Distributed photovoltaic distribution storage and operation method based on artificial intelligence

Abstract

The invention relates to the technical field of energy storage, and discloses a distributed photovoltaic distribution storage and operation method based on artificial intelligence, and the method comprises the steps: obtaining an initial multi-source data set of a distributed photovoltaic system at a user side; based on the initial multi-source data set, obtaining a distribution and storage necessity probability value through feature engineering and gradient boosting tree classification model processing, and judging whether a user side needs distribution and storage; when a user side needs distribution and storage, optimal energy storage configuration is carried out on the distributed photovoltaic system by using the bidirectional long-short-term memory network model and the improved genetic algorithm, and a next-day prediction data set and a target configuration parameter set are obtained; based on the target configuration parameter set and the next-day prediction data set, a target charging and discharging strategy of the distributed photovoltaic system is obtained through dynamic self-adaptive regulation and control method processing; and the target charging and discharging strategy is utilized to control the operation of the distributed photovoltaic system, a distribution and storage operation result is obtained, and the renewable energy consumption efficiency and the comprehensive income of the distributed photovoltaic distribution and storage project are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage, in particular to a distributed photovoltaic storage and operation method based on artificial intelligence. BACKGROUND

[0002] As an important form of clean energy utilization, the efficient consumption and economic operation of distributed photovoltaic rely on the scientific configuration and dispatch of energy storage systems, but the existing technology has the following core problems:

[0003] (1) The storage configuration method lacks scientific basis: the existing method is mostly based on static data (such as historical average), and cannot deeply mine the load-photovoltaic coupling rules hidden in the massive historical data, resulting in that the decision of "with or without storage" lacks data support, and the capacity configuration "big horse pulling small cart" or "insufficient capacity" occurs from time to time, and the precise quantitative determination of the necessity of storage cannot be realized; (2) Poor adaptability of photovoltaic storage system operation strategy: the existing energy storage system mostly uses a fixed threshold-based charging and discharging strategy (such as "valley charging peak discharging"), which cannot adaptively respond to complex real-time situations such as price fluctuations, weather abnormalities, and load steep increases, and it is difficult to maximize the benefits while ensuring the stability of operation, and there is no dynamic regulation mechanism considering economy and battery health; (3) Fragmentation of AI technology application: in the existing technology, artificial intelligence is mostly used only in a single link (such as load prediction), and a full-process intelligent closed loop of "data collection-storage decision-operation optimization-feedback iteration" has not been formed, and the system adaptability cannot be continuously improved through multi-model collaborative optimization; although simple mathematical programming models or classical machine learning models (such as regression, clustering) can process numerical data, they are difficult to understand and utilize market data, weather warnings, and device manuals and other unstructured text information rich in value, and the decision-making dimension is single and lacks foresight, and cannot provide comprehensive data support for energy storage configuration and operation.

[0004] As an important means to solve the intermittency and volatility of distributed photovoltaic, energy storage system can improve the flexibility and reliability of power system. However, at present, the energy storage configuration lacks scientific methods, and often has problems such as unreasonable capacity configuration and low operation efficiency, resulting in high investment cost and poor economic benefit of energy storage. SUMMARY

[0005] The present application provides a distributed photovoltaic storage and operation method based on artificial intelligence to solve the problem that the current energy storage configuration lacks scientific methods and has problems such as unreasonable capacity configuration and low operation efficiency.

[0006] In a first aspect, the present application provides a distributed photovoltaic storage and operation method based on artificial intelligence, which comprises: An initial multi-source data set of a distributed photovoltaic system on a user side is acquired; based on the initial multi-source data set, a feature engineering and a gradient boosting tree classification model processing are performed to obtain a storage necessity probability value; whether the user side needs storage is determined according to the storage necessity probability value; when the user side needs storage, an optimal energy storage configuration is performed on the distributed photovoltaic system by using a bidirectional long short-term memory network model and an improved genetic algorithm, and a next-day prediction data set and a target configuration parameter set are obtained; the bidirectional long short-term memory network model is obtained by stacking a forward long short-term memory network model and a reverse long short-term memory network model; based on the target configuration parameter set and the next-day prediction data set, a dynamic self-adaptive regulation method is performed to obtain a target charging and discharging strategy of the distributed photovoltaic system; the operation of the distributed photovoltaic system is controlled by using the target charging and discharging strategy, and a storage operation result is obtained.

[0007] The distributed photovoltaic storage and operation method based on artificial intelligence provided by the application can mine the load-photovoltaic coupling rules in the data by feature engineering and gradient boosting tree classification model processing, realize quantitative determination of storage necessity, overcome the subjectivity of traditional experience judgment, and solve the defects of lack of scientific basis in existing storage decision. Further, whether the user side needs storage is determined according to the storage necessity probability value, which avoids resource waste and income loss caused by blind storage or no storage, and solves the problems of “big horse pulling small cart” or “insufficient capacity” in capacity configuration. Further, the bidirectional long short-term memory network model and the improved genetic algorithm can accurately predict future key parameters and realize optimal configuration in the whole life cycle, taking into account economy and practicability, reducing energy storage investment cost and improving economic benefit. Further, the dynamic self-adaptive regulation method enables the target charging and discharging strategy to adapt to complex working conditions and balance energy storage economy and battery health. Further, the operation of the distributed photovoltaic system is controlled by using the target charging and discharging strategy, realizing the collaborative and efficient operation of the distributed photovoltaic system and the energy storage, improving the renewable energy consumption efficiency and comprehensive income of the distributed photovoltaic storage project. At the same time, it provides a physical basis for the construction of a regional power side virtual power plant, which helps to promote high proportion of regional new energy consumption.

[0008] In an optional implementation, based on the initial multi-source data set, a feature engineering and a gradient boosting tree classification model processing are performed to obtain a storage necessity probability value, including: Data cleaning is performed on the multi-source data set to obtain a target multi-source data set; feature extraction is performed on the target multi-source data set to obtain a multi-dimensional feature parameter set; the multi-dimensional feature parameters are input into a gradient boosting tree classification model for processing to obtain a storage necessity probability value.

[0009] The application provides a distributed photovoltaic storage and operation method based on artificial intelligence, which can eliminate abnormal values and repair missing values through data cleaning of a multi-source data set, thereby guaranteeing data quality.Further, through feature extraction, the core value of the data can be mined, and multi-dimensional features suitable for gradient boosting tree classification models are formed, which helps to improve the accuracy of the necessity determination of the storage and distribution.Further, the multi-dimensional feature parameters are input into the gradient boosting tree classification model for processing to obtain a storage and distribution necessity probability value, thereby realizing accurate quantification of the storage and distribution demand, overcoming the subjectivity of traditional experience judgment, and solving the defects of the lack of scientific basis for existing storage and distribution decision-making.

[0010] In an optional embodiment, when the user side needs storage and distribution, an optimal energy storage configuration is performed on the distributed photovoltaic system by using a bidirectional long short-term memory network model and an improved genetic algorithm, and a next-day prediction data set and a target configuration parameter set are obtained, including: When the user side needs storage and distribution, a target optimization function and a multi-dimensional data set and a next-day data set of the user side are obtained, the target optimization function is used to represent the maximum net income in the whole life cycle, the multi-dimensional data set and the next-day data set are input into a bidirectional long short-term memory network model for processing to obtain a multi-dimensional prediction data set and a next-day prediction data set, and the target optimization function is solved based on the multi-dimensional prediction data set and a preset constraint condition set by using an improved genetic algorithm to obtain a target configuration parameter set.

[0011] The distributed photovoltaic storage and operation method based on artificial intelligence provided by the application guarantees the economy of the configuration by obtaining a target optimization function and a multi-dimensional data set, and clearly defining the target orientation and data support of the energy storage configuration.Further, the bidirectional long short-term memory network model can accurately predict future key parameters, providing prospective data for optimal capacity solving, and improving the rationality of the configuration.Further, based on the prediction data set and the constraint condition, and by using the improved genetic algorithm for solving, the optimal configuration parameters can be obtained within the compliance and feasibility boundary, taking into account the economy, safety and practicability, thereby reducing the energy storage investment cost and improving the economic benefit.

[0012] In an optional embodiment, based on the target configuration parameter set and the next-day prediction data set, a target charging and discharging strategy of the distributed photovoltaic system is obtained through a dynamic self-adaptive regulation method, including: Based on the target configuration parameter set and the next-day prediction data set, a basic charging and discharging strategy of the distributed photovoltaic system is generated through a deep reinforcement learning method, the deep reinforcement learning method is used to balance the single-period net income and battery loss cost of the distributed photovoltaic system, a first real-time running data set of the distributed photovoltaic system based on the basic charging and discharging strategy is obtained, the target correction strategy is determined by using the first real-time running data set and the next-day prediction data set, and the basic charging and discharging strategy is corrected in real time by using the target correction strategy to obtain the target charging and discharging strategy.

[0013] The distributed photovoltaic storage and operation method based on artificial intelligence provided by the application generates a basic charging and discharging strategy through a deep reinforcement learning method, realizes the advance planning of the next day's charging and discharging, and can reduce battery loss while ensuring single-period income. Further, by obtaining a first real-time operation data set and combining a next-day prediction data set to determine a target correction strategy, the actual working condition dynamic changes can be captured, and the deviation between the prediction and the actual working condition can be accurately identified, and the strategy adjustment direction can be determined. Further, the target correction strategy is used to real-time correct the basic charging and discharging strategy, which improves the adaptability of the strategy to the real-time working condition and ensures the economy and stability of the operation process.

[0014] In an optional embodiment, the operation of the distributed photovoltaic system is controlled by using the target charging and discharging strategy, and a storage operation result is obtained, including: When the operation of the distributed photovoltaic system is controlled based on the target charging and discharging strategy, the battery state of the distributed photovoltaic system is obtained by using the artificial intelligence model; when the battery state is a healthy state, the operation of the distributed photovoltaic system is controlled by using the target charging and discharging strategy, and a storage operation result is obtained; when the battery state is an unhealthy state, the operation parameters of the distributed photovoltaic system are adjusted based on the battery state, and a storage operation result is obtained.

[0015] The distributed photovoltaic storage and operation method based on artificial intelligence provided by the application obtains the battery state by using the artificial intelligence model, can real-time master the battery health condition, and provides a safety basis for operation control. Further, when the battery is healthy, the operation is performed according to the target strategy, which ensures that the distributed photovoltaic system realizes optimal economic operation under the premise of safety. Further, when the battery is unhealthy, the operation parameters are adjusted, which can avoid further damage to the battery, prolong the service life of the battery, and further consider the operation safety and sustainability.

[0016] In an optional embodiment, the method further includes: based on the storage operation result, updating the gradient boosting tree classification model, the bidirectional long short-term memory network model, the improved genetic algorithm, and the dynamic self-adaptive regulation method, respectively.

[0017] The distributed photovoltaic storage and operation method based on artificial intelligence provided by the application updates each model and method based on the storage operation result, forms a whole-process closed-loop optimization, continuously improves the accuracy of storage decision, the rationality of energy storage configuration, and the adaptability of operation strategy, and enhances the comprehensive benefits and generalization ability of long-term operation of the system.

[0018] In a second aspect, the application provides a distributed photovoltaic storage and operation device based on artificial intelligence, which comprises: The acquisition module is used to acquire an initial multi-source data set and a next-day prediction data set of a distributed photovoltaic system on a user side; the first processing module is used to obtain a storage necessity probability value through feature engineering and a gradient boosting tree classification model based on the initial multi-source data set; the judgment module is used to judge whether the user side needs storage based on the storage necessity probability value; the configuration module is used to perform optimal energy storage configuration on the distributed photovoltaic system by using a bidirectional long short-term memory network model and an improved genetic algorithm when the user side needs storage, and obtain a target configuration parameter set, the bidirectional long short-term memory network model being obtained by stacking a forward long short-term memory network model and a reverse long short-term memory network model; the second processing module is used to obtain a target charging and discharging strategy of the distributed photovoltaic system through a dynamic self-adaptive regulation method based on the target configuration parameter set and the next-day prediction data set; and the control module is used to control the operation of the distributed photovoltaic system by using the target charging and discharging strategy, and obtain a storage operation result.

[0019] In a third aspect, the present application provides an electronic device, comprising a memory and a processor, the memory and the processor being communicatively connected with each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the artificial intelligence-based distributed photovoltaic storage and operation method of the first aspect or any of the corresponding embodiments thereof.

[0020] In a fourth aspect, the present application provides a computer readable storage medium, the computer readable storage medium storing computer instructions, and the computer instructions being used to make a computer execute the artificial intelligence-based distributed photovoltaic storage and operation method of the first aspect or any of the corresponding embodiments thereof.

[0021] In a fifth aspect, the present application provides a computer program product, comprising computer instructions, and the computer instructions being used to make a computer execute the artificial intelligence-based distributed photovoltaic storage and operation method of the first aspect or any of the corresponding embodiments thereof. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0023] Figure 1 is a schematic diagram of an application scenario according to an embodiment of the present application; Figure 2 is a flowchart of the artificial intelligence-based distributed photovoltaic storage and operation method according to an embodiment of the present application; Figure 3 This is a schematic diagram of intelligent planning and configuration in an AI-based distributed photovoltaic power generation and storage and operation method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of intelligent operation optimization in an AI-based distributed photovoltaic power generation and storage and operation method according to an embodiment of the present invention; Figure 5 This is a structural block diagram of a distributed photovoltaic power generation, storage and operation device based on artificial intelligence according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] As an optional application scenario of this invention, the specific application environment architecture or specific hardware architecture on which the execution of the AI-based distributed photovoltaic power generation, storage, and operation method depends is described herein. For example... Figure 1 As shown, the architecture system may include at least one terminal device and at least one server. Figure 1 The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0028] The terminal device can be specifically a smart phone, a tablet computer, a notebook computer, a palm computer, or can be a desktop computer, a game console, a smart television, a smart wearable device, a vehicle-mounted terminal, a VR (Virtual Reality) device, an AR (Augmented Reality) device, or the like. The server 103 can be a stand-alone physical server, or can be a server cluster or a distributed system, or can be a cloud server providing cloud services. The network 110 can be a wired network or a wireless network, and examples thereof include but are not limited to the Internet, an intranet, a local area network, a wide area network, a mobile communication network, and a combination thereof.

[0029] The embodiment of the present application provides a distributed photovoltaic storage and operation method based on artificial intelligence. First, the necessity of storage is quantitatively determined through feature engineering and a gradient boosting tree classification model. Then, through a bidirectional long short-term memory network model and an improved genetic algorithm, future key parameters can be accurately predicted and optimal configuration of the whole life cycle benefit is realized. Finally, through a dynamic self-adaptive regulation method, the target charging and discharging strategy can adapt to complex working conditions, balance the energy storage economy and the battery health, realize the collaborative and efficient operation of the distributed photovoltaic and energy storage, and improve the energy consumption efficiency and comprehensive benefit.

[0030] According to the embodiment of the present application, a distributed photovoltaic storage and operation method based on artificial intelligence is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from here.

[0031] In the present embodiment, a distributed photovoltaic storage and operation method based on artificial intelligence is provided, which can be used in the above mobile terminal, such as a mobile phone, a tablet computer, etc. Figure 2 The flowchart of the distributed photovoltaic storage and operation method based on artificial intelligence according to the embodiment of the present application is shown in FIG. 1, which includes the following steps: Figure 2 As shown in FIG. 1, the flowchart includes the following steps: In step S201, the initial multi-source data set of the distributed photovoltaic system on the user side is obtained.

[0032] In an optional embodiment, the distributed photovoltaic system represents a small photovoltaic power supply system deployed on the user side, composed of photovoltaic components, inverters and other devices, which can realize solar power generation and meet the user's own power demand (surplus power can be connected to the network).

[0033] In an optional embodiment, the initial multi-source data set represents the user-side full-dimensional historical and basic data collected to support the storage decision, with a time granularity of 5 minutes for industrial and commercial enterprises, which can include: 1. Load data: historical active power, peak-valley period distribution, daily / monthly load fluctuation coefficient; 2. Photovoltaic data: installed capacity, historical output curve, self-generation and self-use power, on-grid power and proportion, light abandonment record; 3. Market data: time-of-use electricity price (peak / flat / valley period and price), photovoltaic on-grid electricity price, subsidy policy, capacity price; 4. Environmental data: local light intensity, temperature, precipitation and other meteorological data (in the past 3 years), seasonal variation law.

[0034] In an optional embodiment, historical active power, peak-valley period distribution, daily / monthly load fluctuation coefficient and other load data can be obtained through user-side electricity meters and energy management systems.

[0035] Further, installed capacity, historical output curve, self-generation and self-use power, on-grid power and proportion, light abandonment record and other photovoltaic data can be extracted from photovoltaic inverters and metering equipment.

[0036] Further, time-of-use electricity price (peak / flat / valley period and corresponding price), photovoltaic on-grid electricity price, market policy, capacity price and other file data published by local power departments can be sorted out; Further, the database of the meteorological department can be connected to obtain local light intensity, temperature, precipitation and other meteorological data in the past 3 years, as well as seasonal variation law.

[0037] Step S202, based on the initial multi-source data set, after feature engineering and gradient boosting tree classification model processing, the necessary probability value of matching storage is obtained.

[0038] In an optional embodiment, feature engineering represents the preprocessing process of data cleaning and core feature extraction on the initial multi-source data set, the purpose is to improve data quality, mine key information, and form a feature parameter set suitable for the model.

[0039] In an optional embodiment, the gradient boosting tree (GBDT, Gradient Boosting Decision Tree) classification model represents a machine learning model based on the iterative integration of multiple decision trees, which optimizes the loss function through gradient descent, and is used to realize the matching storage / no matching storage binary classification decision, and output the matching storage necessity quantitative result.

[0040] In an optional embodiment, the full-dimensional data (including post-storage benefit data) of 1000+ distributed photovoltaic users is obtained, and the need for matching storage and the need for no matching storage are labeled as positive and negative samples, and a model training data set is formed.

[0041] Further, multi-dimensional feature parameters such as photovoltaic-load matching degree, peak-valley price difference coefficient, light abandonment loss rate, etc. are selected as model input variables.

[0042] Further, the basic hyperparameters such as the number of decision trees, the initial learning rate, and the depth of a single tree are set, and the first decision tree is initialized (with the mean of the sample labels as the initial prediction value). Then, in each iteration, the residual error between the current model prediction value and the true label of the sample is calculated, and a new decision tree is trained with the residual error as the target; the loss function is minimized by gradient descent method to determine the weight of the new decision tree and integrate it into the ensemble model, and the iteration is repeated until the preset number of trees is reached or the loss function converges, and the trained gradient boosting tree classification model is obtained.

[0043] Further, in the training process, the grid search algorithm can also be used to combine and optimize the hyperparameters such as the learning rate, tree depth, and number of leaf nodes, and the optimal hyperparameter combination is determined to ensure that the model classification accuracy is ≥90%.

[0044] In an optional embodiment, the storage necessity probability value represents a 0-100% quantized value output by the GBDT model, which is used to represent the possibility of the user side configuring storage to cover the cost of the benefit.

[0045] In an optional embodiment, by feature engineering to mine the core value of the initial multi-source data and using the GBDT classification model to quantitatively calculate the storage necessity, an objective and accurate storage necessity probability value can be finally obtained.

[0046] In an optional embodiment, when performing feature engineering processing on the initial multi-source data set, first, the local outlier factor (LOF) algorithm is used for data cleaning, and the principal component analysis (PCA) is combined to extract key features, and then the extracted features are input into the gradient boosting tree classification model for processing, thereby realizing the binary quantitative determination of the storage necessity.

[0047] Step S203, determining whether the user side needs storage according to the storage necessity probability value.

[0048] In an optional embodiment, based on the storage necessity probability value, a preset objective decision threshold can be used to further determine whether the user side distributed photovoltaic system needs to configure storage, which replaces the traditional experience judgment and avoids the waste of resources and loss of benefits caused by blind storage or no storage, ensuring the scientificity and economy of the storage decision.

[0049] In an optional embodiment, when the storage necessity probability value is ≥50%, it is determined that storage is needed, otherwise it is determined that storage is not needed.

[0050] Step S204, when the user side needs storage, using the bidirectional long short-term memory network model and the improved genetic algorithm to perform optimal storage configuration on the distributed photovoltaic system, and obtaining the next day prediction data set and the target configuration parameter set.

[0051] In an optional embodiment, photovoltaic output is affected by weather, season, etc., and has both daily periodic local dependence and seasonal long-term dependence. Therefore, in this embodiment, the bidirectional long short-term memory network model (Bi-LSTM model) can capture both forward and backward dependence relationships of time series. Further, photovoltaic and load prediction have uncertainties, and the bidirectional structure helps the model learn confidence information from the complete temporal context.

[0052] In an optional embodiment, the bidirectional long short-term memory network model is obtained by stacking a forward long short-term memory network model and a reverse long short-term memory network model. Further, for each time step, the forward and reverse hidden states are combined to form the final hidden state of the time step.

[0053] Further, compared with the ordinary unidirectional LSTM model, the bidirectional LSTM model can better capture bidirectional dependence information (such as night load affecting daytime energy storage decision), making the distributed photovoltaic storage more accurate and reliable.

[0054] In an optional embodiment, the input of the bidirectional long short-term memory network model can include historical photovoltaic output, light intensity, temperature, historical load demand, workday / holiday, time-of-use electricity price, real-time electricity price, etc. The output can include photovoltaic output prediction data, load demand prediction data, electricity price prediction data, as well as prediction uncertainty and extreme event probability.

[0055] In an optional embodiment, the improved genetic algorithm is an optimization algorithm based on the traditional genetic algorithm and introducing an adaptive crossover mutation operator, which is used to solve the optimal configuration of energy storage under constraints, and improves the optimization efficiency and result accuracy.

[0056] In an optional embodiment, the next-day prediction data set represents the next-day key operating parameter prediction results generated by the bidirectional long short-term memory (Bi-LSTM) model, which can include hourly load demand, photovoltaic output (prediction error ≤10%), time-of-use electricity price and fluctuation trend, and is used to develop a day-ahead energy storage operation plan.

[0057] In an optional embodiment, the target configuration parameter set represents the core configuration parameters of energy storage obtained by model operation with the goal of maximizing the net benefit in the whole life cycle, which can include the optimal energy storage capacity (Eopt) and the maximum charge and discharge power (Pmax).

[0058] In an optional embodiment, for users determined to need storage, long-term key parameters can be accurately predicted by a bidirectional long short-term memory network model, and then an improved genetic algorithm is used to solve the target configuration parameter set that maximizes the net income in the whole life cycle under the constraints of charging and discharging power, SOC, investment payback period, etc., while generating the next day prediction data set to support the next day operation strategy formulation, realizing the unity of economy and practicality of energy storage configuration.

[0059] In step S205, the target charging and discharging strategy of the distributed photovoltaic system is obtained based on the target configuration parameter set and the next day prediction data set through a dynamic self-adaptive regulation method.

[0060] In an optional embodiment, the dynamic self-adaptive regulation method represents a two-stage regulation method for realizing day-ahead planning + real-time correction, which can formulate strategies in advance according to prediction data and adaptively respond to real-time working condition changes, balancing the economy of energy storage and the health of the battery.

[0061] In an optional embodiment, the target charging and discharging strategy represents a quantitative operation instruction generated through two-stage regulation, which can include charging and discharging power in each period, SOC control interval, etc., and can adapt to next day prediction working conditions and real-time fluctuations, ensuring the economic and safe operation of the energy storage system.

[0062] In an optional embodiment, based on the optimal configuration parameters of energy storage and the prediction data of next day key working conditions, the two-stage regulation logic of day-ahead planning and real-time dynamic correction can dynamically generate a target charging and discharging strategy that adapts to complex working conditions, solving the problem of poor adaptability of traditional fixed threshold strategies and realizing the dual goals of maximizing energy storage benefits and prolonging battery life.

[0063] In step S206, the target charging and discharging strategy is used to control the operation of the distributed photovoltaic system, and the storage operation result is obtained.

[0064] In an optional embodiment, the storage operation result represents the full-dimensional quantitative result generated by the distributed photovoltaic system during operation under the cooperative control of the target charging and discharging strategy and the battery health management strategy, which can include energy consumption related data, economic benefit data, equipment operation state data, etc.

[0065] In an optional embodiment, the target charging and discharging strategy is converted into specific operation control instructions of the distributed photovoltaic system, and the operation state is dynamically adjusted in combination with the battery health management strategy, thereby realizing the cooperative and efficient operation of photovoltaic and energy storage, and outputting full-dimensional operation results covering energy consumption, economic benefits, and equipment state.

[0066] The distributed photovoltaic storage and operation method based on artificial intelligence provided by the embodiment can mine the load-photovoltaic coupling rules in the data, realize quantitative determination of the necessity of storage, overcome the subjectivity of traditional experience determination, and solve the defects of lack of scientific basis in existing storage decision. Further, according to the storage necessity probability value, it is judged whether the user side needs storage, which avoids the waste of resources and loss of income caused by blind storage or no storage, and solves the problem of "big horse pulling small cart" or "insufficient capacity" in capacity configuration. Further, by using the bidirectional long short-term memory network model and the improved genetic algorithm, the future key parameters can be accurately predicted and the optimal configuration of the whole life cycle income is realized, which takes into account the economy and practicability, reduces the investment cost of energy storage, and improves the economic benefit. Further, through the dynamic self-adaptive regulation method, the target charging and discharging strategy can adapt to complex working conditions and balance the economy and battery health of energy storage. Further, the operation of the distributed photovoltaic system is controlled by the target charging and discharging strategy, realizing the collaborative and efficient operation of the distributed photovoltaic and energy storage, improving the renewable energy consumption efficiency and comprehensive income of the distributed photovoltaic storage project. At the same time, it provides a physical basis for the construction of a virtual power plant on the regional power supply side, which helps to promote high proportion of regional new energy consumption.

[0067] In some optional embodiments, the step S202 comprises: In step S2021, the multi-source data set is subjected to data cleaning to obtain a target multi-source data set.

[0068] In an optional embodiment, the isolated forest algorithm can be used to identify outliers in the load data, photovoltaic data, market data, and environmental data. Then, the identified outliers are labeled and removed to avoid interfering with the data rules.

[0069] Further, the missing values in the multi-source data set can be repaired by using the linear interpolation method to ensure the continuity and integrity of the data. Finally, the corresponding target multi-source data set can be obtained after the above outlier removal and missing value repair.

[0070] In an optional embodiment, the DBSCAN clustering algorithm can also be used to identify outliers; the K-nearest neighbor (KNN) interpolation method is used to repair the missing values to ensure the time sequence correlation and accuracy of the data.

[0071] In an optional embodiment, the sliding window method (window size is set to 1 hour) can be used to detect data continuity, and the abnormal data whose fluctuation exceeds 3 times the standard deviation of the same period in history is marked, and invalid data is filtered in combination with industry standard threshold; the seasonal trend decomposition (STL) interpolation method is used to repair the missing values to adapt to environmental data and load data with obvious seasonal fluctuation characteristics.

[0072] Step S2022: Extract features from the target multi-source dataset to obtain a multi-dimensional feature parameter set.

[0073] In one optional embodiment, the multidimensional feature parameter set represents a set of features that accurately reflect the photovoltaic-load characteristics, economic benefit potential, and storage demand on the user side. It may include multidimensional feature parameters such as photovoltaic-load matching degree (the proportion of overlap between photovoltaic output and load), peak-valley price difference coefficient (peak-valley electricity price difference / flat-rate electricity price), and curtailment loss rate (curtailed photovoltaic power × grid connection price / total power generation × electricity sales price).

[0074] In one optional embodiment, by mining key information from target multi-source data, core features that can characterize the necessity of storage are constructed and a corresponding multi-dimensional feature parameter set is formed.

[0075] In an optional embodiment, the constructed features can also be standardized (uniformed in scale) to ensure that the weights of each feature are balanced during model training. Furthermore, redundant features can be filtered out.

[0076] Step S2023: Input the multidimensional feature parameters into the gradient boosting tree classification model for processing to obtain the probability value of storage necessity.

[0077] In one optional embodiment, the gradient boosting tree classification model, after training, performs calculations on the input multidimensional feature parameters and outputs a quantitative probability value that objectively reflects the possibility that the user's storage benefits cover the costs, i.e., the storage necessity probability value.

[0078] In some optional implementations, step S204 above includes: Step S2041: When the user side needs to allocate storage, obtain the target optimization function and the user side's multi-dimensional dataset and the next day's dataset.

[0079] In an alternative embodiment, the objective optimization function is used to characterize maximizing the net benefit over the entire lifecycle, as shown in the following equation (1): (1) In the formula: express The added value of self-consumption during a certain period is (electricity sales price - grid connection price) × the amount of electricity stored to replace grid connection; express The arbitrage profit during a given period is calculated as follows: (discharge price × discharge volume - charging price × charging volume). This represents the initial investment in energy storage (cost per unit capacity × capacity, approximately 0.8 yuan / Wh for lithium batteries). This represents the operation and maintenance cost (initial investment × 2% / year × lifespan). Indicates the life cycle (in 20 years).

[0080] In an optional embodiment, the multi-dimensional data set is represented as user-side full-dimensional basic data collected to support energy storage optimal capacity calculation (long-term configuration decision), which can include historical photovoltaic output, illumination intensity, temperature, historical load demand, workday / holiday, time-of-use electricity price, real-time electricity price, and other parameter information.

[0081] In an optional embodiment, the next-day data set is represented as next-day basic data collected to support next-day operation strategy formulation (short-term operation decision), which can include next-day photovoltaic output, illumination intensity, temperature, load demand, electricity price, and other parameter information.

[0082] In step S2042, the multi-dimensional data set and the next-day data set are respectively input into the bidirectional long short-term memory network model for processing to obtain a multi-dimensional prediction data set and a next-day prediction data set.

[0083] In an optional embodiment, the multi-dimensional prediction data set can include hourly load (error ≤10%), photovoltaic output (error ≤10%), and electricity price policy adjustment trend (such as peak valley period change probability, user electricity price increase, and power market on-grid electricity price) in the next year.

[0084] In an optional embodiment, the next-day prediction data set can include hourly load (error ≤10%), photovoltaic output (error ≤10%), and electricity price policy adjustment trend (such as peak valley period change probability, user electricity price increase, and power market on-grid electricity price) in the next 24 hours.

[0085] In an optional embodiment, the bidirectional long short-term memory network model respectively captures long-term data rules and next-day short-term operation characteristics, and thus can respectively generate corresponding accurate prediction data, i.e., the multi-dimensional prediction data set and the next-day prediction data set.

[0086] For example, after the multi-dimensional data set and the next-day data set are input into the bidirectional long short-term memory network model, for the multi-dimensional data set, the bidirectional long short-term memory network model predicts hourly key parameters in the next year; and for the next-day prediction data set, the bidirectional long short-term memory network model predicts hourly key parameters in the next 24 hours.

[0087] In step S2043, based on the multi-dimensional prediction data set and a preset constraint condition set, an improved genetic algorithm is used to solve a target optimization function to obtain a target configuration parameter set.

[0088] In an optional embodiment, the preset constraint condition set is used to guarantee the safety, feasibility, and economy of energy storage configuration, and can include: 1. Charge and discharge power constraints, which are respectively shown in the following relationship formulas (2) and (3): (2) (3) In the formula: represents the energy storage charging power; represents the excess power of the distributed photovoltaic system, i.e., the part of the photovoltaic output exceeding the current load demand of the user; represents the energy storage discharging power; represents the load deficiency power on the user side, i.e., the part of the current load demand exceeding the real-time output of the photovoltaic.

[0089] 2. SOC constraint, as shown in the following relationship (4): (4) In the formula: represents the state of charge of the energy storage, used to reflect the proportion of the current remaining capacity of the energy storage battery to the rated capacity.

[0090] 3. Economic constraint: payback period ≤ 4 years.

[0091] In an optional embodiment, within the preset constraint boundary, the energy storage configuration parameters that maximize the net life cycle benefit are found by the improved genetic algorithm, so as to realize the unity of the economy and feasibility of the configuration.

[0092] For example, a set of initial solutions of energy storage capacity (E) and charging and discharging power (P) are first generated as the initial population of the algorithm. Then, each initial solution is substituted into the target optimization function, and the fitness, i.e., the net life cycle benefit, of each solution is calculated in combination with the preset constraint condition.

[0093] Further, an adaptive crossover and mutation operator is introduced, and a new generation population is generated through selection, crossover and mutation operations, and the iteration is repeated until the fitness converges or the preset iteration number is reached, and then the solution with the highest fitness, i.e., the target configuration parameter set, is selected and output.

[0094] In some optional embodiments, the above step S205 includes: Step S2051, based on the target configuration parameter set and the next-day prediction data set, a basic charging and discharging strategy of the distributed photovoltaic system is generated through a deep reinforcement learning method.

[0095] In an optional embodiment, the deep reinforcement learning method is used to balance the single-period net benefit and battery loss cost of the distributed photovoltaic system.

[0096] In an optional embodiment, based on the target configuration parameters as constraints and the next-day prediction data, the single-period net benefit and battery loss are balanced through deep reinforcement learning, so as to generate a basic charging and discharging strategy adapted to the next-day predicted working condition.

[0097] In an optional embodiment, the next-day predicted energy storage SOC, hourly photovoltaic output, load demand, and time-of-use electricity price are used as the state space, the charging and discharging power (-Pmax~Pmax, negative value for discharging) is used as the action space, and the single-period net income-battery loss cost (positively correlated with the charging and discharging depth and frequency) is used as the reward function.

[0098] Further, the 1000+ user historical operation data are used as the training sample to train the deep reinforcement learning (DQN) model. Then, the target configuration parameters (Eopt, Pmax) and the next-day prediction data set are input, the model iteratively learns the optimal action, and then outputs the basic charging and discharging strategy.

[0099] Further, the basic charging and discharging strategy determines the charging and discharging power and SOC control interval of each period of the next day, and the core logic is: preferentially charging when the photovoltaic output > load and the electricity price is in the low valley; preferentially discharging when the load > photovoltaic output and the electricity price is in the peak.

[0100] Step S2052, obtaining a first real-time operation data set of the distributed photovoltaic system based on the basic charging and discharging strategy.

[0101] In an optional embodiment, the first real-time operation data set can include actual hourly photovoltaic output, user load demand, real-time electricity price, current energy storage SOC, battery temperature, and actual charging and discharging power.

[0102] In an optional embodiment, a real-time data acquisition terminal is deployed, and when the distributed photovoltaic system operates according to the basic charging and discharging strategy, the operation data of the distributed photovoltaic system are collected at a time granularity of 5 minutes, and the corresponding first real-time operation data set is formed.

[0103] Step S2053, determining a target correction strategy using the first real-time operation data set and the next-day prediction data set.

[0104] In an optional embodiment, the first real-time operation data and the next-day prediction data are compared on a time-by-time basis, and deviation scenarios are identified.

[0105] For example, the actual photovoltaic output is more than 10% lower than the prediction, the real-time electricity price is 5% higher than the predicted value, and the load drops suddenly.

[0106] Further, the target correction strategy can be determined based on a preset deviation-correction mapping rule.

[0107] For example, if the actual photovoltaic output is more than 10% lower than the prediction: reduce the charging amount and reserve the capacity to cope with the evening load; if the real-time electricity price surges (5% higher than the predicted value): discharge in advance to lock in high income; if a sudden load drop (such as a holiday user stop production) is detected: suspend discharging and switch to valley charging.

[0108] Step S2054, using the target correction strategy to correct the basic charging and discharging strategy in real time to obtain a target charging and discharging strategy.

[0109] In an optional embodiment, the adjustment instructions in the target correction strategy are used to replace or correct the corresponding content of the basic charging and discharging strategy by time period. Then, it is verified whether the corrected strategy meets the target configuration parameter constraints and the battery safety requirements.

[0110] Further, the fused strategy is converted into specific and executable quantitative control instructions, and the corresponding corrected target charging and discharging strategy is output.

[0111] In some optional embodiments, the above step S206 includes: Step S2061, when the distributed photovoltaic system is controlled to operate based on the target charging and discharging strategy, the battery state of the distributed photovoltaic system is obtained by using an artificial intelligence model.

[0112] In an optional embodiment, a light neural network model is deployed at the edge, and battery operation data is collected at a time granularity of 5 minutes, which can include real-time operation data such as battery SOC, charging and discharging current, voltage, battery temperature, cycle number, and charging and discharging depth.

[0113] Further, the model outputs battery health state results, which can include three categories of health, sub-health, and unhealth, and gives a SOH (battery health degree) quantitative value.

[0114] The training samples can include the operation characteristics of new batteries, aged batteries, and faulty batteries.

[0115] Step S2062, when the battery state is in a healthy state, the operation of the distributed photovoltaic system is controlled by using the target charging and discharging strategy to obtain a storage operation result.

[0116] In an optional embodiment, when the battery state is in a healthy state, the quantitative instructions (charging and discharging power of each time period, SOC control interval) in the target charging and discharging strategy are issued to execution devices such as a power conversion system (PCS) to control the energy storage system to charge and discharge according to the plan.

[0117] Further, the real-time monitoring system cooperates with the operation state to ensure that the photovoltaic output, load demand, and energy storage charging and discharging action are accurately matched, and there is no power conflict or out-of-limit situation.

[0118] Further, full-dimensional operation data can be collected at a time granularity of 5 minutes, which can include photovoltaic actual output, user actual load, energy storage charging and discharging amount, SOC change curve, hourly income, and battery operation parameters, and form a corresponding storage operation result.

[0119] Step S2063, when the battery state is an unhealthy state, adjusting the operation parameters of the distributed photovoltaic system based on the battery state, and obtaining the storage operation result.

[0120] In an optional embodiment, when the battery state is an unhealthy state, the corresponding adjustment rule can be determined according to the unhealthy reason output by the artificial intelligence model.

[0121] For example, the adjustment rule can include: 1. SOC control: dynamically adjusting the interval according to the number of battery cycles (new battery 20%-90%, aged battery 30%-80%); 2. Temperature protection: when the battery temperature is greater than or equal to 35°C, reduce the charging and discharging power to 50%, and when the battery temperature is greater than or equal to 40°C, suspend charging and discharging; 3. Charging and discharging depth optimization: based on the battery health (SOH) adjustment, when SOH is less than or equal to 80%, limit the depth of charging and discharging (single charging and discharging depth is less than or equal to 50%).

[0122] Further, according to the matched rule, the operation parameters such as charging and discharging power and SOC interval are adjusted, temporary control instructions are generated and executed, and it is ensured that the adjusted parameters meet the safe operation boundary of the battery. Further, the photovoltaic and energy storage system is maintained in basic coordinated operation under the adjusted parameters, and the change of the battery state is monitored to avoid the aggravation of the abnormality.

[0123] Further, the operation data of the adjusted system is collected and the corresponding storage operation result is formed.

[0124] In some optional embodiments, the above method further comprises: based on the storage operation result, updating the gradient boosting tree classification model, the bidirectional long short-term memory network model, the improved genetic algorithm and the dynamic adaptive control method respectively.

[0125] In an optional embodiment, by inputting the actual storage operation result into each core model and control method in reverse, and dynamically updating the model parameters and optimizing the algorithm logic, the accuracy of the storage decision, the economic efficiency of the capacity configuration and the adaptability of the operation strategy can be continuously improved, thereby realizing the continuous iterative upgrade of the system performance.

[0126] In an optional embodiment, the actual storage income, the change of the light rejection rate, the actual matching degree of load-photovoltaic in the current storage operation result are extracted, and combined with the multi-dimensional feature parameters of the original training sample, new labeled samples are added, and the training data set is expanded.

[0127] Further, the expanded data set is input into the original GBDT model, and an incremental training method is used to adjust the weight and split node of the decision tree based on the new sample. Further, the learning rate, tree depth and other hyperparameters are re-optimized through grid search to minimize the classification loss function of the new sample set.

[0128] Further, the classification accuracy of the updated model is verified by using the test set, and if it meets the standard, the original model is replaced, and the threshold calibration logic of the storage necessity decision is updated synchronously.

[0129] In an optional embodiment, the actual data in the storage operation result is compared with the original prediction data of the model to calculate the prediction error of each period, and then the abnormal period data with an error exceeding 10% is screened out, and the error reason is analyzed.

[0130] Further, the screened error data, actual operation data and newly added historical data (such as operation records under extreme weather) are integrated into the training set, the Bi-LSTM model is fine-tuned and trained, and the network hidden unit number, iteration number and other hyperparameters are optimized to improve the model's ability to capture extreme conditions and policy changes.

[0131] Further, the updated model is used to re-predict future short-term data to verify whether the prediction error is ≤10%; if it meets the standard, the updated model parameters are solidified for subsequent multi-dimensional prediction and next-day prediction data set generation.

[0132] In an optional embodiment, based on the actual full life cycle net income and investment payback period data in the storage operation result, the parameter weight of the target optimization function is corrected. Further, if there is a constraint boundary adaptation problem in actual operation, the boundary value of the preset constraint condition set is fine-tuned.

[0133] Further, the deviation of the original solution result (target configuration parameter set) of the improved genetic algorithm and the actual operation optimal parameter is calculated, and the probability threshold of the adaptive crossover mutation operator is adjusted. Further, if there is a low optimization efficiency problem, the diversity of population initialization is increased, and the convergence time is shortened.

[0134] Further, the updated algorithm is used to re-solve the target optimization function to verify whether the deviation between the simulated income corresponding to the output target configuration parameter set and the actual income is ≤5%, and after meeting the standard, the updated algorithm is officially enabled.

[0135] In an optional embodiment, the update of the dynamic self-adaptive control method can include: (1) Day-ahead plan optimization rule update: analyze the adaptation difference between the basic charging and discharging strategy and the actual working condition in the storage operation result, adjust the reward function weight of deep reinforcement learning, and expand the subdivision dimension of the action space.

[0136] (2) Real-time correction rule iteration: statistics the actual correction effect of different deviation scenarios (such as insufficient photovoltaic output and load sudden drop), add correction rules corresponding to high-frequency deviation types, and optimize the input features of the lightweight neural network to improve the response speed of deviation identification and correction.

[0137] (3) Battery health management strategy optimization: Based on the actual health state changes of the battery in the storage operation result, the health management parameters are adjusted, and the dynamic adjustment logic of the SOC control interval is updated.

[0138] (4) Update and monitoring: The updated regulation and strategy are integrated into the two-stage operation framework, the operation effect is monitored through short-term trial operation, and the update of the dynamic adaptive regulation method is completed after confirming the standard.

[0139] In an example, an artificial intelligence-based distributed photovoltaic storage and operation method based on artificial intelligence aims to scientifically solve the "necessity" and "economy" problems of energy storage configuration through data-driven intelligent algorithms, and on this basis, to realize the adaptive and forward-looking economic operation of the energy storage system through advanced AI models, and ultimately to achieve the goal of the lowest life cycle cost and the maximum benefit.

[0140] Specifically, the present example aims to solve the problems of lack of scientific basis, poor adaptability of photovoltaic storage system operation strategy, and fragmented application of AI technology in existing distributed photovoltaic storage methods, and realizes the following through artificial intelligence technology: (1) Based on historical data such as user load, photovoltaic characteristics, electricity price policy, and investment economic and technical analysis model, through a machine learning classification model, accurately determine whether to configure energy storage and its capacity; (2) Combine real-time data and artificial intelligence algorithms to dynamically generate the optimal economic operation strategy, maximize the energy storage benefit and prolong the battery life.

[0141] In an optional embodiment, as shown in Figure 3 and Figure 4 The artificial intelligence-based distributed photovoltaic storage and operation method based on artificial intelligence provided by the present example constructs a "data-driven-intelligent decision-making-dynamic optimization" technical framework, including an intelligent storage decision-making module and an adaptive operation optimization module, and the specific steps are as follows: 1. Intelligent storage decision-making module (solving "whether to store" and "how much capacity").

[0142] 1.1 Multi-source data acquisition and feature engineering: Acquire key user-side data (time granularity: 5 minutes for industrial and commercial enterprises): 1.1.1 Load data: historical active power, peak and valley period distribution, daily / monthly load fluctuation coefficient; 1.1.2 Photovoltaic data: installed capacity, historical output curve, self-use power, grid-connected power and proportion, and abandoned light record; 1.1.3 Market data: time-of-use electricity price (peak / flat / valley period and price), photovoltaic grid-connected electricity price, subsidy policy, capacity price; 1.1.4 Environmental data: local light intensity, temperature, precipitation, etc. meteorological data (in the past 3 years), seasonal variation law.

[0143] 1.2 Feature engineering processing: 1.2.1 Data cleaning: using isolation forest algorithm to identify outliers, and repairing missing values by linear interpolation; 1.2.2 Feature extraction: build a core feature set, including "photovoltaic-load matching degree" (the proportion of the coincidence duration of photovoltaic output and load), "peak-valley price difference coefficient" (peak-valley price difference / flat section price), "abandoned light loss rate" (abandoned light power x on-grid electricity price / total power generation x electricity price) and other multi-dimensional feature parameters.

[0144] 1.3 Necessary storage based on machine learning.

[0145] Construct a gradient boosting decision tree (GBDT) classification model to realize "storage / no storage" binary classification decision: (1) Input: multi-dimensional feature parameters extracted in section 1.2.2; (2) Output: storage necessity probability (0-100%); (3) Decision threshold: probability ≥ 50% is determined as "need storage" (storage benefit covers cost), otherwise "no need storage".

[0146] Model training and optimization: (1) Training samples: 1000+ distributed photovoltaic user data (including storage after benefit data), including positive samples (need storage) and negative samples (no need storage); (2) Model optimization: optimize hyperparameters (learning rate, tree depth, etc.) through grid search to make classification accuracy ≥ 90%.

[0147] 1.4 Optimal capacity calculation based on deep learning. For users determined to "need storage", the optimal storage capacity is calculated by the following steps: 1.4.1 Multi-dimensional prediction: in the distributed photovoltaic storage system, a bidirectional LSTM model is used to predict key parameters in the next year: (1) Hourly load (error ≤ 10%), photovoltaic output (error ≤ 10%); (2) Electricity price policy adjustment trend (such as peak-valley period change probability, user electricity price rise, on-grid electricity price in power market).

[0148] The bidirectional LSTM model can be regarded as the stacking of two unidirectional LSTM models, one forward and one backward. For each time step, the forward and backward hidden states are combined to form the final hidden state of the time step. The reasons for using this model are mainly two aspects; (1) Photovoltaic output is affected by weather, season and other factors, both daily and seasonal. Bi-LSTM can capture both forward and backward dependencies.

[0149] (2) There is uncertainty in photovoltaic and load forecasting, and the bidirectional structure helps the model learn confidence information from the complete temporal context.

[0150] The input of the model includes historical photovoltaic output, light intensity, temperature, historical load demand, workday / holiday, time-of-use price, real-time price, etc. The output of the model includes photovoltaic output prediction data, load demand prediction data, and price prediction data, as well as prediction uncertainty and extreme event probability.

[0151] Compared with ordinary unidirectional LSTM model, bidirectional LSTM model can better capture bidirectional dependency information (such as night load affecting daytime energy storage decision), making distributed photovoltaic storage more accurate and reliable.

[0152] 1.4.2 Capacity optimization model: The optimization function is constructed with the goal of maximizing the net benefit in the whole life cycle, as shown in the above relationship (1).

[0153] 1.4.3 Capacity solution: The improved genetic algorithm (introducing adaptive crossover and mutation operators) is used to solve the optimal capacity Eopt, where the constraint conditions include charge and discharge power constraints (relationship (2) and (3) above), SOC constraints (relationship (4) above), and economic constraints (investment payback period ≤ 4 years).

[0154] 2 Adaptive operation optimization module (based on configured dynamic operation strategy).

[0155] 2.1 Two-stage operation strategy framework. Combined with reinforcement learning and real-time data, realize the dynamic regulation and control of "day-ahead plan + real-time correction": 2.1.1 Day-ahead optimization plan: Based on the next day's prediction data (load, photovoltaic output, price) of Bi-LSTM model, a basic strategy is generated using deep reinforcement learning (DQN): (1) State space: predicted SOC, photovoltaic output, load, price; (2) Action space: charge and discharge power (-Pmax~Pmax, negative value for discharge); (3) Reward function: single-period net benefit - battery wear cost (positively correlated with charge and discharge depth and frequency).

[0156] (4) Core logic: prefer to charge when "photovoltaic output > load" and price is low; prefer to discharge when "load > photovoltaic output" and price is high.

[0157] 2.1.2 Real-time dynamic correction: edge deployment of lightweight neural network, real-time data collection every 5 minutes (actual load, photovoltaic output, weather mutation), correction of charging and discharging plan: If the actual photovoltaic output is more than 10% lower than the prediction: reduce the charging capacity, reserve capacity to cope with evening load; if the real-time electricity price surges (more than 5% of the predicted value): discharge in advance, lock high yield; if a load drop is detected (such as holiday user stop production): suspend discharging and switch to valley charging.

[0158] 2.2 Battery health management strategy.

[0159] Real-time monitoring of battery status through AI model, dynamic adjustment of operating parameters: based on the running optimization of 2.1, the battery will give some boundary operating conditions, such as frequent charging and discharging will cause the battery to heat up, leading to unsafe state of the battery. In fact, it is based on the state of running optimization, while also considering the bearing capacity of the battery. Specifically includes: 1. SOC control: dynamically adjust the interval according to the number of battery cycles (new battery 20%-90%, aged battery 30%-80%); 2. Temperature protection: when the battery temperature is ≥35℃, reduce the charging and discharging power to 50%, ≥40℃, suspend charging and discharging; 3. Charging and discharging depth optimization: based on battery health (SOH) adjustment, SOH≤80% limit deep charging and discharging (single charging and discharging depth≤50%).

[0160] The AI-based distributed photovoltaic storage and operation method provided in this example has the following beneficial effects: (1) Precise storage decision: based on historical data of user load, photovoltaic characteristics, electricity price policy, and investment economic and technical analysis model, through machine learning classification model, accurately judge whether to configure storage and its capacity; (2) Maximize economic benefits: adaptive operation strategy makes the net benefit of storage in the whole life cycle increase by more than 20% compared with traditional methods, and the investment recovery period is shortened by more than 2 years; (3) Strong adaptability and generalization: AI model can learn different regional electricity price policies, user load characteristics, and adapt to different types of users in multiple scenarios; (4) Prolong battery life: through health management strategy, battery cycle life is improved by more than 20%.

[0161] A distributed photovoltaic storage and operation device based on artificial intelligence is also provided in the embodiment, which is used to implement the above-mentioned embodiments and preferred embodiments, and has been described above. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware, or a combination of software and hardware is also possible and contemplated.

[0162] The embodiment provides a distributed photovoltaic storage and operation device based on artificial intelligence, as shown in the drawings, which comprises: Figure 5 An acquisition module 501 is configured to acquire an initial multi-source data set and a next-day prediction data set of a distributed photovoltaic system on a user side.

[0163] A first processing module 502 is configured to obtain a storage necessity probability value through feature engineering and a gradient boosting tree classification model based on the initial multi-source data set.

[0164] A judgment module 503 is configured to determine whether the user side needs storage based on the storage necessity probability value.

[0165] A configuration module 504 is configured to perform optimal energy storage configuration on the distributed photovoltaic system by using a bidirectional long short-term memory network model and an improved genetic algorithm to obtain a target configuration parameter set when the user side needs storage, wherein the bidirectional long short-term memory network model is obtained by stacking a forward long short-term memory network model and a reverse long short-term memory network model.

[0166] A second processing module 505 is configured to obtain a target charging and discharging strategy of the distributed photovoltaic system through a dynamic self-adaptive regulation method based on the target configuration parameter set and the next-day prediction data set.

[0167] A control module 506 is configured to control the operation of the distributed photovoltaic system by using the target charging and discharging strategy to obtain a storage operation result.

[0168] The distributed photovoltaic storage and operation device based on artificial intelligence provided by the embodiment can execute the distributed photovoltaic storage and operation method based on artificial intelligence provided by any embodiment of the present application, and has the corresponding function modules and beneficial effects of the execution method. The further function description of each module and unit is the same as that of the corresponding embodiment, and will not be repeated here.

[0169] Figure 6 A structure schematic diagram of an electronic device is provided for the embodiment.

[0170] The following will be specifically described with reference to Figure 6 ​which shows a structural schematic diagram suitable for being used to implement the electronic device in the embodiments of the present application. The electronic device can include a processor (such as a central processor, a graphic processor, etc.) 601, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 602 or programs loaded from a memory 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device are also stored. The processor 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0171] Generally, the following devices can be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a memory 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 can allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device with various devices is shown, but it should be understood that it is not required to implement or have all the shown devices, and more or fewer devices can be alternatively implemented or had.

[0172] In particular, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the embodiments of the present application include a computer program product including a computer program carried on a non-transitory computer readable medium, the computer program containing program codes for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network through the communication device 609, or installed from the memory 608, or installed from the ROM 602. When the computer program is executed by the processor 601, the above-mentioned functions defined in the artificial intelligence-based distributed photovoltaic power storage and operation method of the embodiments of the present application are performed.

[0173] Figure 6 The electronic device shown is only an example and should not bring any limitation to the functions and use range of the embodiments of the present application.

[0174] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the artificial intelligence-based distributed photovoltaic power generation and storage operation method shown in the above embodiments is implemented.

[0175] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0176] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A distributed photovoltaic power generation and storage system operation method based on artificial intelligence, characterized in that, The method includes: Obtain the initial multi-source dataset of the distributed photovoltaic system on the user side; Based on the initial multi-source dataset, the probability value of storage necessity is obtained after feature engineering and gradient boosting tree classification model processing. Determine whether the user side needs storage based on the probability value of storage necessity. When the user side needs to allocate energy storage, the optimal energy storage configuration of the distributed photovoltaic system is performed using a bidirectional long short-term memory network model and an improved genetic algorithm, and the next day's prediction dataset and target configuration parameter set are obtained. The bidirectional long short-term memory network model is obtained by stacking a forward long short-term memory network model and a backward long short-term memory network model. Based on the target configuration parameter set and the next day's prediction dataset, the target charging and discharging strategy of the distributed photovoltaic system is obtained through dynamic adaptive control method. The operation of the distributed photovoltaic system is controlled using the target charging and discharging strategy to obtain the distribution and storage operation results.

2. The method according to claim 1, characterized in that, Based on the initial multi-source dataset, after feature engineering and gradient boosting tree classification model processing, the probability value of storage necessity is obtained, including: The multi-source dataset is cleaned to obtain the target multi-source dataset; Feature extraction is performed on the target multi-source dataset to obtain a multi-dimensional feature parameter set; The multidimensional feature parameters are input into the gradient boosting tree classification model for processing to obtain the probability value of the necessity of storage.

3. The method according to claim 1, characterized in that, When the user side requires energy storage, a bidirectional long short-term memory network model and an improved genetic algorithm are used to optimize the energy storage configuration of the distributed photovoltaic system, resulting in a next-day prediction dataset and a target configuration parameter set, including: When the user side needs to allocate storage, the target optimization function and the user side's multi-dimensional dataset and next day's dataset are obtained. The target optimization function is used to characterize the maximum net income over the entire life cycle. The multidimensional dataset and the next-day dataset are respectively input into the bidirectional long short-term memory network model for processing to obtain the multidimensional prediction dataset and the next-day prediction dataset. Based on the multi-dimensional prediction dataset and the preset constraint set, the improved genetic algorithm is used to solve the objective optimization function to obtain the objective configuration parameter set.

4. The method according to claim 1, characterized in that, Based on the target configuration parameter set and the next-day prediction dataset, the target charging and discharging strategy of the distributed photovoltaic system is obtained through dynamic adaptive control methods, including: Based on the target configuration parameter set and the next day prediction dataset, the basic charging and discharging strategy of the distributed photovoltaic system is generated through deep reinforcement learning. The deep reinforcement learning method is used to control the balance between the single-period net income and battery loss cost of the distributed photovoltaic system. Obtain the first real-time running dataset of the distributed photovoltaic system based on the basic charging and discharging strategy; Using the first real-time running dataset and the next day's predicted dataset, a target correction strategy is determined; The target charging and discharging strategy is obtained by using the target correction strategy to correct the basic charging and discharging strategy in real time.

5. The method according to claim 1, characterized in that, Using the target charge and discharge strategy, the operation of the distributed photovoltaic system is controlled to obtain the distribution and storage operation results, including: When the distributed photovoltaic system is controlled to operate based on the target charging and discharging strategy, the battery status of the distributed photovoltaic system is obtained using an artificial intelligence model; When the battery is in a healthy state, the operation of the distributed photovoltaic system is controlled using the target charging and discharging strategy to obtain the distribution and storage operation results; When the battery is in an unhealthy state, the operating parameters of the distributed photovoltaic system are adjusted based on the battery state, and the distribution and storage operation results are obtained.

6. The method according to claim 5, characterized in that, The method further includes: Based on the results of the storage allocation operation, the gradient boosting tree classification model, the bidirectional long short-term memory network model, the improved genetic algorithm, and the dynamic adaptive regulation method are updated respectively.

7. A distributed photovoltaic power generation, storage, and operation device based on artificial intelligence, characterized in that, The device includes: The acquisition module is used to acquire the initial multi-source dataset and the next day's prediction dataset of the distributed photovoltaic system on the user side; The first processing module is used to obtain the probability value of storage necessity based on the initial multi-source dataset through feature engineering and gradient boosting tree classification model. The judgment module is used to determine whether the user side needs to allocate storage based on the probability value of storage necessity. The configuration module is used to perform optimal energy storage configuration for the distributed photovoltaic system when the user side needs to configure energy storage, using a bidirectional long short-term memory network model and an improved genetic algorithm to obtain a target configuration parameter set. The bidirectional long short-term memory network model is obtained by stacking a forward long short-term memory network model and a reverse long short-term memory network model. The second processing module is used to obtain the target charging and discharging strategy of the distributed photovoltaic system by processing the target configuration parameter set and the next day's prediction dataset through a dynamic adaptive control method. The control module is used to control the operation of the distributed photovoltaic system using the target charging and discharging strategy to obtain the distribution and storage operation results.

8. An electronic device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the distributed photovoltaic power generation and storage and operation method based on artificial intelligence as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to execute the artificial intelligence-based distributed photovoltaic power generation and storage and operation method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute any one of the artificial intelligence-based distributed photovoltaic power generation and storage and operation methods as described in any one of claims 1 to 6.

Images