Photovoltaic storage direct flexible power distribution method and system, electronic equipment, and storage medium

Abstract

The present application provides a photovoltaic storage direct flexible power distribution method and system, electronic equipment, and storage medium, which belongs to the field of new energy power systems. The method includes: obtaining the target parameters of the photovoltaic storage direct flexible power distribution system in the current time period, the target parameters include: system operating parameters, light intensity and grid load, and the system operating parameters include the power generation power, charge state, charge and discharge power, voltage and current, and grid connection parameters of the photovoltaic storage direct flexible power distribution system; inputting the target parameters of the photovoltaic storage direct flexible power distribution system in the current time period into a prediction model for parameter prediction to obtain the system operating parameters in the future time period; and controlling the operation of the photovoltaic storage direct flexible power distribution system based on the system operating parameters in the future time period. The photovoltaic storage direct flexible power distribution method and system, electronic equipment, and storage medium provided in the present application can improve the working efficiency and service life of the photovoltaic storage direct flexible power distribution system.

Description

Technical Field

[0001] This application belongs to the field of new energy power system technology, and more specifically, it relates to a photovoltaic-storage-DC-flexible power distribution method and system, electronic equipment, and storage medium. Background Technology

[0002] The photovoltaic-storage-DC-flexible power distribution system is a DC power distribution system that connects photovoltaic power generation systems, energy storage systems, and electrical equipment, and achieves flexible regulation through intelligent control. As an important measure for optimizing building energy consumption structure, the photovoltaic-storage-DC-flexible power distribution system is attracting increasing attention and research. This system transforms buildings from simple electricity consumption and passive energy saving to green power generation, flexible energy storage, and efficient electricity consumption. Simultaneously, it can connect to the AC power grid, exchanging AC and DC power, making the photovoltaic-storage-DC-flexible power distribution system a comprehensive system interacting with both DC and AC power grids. It represents a crucial technological route for the development of new building energy systems and zero-carbon buildings.

[0003] However, since the photovoltaic-storage-DC-flexible power distribution system is still in the development stage, how to control its operation has become a key issue. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for photovoltaic-storage-direct-flexible power distribution, electronic equipment, and storage medium that can control the operation of the photovoltaic-storage-direct-flexible power distribution system to improve its working efficiency and service life.

[0005] A first aspect of this application provides a method for photovoltaic-storage-DC-flexible power distribution, comprising:

[0006] Obtain the target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period. The target parameters include: system operating parameters, solar irradiance, and grid load. The system operating parameters include the power generation, state of charge, charging and discharging power, voltage, current, and grid connection parameters of the photovoltaic-storage-DC-flexible power distribution system.

[0007] Input the target parameters of the photovoltaic-storage-DC-flexible power distribution system in the current time period into the prediction model to predict the parameters and obtain the system operation parameters in the future time period.

[0008] Control the operation of the photovoltaic-storage-DC-flexible power distribution system based on system operating parameters for a future time period;

[0009] The parameter prediction methods include:

[0010] The target parameters are subjected to feature extraction to obtain a feature vector;

[0011] The feature vectors are encoded based on temporal features to obtain encoded data;

[0012] Feature interaction processing is performed on the encoded data to obtain interactive information;

[0013] The system's operating parameters for future time periods are obtained based on the interactive information.

[0014] A second aspect of this application provides a photovoltaic-storage-DC-flexible power distribution system, comprising:

[0015] The parameter acquisition module is used to acquire the target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period. The target parameters include: system operating parameters, solar irradiance, and grid load. The system operating parameters include the power generation, state of charge, charging and discharging power, voltage, current, and grid connection parameters of the photovoltaic-storage-DC-flexible power distribution system.

[0016] The parameter prediction module is used to input the target parameters of the photovoltaic-storage-DC-flexible power distribution system into the prediction model for the current time period to predict the system operating parameters for the future time period.

[0017] The system operation control module is used to control the operation of the photovoltaic-storage-DC-flexible power distribution system based on the system operation parameters within a future time period.

[0018] Specifically, the parameter prediction module is used during parameter prediction to:

[0019] The target parameters are subjected to feature extraction to obtain a feature vector;

[0020] The feature vectors are encoded based on temporal features to obtain encoded data;

[0021] Feature interaction processing is performed on the encoded data to obtain interactive information;

[0022] The system's operating parameters for future time periods are obtained based on the interactive information.

[0023] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described photovoltaic-storage-DC-flexible power distribution method.

[0024] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described optical-storage-DC-flexible power distribution method.

[0025] The beneficial effects of the photovoltaic-storage direct-drive-flexible power distribution method and system, electronic equipment, and storage medium provided in the embodiments of this application are as follows:

[0026] The photovoltaic-storage-DC-flexible power distribution method and system, electronic equipment, and storage medium provided in this application embodiment acquire key indicators such as system operating parameters, irradiance, and grid load through multi-source data acquisition, establishing a comprehensive system state perception capability. Subsequently, a predictive model architecture is adopted, combining time-series feature encoding and feature interaction technology to achieve high-precision prediction of the future operating state of the system. In this application embodiment, by combining the changes of system operating parameters, irradiance, and grid load over time, as well as the relationships between system operating parameters and the relationships between system operating parameters, irradiance, and grid load, the future operating state of the system can be predicted. This can solve key technical problems such as photovoltaic consumption, energy storage optimization, and grid interaction in photovoltaic-storage-DC-flexible power distribution systems, thereby improving the accuracy of predicting system operating parameters in the future time period, and thus improving the working efficiency and service life of photovoltaic-storage-DC-flexible power distribution systems. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A schematic flowchart of a photovoltaic-storage-DC-flexible power distribution method provided in an embodiment of this application;

[0029] Figure 2 A schematic flowchart of another photovoltaic-storage-DC-flexible power distribution method provided in an embodiment of this application;

[0030] Figure 3 This is a structural block diagram of a photovoltaic-storage-DC-flexible power distribution system provided in an embodiment of this application;

[0031] Figure 4 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0032] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0033] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0034] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a photovoltaic-storage-DC-flexible power distribution method according to an embodiment of this application. The method is executed by an electronic device and may include:

[0035] S101: Obtain the target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period. The target parameters include: system operating parameters, solar irradiance, and grid load. The system operating parameters include the power generation, state of charge, charging and discharging power, voltage, current, and grid connection parameters of the photovoltaic-storage-DC-flexible power distribution system.

[0036] In this embodiment, the power generation of the photovoltaic-storage-direct-flexible power distribution system can be affected by light intensity, temperature, and photovoltaic module efficiency. The monitoring system built into the photovoltaic-storage-direct-flexible power distribution system can be used to sample the above parameters in real time, with a sampling period of 0.1ms.

[0037] Specifically, solar irradiance refers to the solar radiation power received per unit area, which directly affects the short-circuit current of photovoltaic panels. It is influenced by factors such as weather, season, clouds, and obstructions.

[0038] The power generation capacity can be the real-time output power of the photovoltaic array, reflecting the power generation capacity of the photovoltaic system. The real-time power of the photovoltaic-storage-DC-flexible power distribution system can be monitored by the built-in power monitoring module of the power generation system.

[0039] State of Charge (SOC) is the ratio of a battery's remaining charge to its rated capacity at a given moment, usually expressed as a percentage (0%~100%). It is a core parameter for measuring battery health and optimizing energy management strategies. For example, a mobile phone displaying "20% remaining charge" means SOC = 20%. The SOC of a photovoltaic-storage-DC-flexible power distribution system should be maintained within a safe range (e.g., 20%~80%) to extend its lifespan.

[0040] The charging and discharging power can determine the direction of energy flow for energy storage batteries and can be obtained through some monitoring devices, such as energy storage converters.

[0041] The voltage and current can be the voltage and current of the photovoltaic-storage-DC-flexible power distribution system. For example, the voltage of the photovoltaic-storage-DC-flexible power distribution system can be collected by a voltmeter, and the current of the photovoltaic-storage-DC-flexible power distribution system can be collected by an ammeter.

[0042] Grid connection parameters can be phase voltage, phase current, and grid power, which can be obtained, for example, through devices such as smart meters and grid-connected inverters at the grid connection point.

[0043] Specifically, phase voltage is the voltage between the live wire and the neutral wire, which is very common in photovoltaic-storage grid-connected systems. Phase current is the real-time current of each phase in a three-phase system, reflecting the magnitude of power transfer. Grid power can include active power, reactive power, and apparent power.

[0044] S102: Input the target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period into the prediction model to predict the parameters and obtain the system operating parameters for the future time period.

[0045] In this embodiment, the prediction model can be a trained improved transformer neural network, or other network models, which are not limited here.

[0046] Specifically, the target parameters can be input into an improved transformer neural network to obtain the system operating parameters of the photovoltaic-storage-DC-flexible power distribution system for a future time period. Specifically, the target data can be normalized based on the time series data to obtain processed target data, and the statistical characteristics and rate of change characteristics corresponding to these parameters in the system operating parameters can be calculated using a sliding window operation. Secondly, relevant characteristics are calculated based on the normalized system operating parameters, solar irradiance, and grid load. Finally, the normalized target parameters, the statistical characteristics, rate of change characteristics, and correlation characteristics corresponding to various parameters in the system operating parameters can be concatenated to obtain a feature vector. This feature vector is then subjected to one-hot encoding to obtain time feature codes. Finally, information interaction operations are performed on the encoded data based on a multi-head attention mechanism to obtain interactive information, thus obtaining the system operating parameters for the future time period.

[0047] The parameter prediction method in step S102 is as follows: Figure 2 As shown, it can specifically include:

[0048] S1021: Extract features from the target parameters to obtain the feature vector.

[0049] In this embodiment, feature extraction is performed on the target parameters to obtain a feature vector. Specifically, firstly, time series alignment is performed on various parameters. Then, for each type of parameter in the system operating parameters, based on window sliding, the statistical features and rate of change features corresponding to that type of parameter in the system operating parameters are calculated. Secondly, relevant features are calculated based on the normalized system operating parameters, light intensity, and grid load. Finally, the normalized target parameters, the statistical features and rate of change features corresponding to various parameters in the system operating parameters, the first correlation feature, the second correlation feature, and the third correlation feature are concatenated to obtain the feature vector. In this embodiment, through feature extraction, the original parameters are transformed into a high-information-density, low-dimensional, model-readable feature vector, which becomes a bridge connecting the physical system and the intelligent algorithm, ultimately supporting the adaptive optimization operation of the photovoltaic-storage-direct current-flexible system.

[0050] S1022: Encode the feature vector based on temporal features to obtain encoded data.

[0051] In this embodiment, time information is crucial for predicting system operating parameters in the prediction model of the photovoltaic-storage-DC-flexible power distribution system, because parameters such as photovoltaic power generation and grid load often exhibit obvious periodic variation patterns. To fully utilize time information, it is necessary to extract periodic features from the timestamps of the collected target parameters, encode them into a form recognizable by the model, and finally fuse them with other features to form a complete feature vector for prediction.

[0052] Specifically, since periodic information such as hours and days of the week are categorical data, they cannot be directly input into numerical models and therefore require encoding conversion. One-hot encoding (OHE) is a commonly used method that converts each category into a binary vector, where only the corresponding category position is 1, and the rest are 0. After completing the time feature encoding, all relevant features need to be concatenated to form the final input feature vector. Features can include: normalized target parameters, statistical features, rate of change features, and correlation features. For example, the normalized target parameters can be photovoltaic power generation, energy storage battery SOC, and grid load. Statistical features can be mean, variance, maximum, and minimum values. Rate of change features can be slope, difference, etc., which reflect the dynamic changes of the parameters. Correlation features include the first feature, the second feature, and the third feature. Finally, the above feature vector is encoded based on the time-series features to obtain the encoded data.

[0053] S1023: Perform feature interaction processing on the encoded data to obtain interaction information.

[0054] In this embodiment, the feature information in the encoded data can be enriched first. Specifically, a multi-head self-attention mechanism is used to operate independently on the encoded data to obtain multiple attention matrices with different representation capabilities.

[0055] Specifically, this mechanism first projects the encoded data into multiple independent representation subspaces, performing self-attention computation in parallel within each subspace to generate multiple attention matrices with different focuses. These matrices capture different types of important features and dependencies in the encoded data; for example, some attention heads may focus on the temporal variation patterns of light intensity, others on the fluctuation characteristics of grid load, and still others may specifically capture the interaction patterns between the energy storage system state and other parameters. Through this multi-perspective parallel processing, the model can overcome the limitations of a single attention mechanism and comprehensively understand the complex characteristics of system operation from different dimensions.

[0056] Secondly, these matrices can be cross-multiplied with the original encoded data to achieve deep fusion of feature information. This interactive process is not a simple weighted summation, but rather allows each attention matrix to guide the encoded data to be reorganized and strengthened in different feature dimensions in a targeted manner. For example, an attention head that focuses specifically on photovoltaic power generation features will enhance the feature components related to sunlight in the encoded data, while another attention head that focuses on load changes will highlight key information about electricity consumption patterns. This cross-multiplication interaction allows the feature representations learned in different subspaces to complement and verify each other, preserving the basic information of the original encoded data while injecting new feature associations through the attention mechanism.

[0057] S1024: Obtain system operating parameters for a future time period based on interactive information.

[0058] In this embodiment, interactive information is input into a transformer-based prediction model to obtain prediction parameters. The core of this application lies in leveraging the efficient modeling capability of the Transformer model for time-series interactive information. By capturing the long-term correlation of millisecond-level voltage fluctuations, the Transformer model can predict frequency disturbance trends in advance, driving the energy storage system to achieve proactive compensation control. Simultaneously, the model minimizes the root mean square error between the predicted parameters and the true values ​​through end-to-end training, and combines probabilistic prediction techniques to quantify uncertainty, providing robust decision support for high-volatility scenarios. Furthermore, the obtained system operating parameters for the future time period can also correspond to the input system operating parameters. For example, the system operating parameters for the future time period could correspond to the power generation, state of charge, charging and discharging power, voltage, current, and grid connection parameters of a photovoltaic-storage-DC-flexible power distribution system. Moreover, the specific values ​​for the future time period can be determined based on samples from the training process. For example, during training, the training samples might be the target parameters for a certain month in historical data and the target parameters for the following month. In this case, the future time period could be within the next month, without specific limitations.

[0059] S103: Control the operation of the photovoltaic-storage-DC-flexible power distribution system based on system operating parameters within a future time period.

[0060] In this embodiment, based on the predicted parameters, the photovoltaic-storage-DC-flexible power distribution system is controlled in advance, enabling the system to actively adjust to grid power fluctuations.

[0061] Specifically: When a sudden drop in photovoltaic output is predicted, the system increases the energy storage discharge power in advance or activates the backup power supply to compensate for the power shortfall. If a sudden increase in load is predicted, non-critical loads are pre-cut, and the energy storage discharge rate is increased to avoid grid overload. Combining time-of-use pricing signals and SOC safety constraints (20%~80%), the charging and discharging plan is optimized on a rolling basis; for example, charging to 80% during off-peak hours and discharging to 30% during peak hours. Based on the real-time status of the DC bus voltage, the power consumption factor of variable power equipment is dynamically adjusted. This application aims to minimize grid power tracking error by adjusting the current reference command of the converter in real time to ensure that the grid-connected power responds quickly to dispatch requirements.

[0062] As can be seen from the above, the predictive control method for photovoltaic-storage-DC-flexible power distribution system provided in this application achieves multi-source data fusion prediction and active adjustment by improving the Transformer neural network. It can accurately predict the system operating parameters for future time periods, thereby improving the operating efficiency of the photovoltaic-storage-DC-flexible power distribution system, reducing economic costs, and achieving high-precision tracking of grid power. This enables the system to have the ability to autonomously adjust to cope with high fluctuation scenarios and extends the service life of the system.

[0063] In one embodiment of this application, to further improve the accuracy of controlling the photovoltaic-storage-DC-flexible power distribution system, the method may further include: obtaining the charging and discharging state of the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system during the current time period; determining a target control strategy based on the charging and discharging state; the target control strategy is used to control the energy storage battery of the photovoltaic-storage-DC-flexible power distribution system; wherein, controlling the operation of the photovoltaic-storage-DC-flexible power distribution system based on the system operating parameters in the future time period includes: controlling the operation of the photovoltaic-storage-DC-flexible power distribution system based on the system operating parameters and the target control strategy.

[0064] In this embodiment, the charging and discharging state of the energy storage battery within the current time period can be monitored using real-time monitoring sensors. A control strategy is selected based on the charging and discharging state of the energy storage battery within the current time period. The control strategy can be divided into two closed-loop control strategies: a voltage-current dual-loop control strategy and a single current-loop control strategy. Secondly, the photovoltaic-storage-DC-flexible power distribution system is controlled in real-time using system operating parameters for future time periods. The combination of these two approaches ensures the system operates in an optimal state.

[0065] Furthermore, to further improve the accuracy of control over the photovoltaic-storage-DC-flexible power distribution system, in one embodiment of this application, the system acquires the charging and discharging state of the energy storage battery within the current time period through real-time monitoring sensors. This includes charging and discharging power, current direction, and SOC value. Based on this data, the system dynamically selects the optimal control strategy. For example, a voltage-current dual-loop control strategy or a single current-loop control strategy can be used, combined with system operating parameters for future time periods output by a predictive model to form a composite control scheme. This method, which combines real-time state monitoring with predictive control, can improve the overall energy efficiency of the system and increase its lifespan compared to traditional single control strategies.

[0066] Specifically, based on the charging and discharging state, a target control strategy is determined, which may include: when the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is currently charging and the SOC of the energy storage battery is lower than a first threshold, the target control strategy is determined as a first control strategy; when the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is charging and the SOC of the energy storage battery is not lower than the first threshold, the target control strategy is determined as a second control strategy; when the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is discharging and the power generation of the photovoltaic-storage-DC-flexible power distribution system is not less than the power required by the load, the target control strategy is determined as the first control strategy; when the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is discharging and the power generation of the photovoltaic-storage-DC-flexible power distribution system is less than the power required by the load, the target control strategy is determined as the second control strategy.

[0067] The first control strategy is to control the charging and discharging of the energy storage battery through a single-loop control method; the second control strategy is to control the charging and discharging of the energy storage battery through a dual-loop control method.

[0068] In this embodiment, a state of charge (SOC) below 20% is considered a deep discharge risk zone. At this point, the internal active material structure of the battery may be damaged, leading to irreversible capacity decay, and even causing problems such as lithium plating and a surge in internal resistance. Therefore, the first threshold can be set to 20%.

[0069] Specifically, when the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is currently charging and its State of Charge (SOC) is below a first threshold, the target control strategy is determined as the first control strategy. The first control strategy employs single-current-loop control. In this case, the energy storage battery adopts a constant-current charging strategy to quickly increase its capacity and avoid deep discharge that could damage battery life. Single-loop control relies solely on the current-loop PI regulator, resulting in faster response and reduced control complexity.

[0070] When the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is charging and its State of Charge (SOC) is not lower than the first threshold, the target control strategy is determined to be the second control strategy. In this case, a constant voltage and constant current charging mode will be adopted, and the second control strategy employs a dual-loop control strategy for the energy storage battery's voltage and current. For example, the outer loop generates a current reference value based on the DC bus voltage deviation, while the inner loop uses a proportional resonant controller to achieve zero steady-state error. When the SOC is within the safe range, both voltage stability and economy are considered.

[0071] When the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is in a discharged state, and the power generation of the system is not less than the power required by the load, the target control strategy is determined as the first control strategy. Specifically, when the system has sufficient energy, the energy storage only needs to supplement the instantaneous power gap. Single-loop control directly tracks the power reference value and reduces the computational load.

[0072] When the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is in a discharging state, and the power generation of the system is less than the power required by the load, the target control strategy is determined to be the second control strategy. Specifically, when power generation is insufficient, the energy storage needs to support the bus voltage. The outer loop voltage stabilization maintains the DC voltage, while the inner loop current loop responds to load sudden changes. When the state of charge (SOC) is below 10%, derating protection is added to prevent over-discharge.

[0073] As can be seen from the above, in one embodiment of this application, the system dynamically switches control strategies based on the real-time charge / discharge status and SOC value of the energy storage battery: when the battery is in a charging state and SOC < 20%, a single current loop control strategy is adopted to quickly increase the charge through constant current charging mode, avoiding damage from deep discharge; when SOC ≥ 20%, it switches to a voltage and current dual-loop control strategy, using a hybrid charging mode of outer loop voltage regulation and inner loop constant current to maximize charging efficiency. In the discharge scenario, if the photovoltaic power generation is ≥ the load demand, single-loop control is activated to directly track the power command; when the power generation is insufficient, it automatically switches to dual-loop control, maintaining bus stability through outer loop voltage control and achieving precise power allocation through inner loop current control. This hierarchical control mechanism improves the battery's operating efficiency within the SOC safe range, and reduces the battery degradation rate under extreme conditions through derating protection.

[0074] In one embodiment of this application, feature extraction is performed on the target parameters to obtain a feature vector. Specifically, this may include: aligning various parameters, light intensity, and grid load in the system operating parameters according to time series; cleaning outliers from the target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period; normalizing the outlier-cleaned target parameters to obtain normalized target parameters; for each type of parameter in the system operating parameters, calculating the statistical characteristics and rate of change characteristics corresponding to that type of parameter based on window sliding, wherein the statistical characteristics of a type of parameter include at least one of parameter variance, parameter mean, parameter maximum, or parameter minimum, and the rate of change characteristic of a type of parameter characterizes the rate of change of that type of parameter; and performing normalization processing. Based on the normalized system operating parameters and the normalized light intensity, a first correlation feature is calculated; based on the normalized system operating parameters and the normalized grid load, a second correlation feature is calculated; and based on the normalized system operating parameters, a third correlation feature is calculated. The first correlation feature is the correlation between light intensity and system operating parameters; the second correlation feature is the correlation between grid load and system operating parameters; and the third correlation feature is the correlation between various types of system operating parameters. The normalized target parameter, the statistical features and rate of change features corresponding to various parameters in the system operating parameters, the first correlation feature, the second correlation feature, and the third correlation feature are concatenated to obtain a feature vector.

[0075] In this embodiment, the target parameters are aligned using time series to ensure that all data have a corresponding relationship at the same point in time. This step is to enable accurate comparison and calculation of the correlation between different parameters in subsequent analysis. After alignment, the target parameters within the current time period need to undergo outlier cleaning to remove potential noise or erroneous data, ensuring data quality and reliability. Outlier cleaning can employ statistical methods or rule-based methods to identify and process data points that deviate from the normal range. After outlier cleaning, the target parameters need to be normalized. The purpose of normalization is to transform parameters of different dimensions or magnitudes to the same scale range, preventing certain parameters from dominating the analysis results due to excessively large or small values. Normalized data is more conducive to subsequent feature extraction and correlation analysis.

[0076] Next, for each type of parameter in the system operating parameters, statistical characteristics and rate of change characteristics are calculated using a sliding window method. Statistical characteristics may include at least one of variance, mean, maximum, or minimum values, used to describe the distribution and fluctuation of the parameter within the window period. Rate of change characteristics reflect the rate of change of the parameter over time, capturing the dynamic trend of parameter change. These characteristics help to more comprehensively understand the behavior patterns of the parameters. Then, three types of correlation characteristics are further calculated. The first type of correlation characteristic is based on normalized system operating parameters and light intensity. The second type of correlation characteristic is based on normalized system operating parameters and grid load. The third type of correlation characteristic analyzes the interrelationships between various types of system operating parameters. In this embodiment, through the above steps, the system can conduct in-depth analysis of the operating status of the photovoltaic-storage-DC-flexible power distribution system from multiple angles and levels, providing data support for subsequent optimization and control. The entire process emphasizes the integrity and accuracy of the data, ensuring the reliability and practicality of the analysis results.

[0077] Furthermore, after obtaining the first, second, and third correlation features, the statistical features and rate of change features corresponding to various parameters in the normalized target parameters and system operating parameters, as well as the first, second, and third correlation features, are concatenated to obtain a feature vector. This ensures that the obtained feature vector includes not only the vector features corresponding to each parameter, but also the correlation features between light intensity and system operating parameters, the correlation features between grid load and system operating parameters, and the correlation features between various system operating parameters. This improves the accuracy of predicting system operating parameters in the future time period using these feature vectors.

[0078] Furthermore, the method may also include: extracting periodic information from the timestamp information of the collected target parameters; the periodic information includes: hour, day, week and month; performing one-hot encoding on the periodic information to obtain time feature encoding.

[0079] Specifically, the statistical characteristics, rate of change characteristics, first correlation characteristics, second correlation characteristics, and third correlation characteristics of the normalized target parameters, system operating parameters, and various other parameters are concatenated to obtain a feature vector, including:

[0080] The normalized target parameters, statistical features of various parameters in the system operation parameters, as well as the rate of change features, first correlation features, second correlation features, third correlation features, and time feature vectors are concatenated to obtain the feature vector.

[0081] In this embodiment, periodic information, including hours, days, weeks, and months, can be parsed from the system parameter collection timestamps.

[0082] Specifically, the hourly data reflects daily electricity load fluctuations, such as midday solar PV peaks and evening load peaks; the daily data is correlated with monthly electricity plans; the weekly data distinguishes between weekday and weekend load patterns; and the monthly data corresponds to seasonal energy supply and demand changes. One-hot encoding can be used to convert discrete periodic categorical variables into binary vectors, eliminating the misleading effect of numerical values. Finally, feature concatenation processing can be used to fuse information such as real-time status, short-term fluctuations and changes, external correlations, internal couplings, and time-series patterns to obtain feature vectors.

[0083] Specifically, for the hourly cycle encoding, a day is evenly divided into 24 independent time units, with each hour corresponding to a specific position in a vector. This encoding method clearly reflects the diurnal variation characteristics of photovoltaic power generation. For example, the peak power generation at noon is encoded as a unique binary pattern, contrasting sharply with the zero power generation at night. Simultaneously, this encoding naturally incorporates the temporal information of morning and evening load peaks, helping the model identify typical fluctuation patterns in the electricity consumption curve. In the encoding implementation, each hour is mapped to a basis vector in a 24-dimensional space, preserving the absolute positional information of the time point.

[0084] The weekday date encoding uses a 7-dimensional binary vector to represent the day of the week. This encoding method is particularly helpful in distinguishing the differences in load characteristics between weekdays and weekends, such as the regular operation of manufacturing during weekdays and the concentrated electricity consumption patterns of commercial establishments on weekends. Through one-hot encoding, the model can clearly identify the different electricity consumption characteristics of Friday evening and Sunday evening, without misinterpreting consecutive date values ​​as a linear relationship. In practical applications, this encoding can also reflect the impact of special dates, such as public holidays, which can be supplemented with additional flag bits.

[0085] The monthly cycle coding process divides the year into 12 independent time periods, with each month corresponding to an activation bit in the vector. This coding method effectively captures the impact of seasonal variations on system operation, such as the significant difference between summer cooling load and winter heating demand. Through unique thermal coding, the model can clearly identify the changes in sunshine duration in different seasons and the resulting fluctuations in photovoltaic power generation. In the actual coding process, the similarity between adjacent months needs to be automatically learned by the model from the data, rather than being preset based on the similarity of coded values.

[0086] As described above, in one embodiment of this application, the system extracts multi-level periodic features through timestamp parsing, then uses one-hot encoding technology to convert discrete periodic variables into binary feature vectors, and finally concatenates the processed time feature codes with multiple types of features. This provides a complete input for the subsequent prediction model, including time series patterns, operating condition characteristics, and external correlations. Furthermore, when collecting various types of data, the system also collects timestamp information of target parameters to extract periodic information, obtains time feature codes based on the periodic information, and concatenates them with the feature vector obtained above to obtain a new feature vector. This new feature vector is then used to predict system operating parameters, allowing the system to focus on parameter changes at different times or the relationships between parameters at different times. This further improves the accuracy of system parameter prediction in future time periods and enables effective control of the operation of the photovoltaic-storage-DC-flexible power distribution system.

[0087] In one embodiment of this application, the feature vector is encoded based on temporal features to obtain encoded data. Specifically, this may include: dividing the feature vector into continuous time windows along the time dimension, with each window containing N time steps; adding absolute or relative position encoding to each time step to obtain position feature encoding; connecting each time window after position encoding through temporal residuals to obtain temporal residual encoding; applying a multi-head self-attention mechanism to the temporal residual encoding to calculate the correlation between time steps; based on the correlation between time steps, aggregating the features of each time step to obtain aggregated features; gradually abstracting temporal features through a multi-layer network based on the aggregated features; and performing pooling processing on the temporal features to obtain encoded data.

[0088] Specifically, in this embodiment, firstly, absolute positional encoding or relative positional encoding is used to assign explicit positional information to each time step. Absolute positional encoding marks the specific position of each time step within the window using a fixed pattern, enabling the model to identify feature differences across different time periods. Relative positional encoding, on the other hand, captures short-term temporal dependencies by dynamically calculating the distance between time steps. In this embodiment, a combination of these two encoding methods can also be used to maintain the sequential nature of the time series while enhancing the model's sensitivity to local temporal patterns.

[0089] Secondly, the introduction of temporal residual encoding further enhances the model's temporal modeling capabilities. Through the residual connection structure, the model extracts high-order temporal features while retaining important information from the original input, avoiding the gradient vanishing problem in deep network training. This stage of processing enables the effective extraction of local temporal patterns within the window while maintaining feature integrity.

[0090] Subsequently, the application of multi-head self-attention mechanism elevates time series analysis to the global level. By computing multiple attention heads in parallel, the model can analyze the complex dependencies between time steps from different perspectives and identify key time points and potential long-range correlations that affect system operation.

[0091] Then, in the feature aggregation stage, relevance weights calculated based on a self-attention mechanism are used to dynamically weight and fuse features at each time step within the window. This process is essentially a feature selection mechanism, which strengthens the contribution of important time steps and weakens the influence of noisy or irrelevant time steps, making the aggregated features more representative and robust. The subsequent multi-layer network processing realizes hierarchical abstraction of temporal features. Lower-layer networks capture fine-grained short-term fluctuations, while higher-layer networks integrate macroscopic long-term trends, forming a multi-scale understanding of the system's operating rules.

[0092] The final pooling process, a crucial step in feature compression, transforms the variable-length time window features into fixed-dimensional encoded data through statistical summarization. This process not only reduces computational complexity but also enhances the model's robustness to temporal variations, enabling the encoded data to more effectively represent the core features of the entire time window. The entire encoding process is interconnected, progressing from local to global, from micro to macro, gradually building a complete understanding of the system's operational temporal characteristics, providing high-quality feature representations for subsequent prediction tasks. This refined temporal encoding method is particularly suitable for handling operating parameters with complex spatiotemporal correlations in photovoltaic-storage-DC-flexible power distribution systems, effectively improving the accuracy and generalization ability of prediction models.

[0093] As can be seen from the above, the architecture proposed in this application accurately models cross-time-period dependencies through composite positional encoding. Compared with traditional models, it can reduce the mean square error when predicting sudden changes in photovoltaic output and reduce the problem of network gradient vanishing.

[0094] In one embodiment of this application, feature interaction processing is performed on the encoded data to obtain interaction information. Specifically, this may include: operating on the encoded data based on a multi-head attention mechanism to obtain attention matrices representing different capabilities; and performing cross-multiplication of the encoded data and the attention matrices representing different capabilities to obtain interaction information.

[0095] In this embodiment, the encoded data is manipulated using a multi-head attention mechanism to obtain attention matrices representing different capabilities. The encoded data and these attention matrices are then cross-multiplied to obtain interaction information. In this embodiment, the encoded data undergoes multiple independent linear transformations to generate multiple sets of Query(Q), Key(K), and Value(V) matrices, each called a "head". Each "head" learns a different feature representation space (e.g., one head focuses on temporal dependencies, while another focuses on power correlations between devices). The attention heads compute in parallel to capture different types of dependencies in the data.

[0096] Furthermore, the encoded data is recombined according to relevance using the attention matrix of each head to generate new features carrying contextual information.

[0097] In this embodiment, in order to extract richer feature information from the data, a multi-head self-attention mechanism can be used to improve information richness. In the multi-head self-attention process, by performing the same and independent operations, multiple attention matrices with different representation capabilities can be obtained. Using multiple matrices and processing them with different parameters, attention matrices with different capabilities can be obtained.

[0098] In this embodiment, the encoded data can be cross-multiplied element-wise with each attention matrix, and then concatenated to obtain the interaction information.

[0099] As can be seen from the above, this embodiment significantly improves the expressive ability of the time-series characteristics of the photovoltaic-storage system through a multi-head interactive attention mechanism, providing a more robust input for subsequent control decisions. Furthermore, after obtaining the interactive information, the system operating parameters for the future time period are predicted by the model based on the interactive information, thereby controlling the operation of the photovoltaic-storage DC-flexible power distribution system.

[0100] Corresponding to the photovoltaic-storage-DC-flexible power distribution method in the above embodiments, Figure 3 This is a structural block diagram of a photovoltaic-storage-DC-flexible power distribution system according to an embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 3 The photovoltaic-storage-DC-flexible power distribution system 20 includes: a parameter acquisition module 21, a parameter prediction module 22, and a system operation control module 23.

[0101] The parameter acquisition module 21 is used to acquire the target parameters of the photovoltaic-storage-DC-flexible power distribution system in the current time period. The target parameters include: system operating parameters, solar intensity and grid load. The system operating parameters include the power generation, charge state, charging and discharging power, voltage and current and grid connection parameters of the photovoltaic-storage-DC-flexible power distribution system.

[0102] The parameter prediction module 22 is used to input the target parameters of the photovoltaic-storage-DC-flexible power distribution system in the current time period into the prediction model to predict the parameters and obtain the system operation parameters in the future time period.

[0103] System operation control module 23 is used to control the operation of the photovoltaic-storage-DC-flexible power distribution system based on system operation parameters within a future time period;

[0104] Specifically, the parameter prediction module 22 is used for parameter prediction in the following ways:

[0105] The target parameters are subjected to feature extraction to obtain a feature vector;

[0106] The feature vectors are encoded based on temporal features to obtain encoded data;

[0107] Feature interaction processing is performed on the encoded data to obtain interactive information;

[0108] The system's operating parameters for future time periods are obtained based on the interactive information.

[0109] In one embodiment of this application, the system operation control module 23 is further configured to:

[0110] When the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is currently charging and the SOC of the energy storage battery is lower than the first threshold, the target control strategy is determined to be the first control strategy.

[0111] When the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is in a charging state and the SOC of the energy storage battery is not lower than the first threshold, the target control strategy is determined to be the second control strategy.

[0112] When the energy storage battery in the photovoltaic-storage-direct-flexible power distribution system is in a discharging state, and the power generation of the photovoltaic-storage-direct-flexible power distribution system is not less than the power required by the load, the target control strategy is determined as the first control strategy.

[0113] When the energy storage battery in the photovoltaic-storage-direct-flexible power distribution system is in a discharging state, and the power generation of the photovoltaic-storage-direct-flexible power distribution system is less than the power required by the load, the target control strategy is determined to be the second control strategy.

[0114] The first control strategy is to control the charging and discharging of the energy storage battery through a single-loop control method; the second control strategy is to control the charging and discharging of the energy storage battery through a dual-loop control method.

[0115] In one embodiment of this application, the system parameter prediction module 22 is further configured to:

[0116] The various parameters in the system operation parameters, light intensity, and power grid load are aligned in time series.

[0117] The target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period are cleaned of outliers; the cleaned target parameters are then normalized to obtain the normalized target parameters.

[0118] For each type of parameter in the system operating parameters, based on window sliding, the statistical characteristics and rate of change characteristics of that type of parameter are calculated. The statistical characteristics of a type of parameter include at least one of the variance mean, maximum or minimum value. The rate of change characteristics of a type of parameter represent the rate of change of that type of parameter.

[0119] Based on the normalized system operating parameters and the normalized light intensity, a first correlation feature is calculated; based on the normalized system operating parameters and the normalized grid load, a second correlation feature is calculated; and based on the normalized system operating parameters, a third correlation feature is calculated. The first correlation feature is the correlation between light intensity and system operating parameters; the second correlation feature is the correlation between grid load and system operating parameters; and the third correlation feature is the correlation between various types of system operating parameters.

[0120] The statistical features, rate of change features, first correlation features, second correlation features, and third correlation features corresponding to the normalized target parameters, system operating parameters, and various other parameters are concatenated to obtain the feature vector.

[0121] In one embodiment of this application, the parameter prediction module 22 is further configured to:

[0122] Periodic information includes: periodic information extracted from the timestamp information of the collected target parameters; hour, day, week, and month;

[0123] One-hot encoding is performed on periodic information to obtain time feature encoding;

[0124] Specifically, the statistical characteristics, rate of change characteristics, first correlation characteristics, second correlation characteristics, and third correlation characteristics of the normalized target parameters, system operating parameters, and various other parameters are concatenated to obtain a feature vector, including:

[0125] The normalized target parameters, statistical features of various parameters in the system operation parameters, as well as the rate of change features, first correlation features, second correlation features, third correlation features, and time feature vectors are concatenated to obtain the feature vector.

[0126] In one embodiment of this application, the parameter prediction module 22 is further configured to:

[0127] The feature vector is divided into continuous time windows along the time dimension, with each window containing N time steps;

[0128] Add absolute position encoding and relative position encoding to each time step to obtain position feature encoding;

[0129] Each time window after location encoding is concatenated using temporal residuals to obtain temporal residual encoding;

[0130] A multi-head self-attention mechanism is applied to the temporal residual coding to calculate the correlation between time steps;

[0131] Based on the correlation between time steps, feature aggregation is performed on the features of each time step to obtain aggregated features;

[0132] Based on aggregated features, temporal features are gradually abstracted through a multi-layer network;

[0133] The temporal features are pooled to obtain encoded data.

[0134] In one embodiment of this application, the parameter prediction module 22 is further configured to:

[0135] The encoded data is manipulated based on a multi-head attention mechanism to obtain attention matrices representing different capabilities;

[0136] Interaction information is obtained by cross-multiplying the encoded data with attention matrices representing different capabilities.

[0137] See Figure 4 , Figure 4 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 4The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules in the aforementioned system embodiments, for example... Figure 3 The functions of the parameter acquisition module 21, parameter prediction module 22, and system operation control module 23 are shown.

[0138] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), but it may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0139] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0140] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example,

[0141] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation methods described in the photovoltaic-storage-direct-flexible power distribution method provided in the embodiments of this application, or they can execute the implementation methods of the electronic devices described in the embodiments of this application, which will not be repeated here.

[0142] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to implement these processes. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or system capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0143] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0144] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0145] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0146] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0147] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0148] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0149] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for photovoltaic-storage-DC-flexible power distribution, characterized in that, include: Obtain the target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period. The target parameters include: system operating parameters, solar irradiance, and grid load. The system operating parameters include the power generation, state of charge, charging and discharging power, voltage, current, and grid connection parameters of the photovoltaic-storage-DC-flexible power distribution system. Input the target parameters of the photovoltaic-storage-DC-flexible power distribution system in the current time period into the prediction model to predict the parameters and obtain the system operation parameters in the future time period. Based on the system operating parameters within the aforementioned future time period, control the operation of the photovoltaic-storage-DC-flexible power distribution system. The parameter prediction methods include: The target parameters are subjected to feature extraction to obtain a feature vector; The feature vector is encoded based on temporal features to obtain encoded data; The encoded data is subjected to feature interaction processing to obtain interaction information; The system operating parameters for the future time period are obtained based on the interactive information.

2. The photovoltaic-storage-DC-flexible power distribution method according to claim 1, characterized in that, The method further includes: Obtain the charging and discharging status of the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system within the current time period; Based on the charging and discharging state, a target control strategy is determined; the target control strategy is used to control the energy storage battery of the photovoltaic-storage-DC-flexible power distribution system. The control of the photovoltaic-storage-DC-flexible power distribution system based on system operating parameters within the future time period includes: The operation of the photovoltaic-storage-DC-flexible power distribution system is controlled based on the system operating parameters and the target control strategy.

3. The photovoltaic-storage-DC-flexible power distribution method as described in claim 2, characterized in that, The step of determining the target control strategy based on the charging and discharging state includes: When the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is currently charging and the SOC of the energy storage battery is lower than the first threshold, the target control strategy is determined to be the first control strategy. When the energy storage battery in the photovoltaic-storage-DC-flexible power distribution system is in a charging state and the SOC of the energy storage battery is not lower than the first threshold, the target control strategy is determined to be the second control strategy. When the energy storage battery in the photovoltaic-storage-direct-flexible power distribution system is in a discharging state, and the power generation of the photovoltaic-storage-direct-flexible power distribution system is not less than the power required by the load, the target control strategy is determined to be the first control strategy. When the energy storage battery in the photovoltaic-storage-direct-flexible power distribution system is in a discharging state, and the power generation of the photovoltaic-storage-direct-flexible power distribution system is less than the power required by the load, the target control strategy is determined to be the second control strategy. The first control strategy is to control the charging and discharging of the energy storage battery through a single-loop control method; the second control strategy is to control the charging and discharging of the energy storage battery through a dual-loop control method.

4. The photovoltaic-storage-DC-flexible power distribution method as described in claim 1, characterized in that, The step of extracting features from the target parameters to obtain a feature vector includes: The various parameters in the system operating parameters, the light intensity, and the power grid load are aligned in time series. The target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period are cleaned of outliers; the target parameters after outlier cleansing are normalized to obtain the normalized target parameters. For each type of parameter in the system operating parameters, based on window sliding, the statistical characteristics and rate of change characteristics of that type of parameter are calculated. The statistical characteristics of a type of parameter include at least one of the following: parameter variance, parameter mean, parameter maximum or parameter minimum. The rate of change characteristics of a type of parameter represent the rate of change of that type of parameter. Based on the normalized system operating parameters and the normalized light intensity, a first correlation feature is calculated; based on the normalized system operating parameters and the normalized grid load, a second correlation feature is calculated; and based on the normalized system operating parameters, a third correlation feature is calculated. The first correlation feature is the correlation between light intensity and system operating parameters; the second correlation feature is the correlation between grid load and system operating parameters; and the third correlation feature is the correlation between various types of system operating parameters. The normalized target parameter, the statistical features and rate of change features of various parameters in the system operation parameters, the first correlation feature, the second correlation feature and the third correlation feature are concatenated to obtain the feature vector.

5. The photovoltaic-storage-DC-flexible power distribution method according to claim 4, characterized in that, The method further comprises: Extract periodic information from the timestamp information of the collected target parameters; the periodic information includes: hour, day, week, and month; One-hot encoding is performed on the periodic information to obtain time feature encoding; The step of concatenating the normalized target parameters, statistical features and rate of change features corresponding to various parameters in the system operating parameters, the first correlation feature, the second correlation feature, and the third correlation feature to obtain the feature vector includes: The normalized target parameter, the statistical features and rate of change features of various parameters in the system operation parameters, the first correlation feature, the second correlation feature, the third correlation feature and the time feature vector are concatenated to obtain the feature vector.

6. The photovoltaic-storage-DC-flexible power distribution method as described in claim 1, characterized in that, The process of encoding the feature vector based on temporal features to obtain encoded data includes: The feature vector is divided into continuous time windows along the time dimension, and each window contains N time steps; Add absolute or relative position codes to each time step to obtain position feature codes; Each time window after location encoding is concatenated using temporal residuals to obtain temporal residual encoding; A multi-head self-attention mechanism is applied to the temporal residual coding to calculate the correlation between time steps; Based on the correlation between the time steps, the features of each time step are aggregated to obtain aggregated features; Based on aggregated features, temporal features are gradually abstracted through a multi-layer network; The temporal features are pooled to obtain the encoded data.

7. The photovoltaic-storage-DC-flexible power distribution method as described in claim 1, characterized in that, The encoded data is subjected to feature interaction processing to obtain interaction information, including: The encoded data is manipulated based on a multi-head attention mechanism to obtain attention matrices representing different capabilities; The interaction information is obtained by cross-multiplying the encoded data with the attention matrices representing different capabilities.

8. A photovoltaic-storage-DC-flexible power distribution system, characterized in that, include: The parameter acquisition module is used to acquire the target parameters of the photovoltaic-storage-DC-flexible power distribution system within the current time period. The target parameters include: system operating parameters, solar intensity and grid load. The system operating parameters include the power generation, charge state, charging and discharging power, voltage and current and grid connection parameters of the photovoltaic-storage-DC-flexible power distribution system. The parameter prediction module is used to input the target parameters of the photovoltaic-storage-DC-flexible power distribution system in the current time period into the prediction model to predict the parameters and obtain the system operating parameters in the future time period. The system operation control module is used to control the operation of the photovoltaic-storage-DC-flexible power distribution system based on the system operation parameters within the future time period. Specifically, the parameter prediction module is used for parameter prediction as follows: The target parameters are subjected to feature extraction to obtain a feature vector; The feature vector is encoded based on temporal features to obtain encoded data; The encoded data is subjected to feature interaction processing to obtain interaction information; The system operating parameters for the future time period are obtained based on the interactive information.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images