicted reactive power at grid connection points for each scheduling cycle Indicates the first The conjugate complex value of the predicted voltage at the grid connection point for each scheduling cycle. Indicates the first Each scheduling cycle node The predicted value of the injection current, For nodes in the node admittance matrix Self-guided nanometer capacity, For nodes in the node admittance matrix and The mutual conductivity, It is the imaginary unit.

[0034] It should be noted that self-admittance is used to characterize the admittance characteristics of a node itself to current, and is obtained by superimposing the branch admittance and parallel admittance related to the node. It is used to describe the influence of voltage changes on the current of the node. Conjugate complex values ​​are used to handle the complex relationships between voltage, current and power in AC circuits. By taking the conjugate of the voltage complex value, a form matching the complex power calculation is obtained. The node admittance matrix is ​​used to describe the admittance coupling relationship between nodes in the entire power grid. Each matrix element represents the linear relationship between current and voltage between nodes and is the core parameter for power flow solution. Mutual admittance is used to characterize the admittance coupling strength between two different nodes. It is given by the line admittance connecting the two nodes and is used to describe the electrical mutual influence relationship between nodes.

[0035] S2.6: Based on the fluctuation cause label, fluctuation intensity and expected duration, perform scenario-level correction on the predicted values ​​of grid connection point voltage and grid connection point current, calculate the corresponding active power upper limit, reactive power range and allowable voltage deviation, and form a dynamic grid connection margin vector and grid connection constraint weight vector. Furthermore, based on the disturbance source category corresponding to the fluctuation cause label, the predicted values ​​of grid-connected point voltage and current are categorized and organized, so that each disturbance source corresponds to an independent voltage fluctuation amplitude range and current change range. This allows the grid-connected point voltage and current predicted values ​​to form hierarchical data sets for calculating active power limits, reactive power ranges, and allowable voltage offsets according to the disturbance category. Linear proportional adjustment is performed on the voltage and current predicted values ​​according to the fluctuation intensity and expected duration, so that different disturbance scenarios correspond to different voltage and current boundaries. Based on the processed voltage and current predicted values, the allowable active power limit, allowable reactive power range, and allowable voltage offset in the current scenario are calculated. These three types of constraints are combined into a dynamic grid-connected margin vector, and a grid-connected constraint weight vector is generated based on the scenario change degree in each scheduling cycle.

[0036] S2.7: Arrange the dynamic grid connection margin vector according to the scheduling cycle to form a dynamic grid connection margin time series; Furthermore, the dynamic grid connection margin vector is arranged in chronological order according to the scheduling cycle, and the arranged dynamic grid connection margin vector is sequentially concatenated into a dynamic grid connection margin time series covering multiple scheduling cycles. Each time position in the dynamic grid connection margin time series includes the active power limit, reactive power range, and allowable voltage offset.

[0037] S3: Input the energy storage operating condition data into the lifetime model, calculate the health status and remaining lifetime, and combine the lifetime decay relationship and dynamic grid connection margin time series to obtain the lifetime constraint set and joint constraint time series; S3.1: Construct the lifetime model input vector based on the energy storage state of charge, energy storage temperature, and energy storage cycle count in the energy storage operating condition data; Furthermore, in the energy storage operating condition data, the energy storage state of charge, energy storage temperature, and energy storage cycle count are read for each scheduling cycle. The energy storage state of charge is taken as the first component, the energy storage temperature as the second component, and the energy storage cycle count as the third component. These components are then concatenated in a fixed order to form an ordered set of values, which is named the lifetime model input vector.

[0038] S3.2: Input the lifespan model input vector into the lifespan model to calculate the health status and remaining lifespan; Furthermore, within each scheduling cycle, the lifetime model input vector is fed into the lifetime model for lifetime estimation. After receiving the lifetime model input vector, the lifetime model first performs fatigue accumulation and thermal stress assessment on the energy storage state of charge component and energy storage temperature component, and then performs cycle accumulation damage assessment in combination with the energy storage cycle number component. Through the mapping relationship within the lifetime model, the health status and remaining lifetime results for the corresponding scheduling cycle are given.

[0039] S3.3: Based on the health status and remaining lifetime, combined with the lifetime decay relationship, obtain the maximum charge / discharge power, power change rate and cycle count limit, and combine them into a lifetime constraint set; Furthermore, using health status as a quantitative indicator reflecting the current available lifetime margin and remaining lifetime as a quantitative indicator reflecting the remaining available cyclic resources, both health status and remaining lifetime are input into the lifetime decay relationship description. Through the lifetime decay relationship, health status and remaining lifetime are converted into the maximum sustainable charge / discharge power, acceptable power change rate, and allowable cycle count limit. The maximum charge / discharge power is used to limit the amplitude range of energy storage charging and discharging power within each scheduling cycle; the power change rate is used to limit the gradient of energy storage charge / discharge power changes between adjacent scheduling cycles; and the cycle count limit is used to limit the number of charge / discharge cycles allowed within a scheduling cycle. The maximum charge / discharge power, power change rate, and cycle count limit are concatenated in a fixed order to form a lifetime constraint set.

[0040] S3.4: Expand the lifetime constraint set according to the scheduling cycle to generate a lifetime constraint time series, and align it with the dynamic grid connection margin time series on a unified time axis to form a joint constraint time series; Furthermore, using the scheduling cycle as an index, the maximum charge / discharge power, power change rate, and cycle count limits contained in the lifetime constraint set are expanded in chronological order. For each scheduling cycle, the corresponding three types of lifetime constraints are recorded, forming a lifetime constraint time series arranged by scheduling cycle. The lifetime constraint time series and the dynamic grid connection margin time series are aligned on the time axis using a unified scheduling cycle length and a unified start time. The lifetime constraint time series component of each scheduling cycle is combined with the dynamic grid connection margin time series component of the same scheduling cycle to form a constraint set that simultaneously considers lifetime constraints and grid connection constraints within that scheduling cycle. The constraint sets corresponding to all scheduling cycles are arranged to form a joint constraint time series.

[0041] S4: Based on the lifetime constraint set and the joint constraint time series, a multi-objective scheduling function is used to obtain a set of candidate scheduling schemes, and a rolling solution is performed on the photovoltaic curtailment ratio sequence and the photovoltaic output prediction curve to generate scheduling instructions; S4.1: Based on the lifetime constraint set and the joint constraint time series, the photovoltaic curtailment ratio and energy storage charging and discharging power of each scheduling cycle are used as variables to be determined, a multi-objective scheduling function is constructed, and the value of the multi-objective scheduling function is calculated; Furthermore, based on the lifetime constraint set and the joint constraint time series, two variables to be determined are set for each scheduling cycle: the photovoltaic curtailment ratio and the energy storage charging and discharging power. For each combination of variables, three evaluation quantities are calculated: photovoltaic curtailment amount, grid connection point power fluctuation amplitude, and lifetime consumption cost. The photovoltaic curtailment amount, grid connection point power fluctuation amplitude, and lifetime consumption cost are weighted and superimposed according to preset weights to form a multi-objective scheduling function. The multi-objective scheduling function value is obtained by calculating the multi-objective scheduling function.

[0042] S4.2: Based on the fluctuation cause label and the grid connection constraint weight vector, adjust the weight coefficients in the multi-objective scheduling function to form a set of candidate scheduling schemes; Furthermore, using the disturbance source category corresponding to the fluctuation cause label as the distinguishing criterion, disturbance sources belonging to the rapidly changing category and those belonging to the slowly changing category are respectively assigned different emphases in photovoltaic curtailment evaluation and power fluctuation evaluation. Then, the change amplitude of grid connection safety constraints corresponding to each component in the grid connection constraint weight vector is used as an adjustment reference to ensure a quantitative mapping relationship between the weight coefficients of photovoltaic curtailment, grid connection point power fluctuation amplitude, and lifetime consumption cost in the multi-objective scheduling function and the disturbance category and grid connection safety constraint change amplitude. This provides a basis for the evaluation of photovoltaic curtailment in the multi-objective scheduling function. The weighting coefficients corresponding to the amount of curtailed photovoltaic power, the power fluctuation amplitude at the grid connection point, and the lifetime consumption cost are adjusted to make the multi-objective scheduling function pay more attention to the power fluctuation amplitude at the grid connection point in high-risk scenarios, more attention to the lifetime consumption cost in scenarios with low lifetime margin, and more attention to the amount of curtailed photovoltaic power in scenarios with severe curtailment. Subsequently, within the feasible range defined by the lifetime constraint set and the joint constraint time series, the multi-objective scheduling function value is calculated by combining multiple sets of values ​​for the photovoltaic curtailment ratio and the energy storage charging and discharging power. The variable combinations whose multi-objective scheduling function values ​​are in the better range are selected and organized into a candidate scheduling scheme set.

[0043] S4.3: Perform rolling optimization on the candidate scheduling scheme set in chronological order to select the photovoltaic curtailment ratio sequence and the energy storage charging and discharging power sequence; Furthermore, rolling time windows are established according to the scheduling cycle time sequence. Within each rolling time window, candidate scheduling schemes belonging to the current rolling time window are extracted from the candidate scheduling scheme set. The multi-objective scheduling function value of each candidate scheduling scheme is recalculated and evaluated in conjunction with the scheduling results already selected in the previous rolling time window. This avoids excessive jumps in photovoltaic curtailment ratio and energy storage charging and discharging power in adjacent rolling time windows. The photovoltaic curtailment ratio and energy storage charging and discharging power values ​​in each candidate scheduling scheme are compared for consistency with the corresponding constraints in the lifetime constraint set and joint constraint time series. Only candidate scheduling schemes that meet the constraint range are retained. The magnitude of the change in photovoltaic curtailment ratio and energy storage charging and discharging power between the multi-objective scheduling function value of each candidate scheduling scheme and the scheduling scheme already determined in the previous rolling time window is compared. Candidate scheduling schemes with change magnitudes within the allowable range defined by the lifetime constraint set and joint constraint time series and smaller multi-objective scheduling function values ​​are selected and combined in chronological order to form a photovoltaic curtailment ratio sequence and an energy storage charging and discharging power sequence.

[0044] S4.4: Combine the photovoltaic power curtailment ratio sequence with the photovoltaic power output prediction curve to form a photovoltaic grid-connected power output reference curve, convert the energy storage charging and discharging power sequence into an energy storage power reference curve, and generate dispatch instructions; Furthermore, within each scheduling cycle, the photovoltaic (PV) power curtailment ratio corresponding to the scheduling cycle in the PV power curtailment ratio sequence is multiplied by the PV power output prediction value corresponding to the scheduling cycle in the PV power output prediction curve to obtain the PV grid-connected power output reference value for that scheduling cycle. The PV grid-connected power output reference values ​​for all scheduling cycles are arranged in chronological order to form a PV grid-connected power output reference curve. Simultaneously, the energy storage charging and discharging power in each scheduling cycle of the energy storage charging and discharging power sequence is used as the energy storage active power reference value. Combined with power factor requirements or reactive power control strategies, the energy storage reactive power reference value is derived. The energy storage active power reference value and the energy storage reactive power reference value are arranged in chronological order to form an energy storage power reference curve. Within each scheduling cycle, the values ​​at corresponding positions of the PV grid-connected power output reference curve and the energy storage power reference curve are converted into active power command values ​​and reactive power command values ​​for the PV inverter and the energy storage converter, and combined to form a scheduling command.

[0045] In summary, this invention constructs a dynamic grid-connection margin time series based on fluctuation cause identification, enabling scheduling to adaptively adjust photovoltaic curtailment and energy storage responses according to disturbance type; and aligns the lifetime constraint time series derived from health and remaining lifetime with the grid-connection margin to form a joint constraint, so that scheduling can simultaneously meet grid friendliness and lifetime protection requirements in rolling solution, thereby obtaining scheduling instructions that take into account safety, stability and lifetime extension under uncertain operating conditions.

[0046] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent optimization scheduling of photovoltaic-storage charging stations, characterized in that: include, Collect the operating status data of the photovoltaic-storage charging station, and perform alignment and numerical normalization processing to form photovoltaic fluctuation feature vectors and energy storage operating condition data; The photovoltaic fluctuation feature vector is input into the cause identification model to perform cause classification operation, and the fluctuation cause label, intensity and expected duration are obtained. Short-time power flow prediction is performed on voltage and current to calculate the active power limit, reactive power range and allowable voltage deviation, forming a dynamic grid-connection margin time series. Input energy storage operating condition data into the lifetime model to calculate health status and remaining lifetime, and combine lifetime decay relationship and dynamic grid connection margin time series to obtain lifetime constraint set and joint constraint time series; Based on the dynamic grid-connection margin vector and lifetime constraint set, a multi-objective scheduling function is used to obtain a set of candidate scheduling schemes, and a rolling solution is performed on the photovoltaic curtailment ratio sequence and photovoltaic output prediction curve to generate scheduling instructions.

2. The intelligent optimization scheduling method for photovoltaic-storage charging stations as described in claim 1, characterized in that: The operating status data of the photovoltaic-storage charging station includes irradiance, component temperature, inverter active power, inverter reactive power, grid connection point voltage, grid connection point current, energy storage state of charge, energy storage temperature, and energy storage cycle count.

3. The intelligent optimization scheduling method for photovoltaic-storage charging stations as described in claim 2, characterized in that: The specific steps for forming the photovoltaic fluctuation feature vector and energy storage operating condition data are as follows. The operation status data of the photovoltaic and energy storage charging station is time-aligned, and the missing time records are interpolated and multiple records at the same time are merged to obtain the operation status data sequence of the photovoltaic and energy storage charging station. Statistical calculations are performed on the operating status data sequence of photovoltaic and energy storage charging stations to generate candidate data for photovoltaic fluctuation feature vectors and candidate data for energy storage operating conditions. Numerical normalization is performed on the candidate data of photovoltaic fluctuation feature vector and energy storage operating condition data to form photovoltaic fluctuation feature vector and energy storage operating condition data.

4. The intelligent optimization scheduling method for photovoltaic-storage charging stations as described in claim 3, characterized in that: The process involves inputting the photovoltaic fluctuation feature vector into the cause identification model to perform cause classification operations, obtaining the fluctuation cause label, intensity, and expected duration. The specific steps are as follows: Based on the operational status data of photovoltaic-storage charging stations, and combined with a multi-layer feature mapping structure based on forward computation, the initial structure of the cause identification model is constructed, and the initial cause identification model is generated. Based on the initial structure of the cause identification model, the historical photovoltaic-storage-charging station operation status data is divided into a training set and a validation set to generate a training sample set. The training sample set is input into the initial cause identification model, the classification error output by the initial cause identification model is calculated, and the feature mapping parameters are iteratively updated to obtain the cause identification model. The photovoltaic fluctuation feature vector is input into the cause identification model, and forward calculation is performed to output the fluctuation cause label, fluctuation intensity and expected duration.

5. The intelligent optimization scheduling method for photovoltaic-storage charging stations as described in claim 4, characterized in that: The specific steps for forming the dynamic grid connection margin time series are as follows: The grid connection point voltage and grid connection point current are combined to form short-time power flow prediction data. Based on the current power grid topology and node admittance matrix, rolling power flow calculation is performed to generate predicted values ​​for grid connection point voltage and grid connection point current. Based on the fluctuation cause label, fluctuation intensity and expected duration, the predicted values ​​of grid connection point voltage and grid connection point current are corrected according to the scenario classification, and the corresponding active power upper limit, reactive power range and allowable voltage deviation are calculated to form a dynamic grid connection margin vector and a grid connection constraint weight vector. The dynamic grid connection margin vector is arranged according to the scheduling cycle to form a dynamic grid connection margin time series.

6. The intelligent optimization scheduling method for photovoltaic-storage charging stations as described in claim 5, characterized in that: The specific steps for inputting energy storage operating condition data into the lifespan model to calculate the health status and remaining lifespan are as follows. Based on the energy storage state of charge, energy storage temperature, and energy storage cycle number in the energy storage operating condition data, construct the lifetime model input vector; Input the lifespan model input vector into the lifespan model to calculate the health status and remaining lifespan.

7. The intelligent optimization scheduling method for photovoltaic-storage charging stations as described in claim 6, characterized in that: The specific steps for obtaining the lifetime constraint set and the joint constraint time series are as follows: Based on health status and remaining lifetime, combined with lifetime decay relationship, the maximum charge / discharge power, power change rate and cycle count limit are obtained and combined into a lifetime constraint set; The lifetime constraint set is expanded according to the scheduling cycle to generate a lifetime constraint time series, which is then aligned with the dynamic grid connection margin time series on a unified time axis to form a joint constraint time series.

8. The intelligent optimization scheduling method for photovoltaic-storage charging stations as described in claim 7, characterized in that: The specific steps for obtaining the candidate scheduling scheme set are as follows: Based on the lifetime constraint set and joint constraint time series, the photovoltaic curtailment ratio and energy storage charging and discharging power of each scheduling cycle are used as variables to be determined. A multi-objective scheduling function is constructed and the value of the multi-objective scheduling function is calculated. Based on the fluctuation cause label and the grid connection constraint weight vector, the weight coefficients in the multi-objective scheduling function are adjusted to form a set of candidate scheduling schemes.

9. The intelligent optimization scheduling method for photovoltaic-storage charging stations as described in claim 8, characterized in that: The specific steps for generating scheduling instructions are as follows: The candidate scheduling scheme set is optimized in chronological order to select the photovoltaic power curtailment ratio sequence and the energy storage charging and discharging power sequence. The photovoltaic power curtailment ratio sequence is combined with the photovoltaic power output prediction curve to form a photovoltaic grid-connected power output reference curve. The energy storage charging and discharging power sequence is converted into an energy storage power reference curve, and dispatch instructions are generated.

10. A smart optimization scheduling system for photovoltaic-storage-charging stations, based on the smart optimization scheduling method for photovoltaic-storage-charging stations according to any one of claims 1 to 9, characterized in that: include, The data acquisition module is used to collect the operating status data of the photovoltaic-storage charging station, and perform alignment and numerical normalization processing to form photovoltaic fluctuation feature vectors and energy storage operating condition data. The cause identification module is used to input the photovoltaic fluctuation feature vector into the cause identification model to perform cause classification operation, obtain the fluctuation cause label, intensity and expected duration, and perform short-time power flow prediction on voltage and current, calculate the active power limit, reactive power range and allowable voltage deviation, and form a dynamic grid-connection margin time series. The lifetime calculation module is used to input energy storage operating condition data into the lifetime model, calculate the health status and remaining lifetime, and combine the lifetime decay relationship and dynamic grid connection margin time series to obtain the lifetime constraint set and joint constraint time series. The optimization scheduling module is used to obtain a set of candidate scheduling schemes based on the dynamic grid connection margin vector and lifetime constraint set, and to perform rolling solution on the photovoltaic curtailment ratio sequence and photovoltaic output prediction curve to generate scheduling instructions.

Images