Control method, program product, medium and equipment for light storage and charging energy unit

Abstract

The invention provides a control method of an optical storage and charging energy unit, a program product, a medium and equipment, and relates to the field of power supply and consumption. The control method comprises the following steps: collecting electric energy historical data of a light storage and charging energy unit, and obtaining electric energy influence factor information corresponding to the electric energy historical data; processing the electric energy historical data and the influence factor information into a multi-dimensional data set with a uniform data format; extracting a core feature set from the multi-dimensional data set; clustering the light storage and charging energy units by using the core feature set to obtain unit groups with similar electric energy application behaviors; and using the AI control model corresponding to the unit group to control the optical storage and charging energy unit. According to the scheme, various factors can be comprehensively considered to control the electric energy use condition of the light storage and charging energy unit, so that the adaptability and effectiveness of overall control are improved.

Description

Technical Field

[0001] This invention relates to the field of power supply and utilization, and in particular to a control method, program product, medium and equipment for a photovoltaic energy storage and charging unit. Background Technology

[0002] A photovoltaic-storage-charging (PV-SCD) energy unit is a comprehensive energy solution integrating photovoltaic power generation, energy storage systems, and charging facilities. For example, residential and small-scale commercial electricity users can operate as independent PV-SCD-charging units after installing photovoltaic and energy storage equipment. PV-SCD-charging units are characterized by their large number, large scale, and significant differences in features. With the widespread use of distributed renewable energy and electric vehicles, various uncertainties are coupled together, significantly altering the energy usage characteristics of PV-SCD-charging units. This poses a significant challenge to the safe and efficient operation of power systems, especially distribution systems that directly serve users.

[0003] To address these challenges, more effective control methods for photovoltaic, energy storage, and charging energy units are needed to regulate power supply and consumption in a targeted manner, meet users' energy needs, and reduce the impact on the power grid.

[0004] Compared to large-scale power grid control, the energy usage characteristics of photovoltaic (PV), energy storage, and charging (ESC) energy units are characterized by significant differences in user energy consumption behavior, numerous external interferences, complex time-series patterns, and insufficient predictive reliability. This makes existing system-level power grid operation control methods unsuitable for direct application. Consequently, existing control methods for ESC energy units cannot adapt to user characteristics when using general-purpose models for prediction and control, resulting in high control failure rates. Furthermore, the limited data volume of ESC energy units themselves leads to poor training convergence and limited model accuracy. This results in a lack of efficient and accurate methods for ESC energy units in current technology. Summary of the Invention

[0005] One object of the present invention is to provide a control method for a photovoltaic energy storage and charging unit based on an AI model that at least solves any of the above-mentioned technical problems.

[0006] A further objective of this invention is to address the problem of significant differences in users' electricity usage behavior.

[0007] A further objective of this invention is to enhance the foresight and precision of control.

[0008] Specifically, this invention provides a control method for a photovoltaic energy storage and charging unit based on an AI model. The method includes: Collect historical power data of photovoltaic energy storage and charging units, and obtain information on power influencing factors corresponding to the historical power data; Historical electricity data and influencing factor information are processed into a multi-dimensional dataset with a unified data format. Extract the core feature set from the multi-dimensional dataset; Clustering of photovoltaic, energy storage and charging units using core feature sets yields groups of units with similar electricity application behaviors; The photovoltaic energy storage and charging unit is controlled using an AI control model corresponding to the unit group.

[0009] Optionally, the steps of controlling the photovoltaic energy storage and charging unit using an AI control model corresponding to the unit group include: Obtain the pre-built temporal prediction network model and AI control model for each unit group; A time-series prediction network model is used to predict real-time data of photovoltaic energy storage and charging units to obtain power prediction data; The AI ​​control model generates control commands based on real-time data and power prediction data, and the photovoltaic energy storage and charging unit executes the control commands.

[0010] Optionally, the steps of using a time-series prediction network model to predict real-time data of photovoltaic energy storage and charging units include: Collect real-time data from photovoltaic energy storage and charging units. The real-time data includes real-time power data and information on factors affecting power. Determine the unit group to which the photovoltaic, energy storage, and charging energy unit belongs; The time-series prediction network model of the unit group to which the photovoltaic, energy storage and charging energy unit belongs is used to predict real-time power data and real-time power influencing factors to obtain power prediction data.

[0011] Optionally, the steps for collecting historical power data of the photovoltaic energy storage and charging unit include: collecting raw power data from the smart meter terminal and power supply terminal of the photovoltaic energy storage and charging unit, performing numerical processing on the raw power data according to a preset sampling period, and obtaining historical power data. The steps for obtaining information on factors affecting electricity include: acquiring meteorological data from an environmental data management platform, determining the date stamp data corresponding to the meteorological data based on a calendar database, normalizing the meteorological data and the date stamp data, and obtaining information on factors affecting electricity.

[0012] Optionally, the steps of processing historical electricity data and information on electricity influencing factors into a multi-dimensional dataset with a unified data format include: Time alignment processing is performed on historical power data and information on factors affecting power to ensure that the two correspond in the time dimension; The time-aligned historical power data and power influencing factor information are converted according to a preset data format; The historical electricity data and information on factors affecting electricity, after being converted into different data formats, are organized into a multi-dimensional dataset.

[0013] Optionally, the steps for extracting the core feature set from the multi-dimensional dataset include: Calculate the quantile autocovariance of historical electricity data in a multi-dimensional dataset to generate multiple behavioral features that reflect the fluctuation patterns of historical electricity data. The information on power influencing factors in multi-dimensional datasets is feature-encoded to generate external influence features; Combining behavioral characteristics and external influence characteristics to form multidimensional initial characteristics; The initial multidimensional features are reduced in dimensionality to obtain the core feature set.

[0014] Optionally, the steps for dimensionality reduction of the multidimensional initial features include: Calculate the covariance matrix of the multidimensional initial features; Find the eigenvalues ​​of the covariance matrix and the corresponding orthogonalized unit eigenvectors; Sort the dimensions of the initial multidimensional features according to their feature values; Extract the features corresponding to the top-ranked feature values ​​to form the initial multidimensional features after dimensionality reduction.

[0015] Optionally, the step of extracting the features corresponding to the top-ranked feature values ​​includes: Calculate the cumulative variance contribution rate sequentially according to the order of the eigenvalues; The dimension of the extracted features is determined based on the magnitude of the cumulative variance contribution rate.

[0016] Optionally, the step of clustering photovoltaic-storage-charging energy units using a core feature set includes: Calculate the similarity between the core feature sets corresponding to photovoltaic, energy storage, and charging energy units; Based on similarity, hierarchical clustering is used to merge the photovoltaic, energy storage, and charging units into multiple initial unit groups; The initial unit groups were optimized using fuzzy mean clustering optimized by a genetic algorithm to obtain unit groups with similar power application behaviors.

[0017] According to another aspect of the present invention, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the steps of the control method for the AI ​​model-based photovoltaic energy storage and charging unit described above.

[0018] According to another aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the control method for the photoelectric energy storage and charging unit based on the AI ​​model described above are implemented.

[0019] According to another aspect of the present invention, a computer device is also provided, which includes a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the control method for the photoelectric energy storage and charging unit based on the AI ​​model described above.

[0020] The control method for photovoltaic-storage-charging energy units based on AI models of this invention processes historical power data and information on power influencing factors into a multi-dimensional dataset and extracts core feature sets, then clusters them and constructs predictive models according to the categories. Units with similar power application behaviors are grouped and controlled using corresponding AI control models. The AI ​​(Artificial Intelligence) control model aggregates power data from a large number of users within the group, avoiding the problem of poor model convergence caused by insufficient power data from a single user. Furthermore, the similarity of power application behaviors within the same group avoids the problem of inconsistent control due to large differences in user power usage behaviors. This allows for comprehensive consideration of multiple factors to control the power usage of photovoltaic-storage-charging energy units, providing a scientific basis for energy-saving guidance for users and improving the accuracy and effectiveness of power management. This scheme accurately captures the differences in power application behaviors between units, making the control more tailored to the characteristics of each group and improving the overall adaptability and effectiveness of the control.

[0021] Furthermore, the present invention collects raw power data from a specific terminal and processes it according to a preset cycle, and obtains power influencing factor information from a specific platform and normalizes it, making the acquired data more standardized and usable, which is beneficial for subsequent accurate analysis and modeling and improves the quality of data processing.

[0022] Furthermore, the present invention performs a dimensionality reduction operation on the multidimensional initial features based on the covariance matrix, which reasonably reduces the feature dimensions from a mathematical perspective, thereby reducing computational complexity and improving subsequent processing efficiency while retaining the main information.

[0023] Furthermore, the present invention uses an LSTM (Long Short-Term Memory) time series model for training and combines it with a stochastic gradient descent algorithm for optimization. By leveraging the advantages of LSTM in processing time series data, the optimized training improves the performance of the prediction model, thereby increasing prediction accuracy.

[0024] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0025] The following sections will describe some specific embodiments of the invention in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings: Figure 1 This is a schematic diagram of a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the processing of a multi-dimensional dataset with a unified data format in a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. Figure 3 This is a schematic diagram illustrating the processing of a multi-dimensional dataset with a unified data format in a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. Figure 4 This is a schematic diagram illustrating the dimensionality reduction of multidimensional initial features in a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. Figure 5 This is a schematic diagram illustrating the use of a core feature set to cluster photovoltaic energy storage and charging units in a control method for photovoltaic energy storage and charging units based on an AI model according to an embodiment of the present invention. Figure 6 This is a schematic diagram illustrating the construction and application of a time-series prediction network model in a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. Figure 7 This is a schematic diagram of a computer program product according to an embodiment of the present invention; Figure 8 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present invention; Figure 9 This is a schematic block diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0026] Those skilled in the art should understand that the embodiments described below are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. These partial embodiments are intended to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention. Based on the embodiments provided by the present invention, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of the present invention.

[0027] This invention provides a control method for a photovoltaic-storage-charging energy unit based on an AI model, used to analyze the energy application status of the unit. The photovoltaic-storage-charging energy unit is an independent user with independent energy metering requirements, typically a low-voltage distribution network user. Energy data generally includes electricity consumption data (e.g., electricity consumption data, load power data), photovoltaic power generation data (e.g., power generation data, power output data), and energy storage charging and discharging data (e.g., charging and discharging power, charging and discharging status, charging and discharging time periods, etc.).

[0028] Figure 1 This is a schematic diagram of a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. The control method for the photovoltaic energy storage and charging unit based on an AI model may include: Step S101: Collect historical power data of the photovoltaic energy storage and charging unit, and obtain information on power influencing factors corresponding to the historical power data.

[0029] The steps for collecting historical power data from the photovoltaic-storage-charging energy unit include: collecting raw power data from the smart meter terminal of the photovoltaic-storage-charging energy unit and the power supply and consumption equipment terminal (such as power load equipment, photovoltaic control equipment, energy storage control equipment, etc.); processing the raw power data according to a preset sampling period to obtain historical power data. Historical power data can be obtained through communication with smart meters and power supply and consumption terminals (such as load equipment with power metering or electrical signal measurement functions, and control equipment of the photovoltaic-storage equipment). Historical power data may include: user ID, collection time, power consumption, power generation, power output, power equipment on / off status data, photovoltaic operation data, energy storage charging and discharging status data, current value, voltage value, harmonic content, etc.

[0030] Alternatively, historical power data can be obtained from the data servers of power companies, load equipment, or photovoltaic and energy storage equipment.

[0031] The steps for obtaining information on factors influencing electricity consumption may include: acquiring meteorological data from an environmental data management platform, determining the corresponding date stamp data based on a calendar database, and normalizing both the meteorological and date stamp data to obtain the electricity consumption influencing factor information. This information can be obtained through communication with the environmental data management platform (e.g., a meteorological platform) and calendar database. The electricity consumption influencing factor information may include: meteorological data (daily high, low, and average temperatures, humidity, precipitation, and weather type (sunny / cloudy / rainy / snowy, etc.)) and date data (month, day of the week, and whether it is a holiday). This information will affect the electricity consumption of the photovoltaic-storage-charging energy unit. In addition to meteorological and date stamp data, factors such as the surrounding commercial activities may also be considered.

[0032] Step S102: Process historical electricity data and information on electricity influencing factors into a multi-dimensional dataset with a unified data format.

[0033] Step S103: Extract the core feature set from the multi-dimensional dataset.

[0034] Step S104: Cluster the photovoltaic, energy storage and charging energy units using the core feature set to obtain unit groups with similar power application behaviors.

[0035] Step S105: The photovoltaic energy storage and charging unit is controlled using an AI control model corresponding to the unit group. One possible control method is as follows: A time-series prediction network model and an AI control model are pre-built for each unit group; the time-series prediction network model is used to predict real-time data of the photovoltaic energy storage and charging unit to obtain power prediction data; the AI ​​control model generates control commands based on the real-time data and power prediction data, and the photovoltaic energy storage and charging unit executes the control commands.

[0036] The control method for photovoltaic-storage-charging energy units based on AI models described in the above embodiments processes historical electricity data and information on electricity influencing factors into a multi-dimensional dataset and extracts core feature sets. Then, it clusters the data and constructs predictive models by category. Units with similar electricity application behaviors are grouped together to construct predictive models and AI control models respectively. This aggregates electricity data from a large number of users within each group, avoiding the problem of poor model convergence caused by insufficient electricity data from a single user. Furthermore, the similarity in electricity application behaviors within the same group avoids the problem of inconsistent user characteristics caused by large differences in user electricity usage behavior. Therefore, it can comprehensively consider multiple factors to classify and predict the electricity usage of photovoltaic-storage-charging energy units, providing a scientific basis for electricity usage regulation and user energy-saving guidance, and improving the accuracy and effectiveness of electricity management.

[0037] Figure 2 This is a schematic diagram illustrating the processing of a multi-dimensional dataset with a unified data format in a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. The steps described above for processing historical electricity data and information on electricity influencing factors into a multi-dimensional dataset with a unified data format may include: Step S201: Perform time alignment processing on historical power data and information on power influencing factors to ensure that the two correspond in the time dimension.

[0038] Time alignment processing of historical electricity data and information on factors influencing electricity can be performed to handle missing values ​​and outliers based on the collection period. For missing values, a smoothing method using average values ​​can be used to fill in missing data at a specific point in time. For example, the average of several previous and subsequent sampling points can be used as the imputed value.

[0039] Outliers can be identified by numerical anomalies, such as a sudden increase in load value reaching a preset multiple, or a negative humidity value or a value that significantly exceeds the limit. Outliers can be discarded and filled in using a method similar to missing values.

[0040] For continuous power values ​​and other consecutive data, min-max normalization can be used to normalize the [0,1] interval to avoid interference from magnitude differences in model training. That is, the maximum power value is treated as 1 and the minimum value as 0. Then, the data between the minimum and maximum values ​​is transformed into the [0,1] interval to eliminate the influence of the absolute magnitude of the values.

[0041] Step S202 involves converting the time-aligned historical power data and power influencing factor information into a preset data format. The data format can be defined based on empirical values, selecting a format that meets the accuracy requirements to eliminate the influence of dimensions.

[0042] Step S203: Organize the historical power data and power influencing factor information after data format conversion into a multi-dimensional dataset.

[0043] The above process makes the acquired data more standardized and usable, which is conducive to subsequent accurate analysis and modeling and improves the quality of data processing.

[0044] Figure 3 This is a schematic diagram illustrating the processing of a multi-dimensional dataset with a unified data format in a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. The steps described above for extracting the core feature set from the multi-dimensional dataset may include: Step S301: Calculate the quantile autocovariance (QC) of the historical electricity data in the multi-dimensional dataset to generate multiple behavioral features reflecting the fluctuation patterns of the historical electricity data. The calculation process may include: calculating the optimal values ​​of multiple key quantiles; calculating multiple quantiles using a bouncing loss function; and obtaining behavioral features through the quantile autocovariance. These behavioral features can reflect the fluctuation patterns of the historical electricity data. Given that the time series of the historical electricity data is {X1,……,Xn}, its QC calculation expression is as follows: In the formula For time series at quantiles τ∈(0,1) and The autocovariance is given by j, where j is the time delay.

[0045] The optimal conditional value q based on the time series under the corresponding quantile condition. τ The corresponding indicator function.

[0046] Step S302: Encode the information on power influencing factors in the multi-dimensional dataset to generate external influence features; Step S303: Merge the behavioral features and external influence features to form multidimensional initial features; the multidimensional initial features include historical power data features and external influence features.

[0047] Step S304: Dimensionality reduction is performed on the multidimensional initial features to obtain the core feature set.

[0048] The above process generates behavioral features by calculating quantile autocovariance, encodes and merges information features of power influencing factors, and reduces dimensionality to extract a more representative core feature set, reduce data redundancy, highlight key features, and improve the accuracy of clustering and prediction.

[0049] Dimensionality reduction can be achieved using the PCA (Principal Components Analysis) algorithm. Figure 4 This is a schematic diagram illustrating the dimensionality reduction of multidimensional initial features in a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. The steps for dimensionality reduction of the multidimensional initial features may include: Step S401: Calculate the covariance matrix of the multidimensional initial features.

[0050] Step S402: Solve for the eigenvalues ​​of the covariance matrix and the corresponding orthogonalized unit eigenvectors; Step S403: Sort the dimensions of the initial multidimensional features according to their feature values. The sorting method is to arrange them in descending order.

[0051] Step S406: Extract the features corresponding to the top-ranked feature values ​​to form the multidimensional initial features after dimensionality reduction.

[0052] The above steps, before extracting the features corresponding to the top-ranked feature values, may also include: Step S404: Calculate the cumulative variance contribution rate sequentially according to the order of the eigenvalues. That is, calculate the cumulative variance contribution rate one by one according to the sorting.

[0053] Step S405: Determine the dimension of the extracted features based on the magnitude of the cumulative variance contribution rate. That is, retain the dimensions whose cumulative variance contribution rate is greater than or equal to a set threshold (e.g., 85%, 90%). This ensures that personalized differences in user behavior are not lost, nor is the impact of external factors on the user load ignored. For example, if the cumulative variance contribution rate of the first r influencing factors is greater than or equal to the set threshold, then r is taken as the dimension of the extracted features, i.e., r-dimensional features are retained.

[0054] The above steps calculate the quantile autocovariance of historical electricity data to generate behavioral features, effectively capturing the fluctuation patterns of historical electricity data and providing crucial information for analyzing household electricity application behavior. External influence features are generated by feature encoding of electricity influencing factors and then merged with behavioral features to form multidimensional initial features. This comprehensively considers both internal electricity application behavior and external influencing factors, making the data features more comprehensive. Dimensionality reduction of the multidimensional initial features yields a core feature set, removing redundant information, highlighting key features, and improving the efficiency and accuracy of subsequent analysis.

[0055] By calculating the covariance matrix, solving for the eigenvalues ​​and the corresponding orthogonalized unit eigenvectors, the multidimensional initial features are sorted by eigenvalues ​​and the features corresponding to the top eigenvalues ​​are extracted for dimensionality reduction. This reasonably reduces the feature dimension, reduces computational complexity while retaining the main information, makes subsequent processing more efficient, and helps improve the model's generalization ability.

[0056] The above steps calculate the cumulative variance contribution rate in the order of eigenvalues ​​and determine the dimension of the extracted features based on its magnitude. This provides a scientific quantitative basis for dimensionality reduction operations, avoids the blindness of feature extraction, and can optimize the feature set while preserving data information to the greatest extent, thereby further improving the accuracy and effectiveness of data analysis.

[0057] Figure 5 This is a schematic diagram illustrating the use of a core feature set to cluster photovoltaic energy storage and charging units in a control method for photovoltaic energy storage and charging units based on an AI model according to an embodiment of the present invention. The step of clustering photovoltaic energy storage and charging units using the core feature set may include: Step S501: Calculate the similarity between the core feature sets corresponding to the photovoltaic, energy storage, and charging units. The similarity can be measured using Euclidean distance.

[0058] Step S502: Based on similarity, hierarchical clustering is used to merge the photovoltaic energy storage and charging units into multiple initial unit groups. The hierarchical clustering method first treats each user as a separate cluster, and then gradually merges the closest clusters. During this process, the "profile coefficient" is used to judge the clustering effect, and the cluster with the largest profile coefficient is taken as the initial unit group.

[0059] The silhouette coefficient (SC) is a method for evaluating clustering results. It reflects the clustering effect in terms of both intra-cluster clustering and inter-cluster separation. The SC calculation expression is as follows: SC(i) = [u(i) - v(i)] / max(u(i), v(i)), where SC(i) ranges from -1 to 1, and SC(i) = 1 indicates that both intra-cluster cohesion and inter-cluster separation are optimal. u(i) is the intra-cluster cohesion, representing the average distance between unit i and all units in its cluster. v(i) is the inter-cluster separation, representing the average distance between unit i and all units in its nearest neighboring cluster. The silhouette coefficients of all units are averaged to obtain the silhouette coefficients of the clustering result.

[0060] Step S503 involves optimizing multiple initial unit groups using fuzzy mean clustering optimized by a genetic algorithm to obtain unit groups with similar energy application behaviors. The genetic algorithm optimization can combine genetic algorithm (GA) and fuzzy C-means clustering (FCM). The optimization process is as follows: First, initialize the population by randomly generating a set of candidate cluster centers as the initial population; then evaluate the fitness by using the FCM algorithm to cluster each candidate cluster center and calculating the fitness (e.g., segmentation accuracy or silhouette coefficient); then perform a selection operation, choosing the optimal candidate cluster center based on the fitness, retaining high-quality solutions; next, perform crossover and mutation operations on the selected candidate cluster centers to generate new candidate solutions, thereby expanding the search range; finally, perform iterative optimization, repeating the evaluation, selection, crossover, and mutation steps until the termination condition (e.g., maximum number of iterations or fitness convergence) is met.

[0061] The above process calculates the similarity between the core feature sets of photovoltaic, energy storage, and charging energy units, and based on this, uses hierarchical clustering to obtain multiple initial unit groups, initially grouping users with similar electricity application behaviors together. Then, fuzzy mean clustering optimized by a genetic algorithm is used to further refine the initial groups, enabling more accurate mining of similarities between users and obtaining more precise unit groups with similar electricity application behaviors. This provides a foundation for targeted electricity usage prediction and personalized services.

[0062] Figure 6 This is a schematic diagram illustrating the construction and application of a time-series prediction network model in a control method for a photovoltaic energy storage and charging unit based on an AI model according to an embodiment of the present invention. The steps described above for constructing a time-series prediction network model for each unit group may include: Step S601: Train the LSTM time series model using the core feature set of each unit group; Step S602: Use the stochastic gradient descent algorithm to optimize and train the LSTM time series model to obtain the time series prediction network model.

[0063] After the step of constructing the temporal prediction network model for each unit group, the method of this embodiment further includes: Step S603: Collect real-time power data of the photovoltaic energy storage and charging unit, and obtain real-time power influencing factor information; Step S604: Determine the unit group to which the photovoltaic energy storage and charging unit belongs; Step S605: Use the time-series prediction network model of the unit group to which the photovoltaic, energy storage and charging energy unit belongs to predict the real-time power data and the real-time power influencing factor information to obtain power prediction data.

[0064] In step S606, the AI ​​control model generates control commands based on real-time data and power prediction data, and the photovoltaic energy storage and charging unit executes the control commands.

[0065] The above model construction process utilizes the core feature sets of each unit group to train LSTM time-series models separately, fully leveraging the advantages of LSTM in processing time-series data to effectively learn the time-dimensional patterns of energy usage in photovoltaic-storage-charging energy units. Optimizing the LSTM time-series models using the stochastic gradient descent algorithm accelerates model convergence, improves model performance and prediction accuracy, and thus provides more reliable energy prediction data for photovoltaic-storage-charging energy units within the unit group.

[0066] The AI ​​control model calculates a set of optimal control parameters based on real-time data and power forecast data, according to the control characteristics of corresponding unit groups (such as economic control preferences, system stability and safety preferences, grid dependence preferences, etc.), and formulates a sequence of control commands. This sequence of commands is then distributed to the local controllers of various devices (such as photovoltaic inverters, energy storage converters (PCS), battery management systems (BMS), charging pile controllers, and electrical equipment controllers) via Internet of Things (IoT) communication protocols (such as MQTT, Modbus, CAN bus, etc.). Upon receiving the commands, each local controller parses and executes the corresponding operations, such as adjusting power or switching switch states. The AI ​​control model continuously monitors the effects of command execution and uses this real-time data as the basis for subsequent predictions and control, forming a closed-loop control system. This allows the AI ​​model to continuously learn and optimize its decision-making strategies, adapting to constantly changing environments and equipment states.

[0067] The above model collects real-time power data and real-time power influencing factors information of photovoltaic, energy storage and charging energy units during the process. After determining the user's group, it uses the time series prediction network model of the corresponding group to make predictions. It can obtain the future power usage of photovoltaic, energy storage and charging energy units in a timely and accurate manner, providing real-time data support for real-time power dispatch and real-time energy saving suggestions for users, and enhancing the practicality and timeliness of power data analysis.

[0068] This embodiment also provides a computer program product 810, a computer-readable storage medium 820, and a computer device 830. Figure 7 This is a schematic diagram of a computer program product 810 according to an embodiment of the present invention. Figure 8 This is a schematic diagram of a computer-readable storage medium 820 according to an embodiment of the present invention. Figure 9 This is a schematic block diagram of a computer device 830 according to an embodiment of the present invention.

[0069] Computer program product 810 includes computer program 811, which, when executed by processor 831, implements the steps of the control method for the photoelectric energy storage and charging unit based on the AI ​​model described above. Computer-readable storage medium 820 stores the aforementioned computer program 811, which, when executed by processor 831, implements the steps of the control method for the photoelectric energy storage and charging unit based on the AI ​​model described above. Computer device 830 may include memory 832, processor 831, and computer program 811 stored in memory 832 and running on processor 831.

[0070] The computer program 811 used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, status setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages ​​and procedural programming languages.

[0071] Computer program 811 may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer via any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, to perform aspects of the invention, electronic circuits including, for example, programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs) may execute computer-readable program instructions to personalize the electronic circuits by utilizing state information of computer-readable program instructions.

[0072] For the purposes of this embodiment, computer program product 810 is a related product that includes computer program 811.

[0073] For the purposes of this embodiment, a computer-readable storage medium 820 is a tangible device capable of holding and storing a computer program 811. It can be any device capable of containing, storing, communicating, propagating, or transmitting the computer program 811 for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable storage medium 820 include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable optical disc read-only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanical encoding device, and any suitable combination thereof.

[0074] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A control method for a photovoltaic energy storage and charging unit based on an AI model, characterized in that... include: Collect historical power data of the photovoltaic energy storage and charging unit, and obtain information on power influencing factors corresponding to the historical power data; The historical power data and the influencing factor information are processed into a multi-dimensional dataset with a unified data format. Extract the core feature set from the multi-dimensional dataset; The photovoltaic energy storage and charging units are clustered using the core feature set to obtain unit groups with similar power application behaviors. The photovoltaic energy storage and charging unit is controlled using an AI control model corresponding to the unit group.

2. The control method for a photovoltaic energy storage and charging unit based on an AI model according to claim 1, wherein, The step of controlling the photovoltaic energy storage and charging unit using the AI ​​control model corresponding to the unit group includes: Obtain the pre-built temporal prediction network model for each of the said units and the AI ​​control model; The time-series prediction network model is used to predict the real-time data of the photovoltaic energy storage and charging unit to obtain power prediction data; The AI ​​control model generates control commands based on the real-time data of the photovoltaic energy storage and charging unit and the power prediction data, and the photovoltaic energy storage and charging unit executes the control commands.

3. The control method for a photovoltaic energy storage and charging unit based on an AI model according to claim 2, wherein, The steps for predicting the real-time data of the photovoltaic energy storage and charging unit using the time-series prediction network model include: Collect real-time data from the photovoltaic energy storage and charging unit, including real-time power data and information on real-time power influencing factors; Determine the group of units to which the photovoltaic energy storage and charging unit belongs; The time-series prediction network model of the unit group to which the photovoltaic energy storage and charging unit belongs is used to predict the real-time power data and the real-time power influencing factor information to obtain the power prediction data.

4. The control method for a photovoltaic energy storage and charging unit based on an AI model according to claim 1, wherein, The step of collecting historical power data of the photovoltaic energy storage and charging unit includes: collecting raw power data from the smart meter terminal and power supply terminal of the photovoltaic energy storage and charging unit, and performing numerical processing on the raw power data according to a preset sampling period to obtain the historical power data. The steps for obtaining information on factors affecting electricity include: obtaining meteorological data from an environmental data management platform, determining the date stamp data corresponding to the meteorological data based on a calendar database, and normalizing the meteorological data and the date stamp data to obtain the information on factors affecting electricity.

5. The control method for a photovoltaic energy storage and charging unit based on an AI model according to claim 1, wherein, The step of processing the historical power data and the information on power influencing factors into a multi-dimensional dataset with a unified data format includes: The historical power data and the information on power influencing factors are time-aligned to ensure that they correspond in the time dimension. The historical power data and the power influencing factor information, after the time alignment process, are converted according to a preset data format. The historical power data and the information on power influencing factors, after being converted to the aforementioned data format, are organized into the multi-dimensional dataset.

6. The control method for a photovoltaic energy storage and charging unit based on an AI model according to claim 1, wherein, The step of extracting the core feature set from the multi-dimensional dataset includes: The quantile autocovariance of the historical electricity data in the multi-dimensional dataset is calculated to generate multiple behavioral features that reflect the fluctuation patterns of the historical electricity data. The information on power influencing factors in the multi-dimensional dataset is feature-encoded to generate external influence features; The behavioral features and the external influence features are combined to form multidimensional initial features; The multidimensional initial features are reduced in dimensionality to obtain the core feature set.

7. The control method for a photovoltaic energy storage and charging unit based on an AI model according to claim 6, wherein, The step of reducing the dimensionality of the multidimensional initial features includes: Calculate the covariance matrix of the multidimensional initial features; Solve for the eigenvalues ​​of the covariance matrix and the corresponding orthogonalized unit eigenvectors; The dimensions of the multidimensional initial features are sorted according to their feature values; Extract the features corresponding to the top-ranked feature values ​​to form the multidimensional initial features after dimensionality reduction.

8. The control method for a photovoltaic energy storage and charging unit based on an AI model according to claim 7, wherein, Before the step of extracting the features corresponding to the top-ranked feature values, the following is included: Calculate the cumulative variance contribution rate sequentially according to the order of the eigenvalues; The dimension of the extracted features is determined based on the magnitude of the cumulative variance contribution rate.

9. The control method for a photovoltaic energy storage and charging unit based on an AI model according to any one of claims 1 to 8, wherein, The step of clustering the photovoltaic energy storage and charging units using the core feature set includes: Calculate the similarity between the core feature sets corresponding to the photovoltaic energy storage and charging units; Based on the similarity, hierarchical clustering is used to merge the photovoltaic energy storage and charging units into multiple initial unit groups; The initial unit groups are optimized using fuzzy mean clustering optimized by a genetic algorithm to obtain unit groups with similar electrical energy application behaviors.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the control method for the photoelectric energy storage and charging unit based on the AI ​​model as described in any one of claims 1 to 9.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the control method for the photoelectric energy storage and charging unit based on the AI ​​model as described in any one of claims 1 to 9.

12. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the control method for the photoelectric energy storage and charging unit based on the AI ​​model as described in any one of claims 1 to 9.

Images