Modular photovoltaic carport micro-grid system based on ai light storage cooperation

Abstract

This invention relates to the field of smart microgrid technology, specifically disclosing a modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy. The system includes: a hardware construction module, a load forecasting module, a photovoltaic power generation forecasting module, and a global optimization module. The system constructs a modular photovoltaic carport, collects multi-source operating data using a deployed sensor network, sets a load forecasting cycle, builds a historical charging database and performs periodic statistical forecasts, generates a charging probability distribution for each track section within the load forecasting cycle, determines whether to trigger mobile charging terminal location adjustments, divides the control cycle based on the load forecasting cycle, constructs and uses a photovoltaic time-series forecasting model for prediction, generates a photovoltaic power generation forecast curve within the control cycle, and obtains a global prediction data package within the control cycle by combining the charging probability distribution. Based on the global prediction data package, a multi-objective optimization function is constructed and solved, and the corresponding optimal control instruction set is output.

Description

Technical Field

[0001] This invention relates to the field of smart microgrid technology, specifically to a modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy. Background Technology

[0002] With the increasing proportion of new energy power generation, represented by photovoltaics, the integrated photovoltaic-storage-charging model, which combines photovoltaic power generation, energy storage, and electric vehicle charging, is seen as a key path for building a new power system. However, while the industry is expanding on a large scale, its underlying structural contradictions and coordination dilemmas are becoming increasingly prominent. After shifting from policy-driven to market-driven approaches, traditional technical solutions are no longer sufficient to support the requirements of high-quality and high-economic-efficiency development.

[0003] Existing technologies face several profound technical challenges: First, there is the dual challenge of inaccurate prediction and imprecise scheduling at the operational level. Photovoltaic power generation is inherently intermittent and volatile, and its output is significantly affected by weather. Traditional scheduling methods have limited accuracy in predicting photovoltaic power, especially on short timescales of minutes to hours, where the error is large. At the same time, charging loads are highly random, and traditional solutions struggle to effectively predict their spatiotemporal distribution. Both of these factors contribute to significant uncertainties at both the source and load ends, resulting in a lack of reliable data foundation for the optimal scheduling of microgrids. Existing energy management systems often employ rule-based control based on experience or fixed thresholds, leading to delayed responses and an inability to achieve dynamic and refined coordination between photovoltaic, energy storage, and charging loads at the millisecond to minute level. This directly results in poor system stability, low local renewable energy consumption rates, and large fluctuations in grid interaction power.

[0004] On the other hand, the charging service model is rigid and the resource utilization rate is low. Currently, photovoltaic carports generally use fixed charging piles, whose location and power are not changeable. This causes charging resources to form islands in space and time, which cannot adapt to the dynamic changes in parking and charging needs and affect the user experience.

[0005] To address the aforementioned issues, this invention proposes a modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy. Summary of the Invention

[0006] The purpose of this invention is to provide a modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy to solve the aforementioned background problems.

[0007] The objective of this invention can be achieved through the following technical solution: a modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy, comprising:

[0008] Hardware construction module: Construct a modular photovoltaic carport composed of standardized hardware components, and combine it with the deployment of a sensor network to collect multi-source operating data;

[0009] Load forecasting module: Sets the load forecasting period, builds a historical charging database based on multi-source operation data and performs periodic statistical forecasts, generates the charging probability value distribution of each track section within the load forecasting period, and determines whether to trigger the adjustment of the mobile charging terminal location.

[0010] Photovoltaic power generation prediction module: Based on the load prediction cycle, the control cycle is divided. Based on multi-source operation data, a photovoltaic time series prediction model is constructed and used for prediction to generate the photovoltaic power generation prediction curve within the control cycle. Combined with the charging probability value distribution, the global prediction data package within the control cycle is obtained.

[0011] Global optimization module: Based on global prediction data packets, it constructs and solves a multi-objective optimization function aimed at improving the stability of microgrid operation and reducing operating costs, outputs the corresponding optimal control command set and executes it;

[0012] Furthermore, the standardized hardware components include a standardized steel structure canopy, precast concrete foundations and embedded parts, an integrated photovoltaic system, modular energy storage units, and an intelligent rail-mounted charging system. The intelligent rail-mounted charging system includes a rail subsystem and a mobile charging terminal. The rail subsystem is a rigid sliding contact rail laid under the main beam of the canopy, which integrates high-voltage electricity and communication buses and is divided into multiple numbered rail sections. The mobile charging terminal is an integrated mobile device that integrates a servo-driven trolley, a highly flexible oil-resistant and flame-retardant cable, and an intelligent charging gun.

[0013] Furthermore, the charging probability value distribution is generated as follows:

[0014] At the beginning of each load forecast cycle, the charging probability value of each track segment within the current load forecast cycle is obtained and extracted based on the charging probability forecast matrix. Combined with real-time load-side data, the probability value corresponding to the track segment with charging demand is updated to 1, and finally the distribution of charging probability values ​​of each track segment within the load forecast cycle is obtained.

[0015] Furthermore, the charging probability prediction matrix is ​​obtained as follows:

[0016] Define time characteristics including date type and time period, and define historical runtime segments with the current time as the end point and a duration not exceeding the preset standard historical duration;

[0017] Based on the historical charging database, the average value of all instantaneous charging loads of any track segment within the historical operating period at the same time characteristic is used as the historical load characteristic. After normalizing the historical load characteristic, the charging probability value is obtained. All charging probability values ​​constitute a charging probability prediction matrix with track segment number as the row and time characteristic as the column.

[0018] Furthermore, the method for adjusting the location of the mobile charging terminal is as follows:

[0019] Compare the charging probability value of each track section with the preset probability standard, mark the track section where the charging probability value reaches the preset probability standard and there is no mobile charging terminal, and sort them in descending order of probability value. If there is an idle mobile charging terminal, the adjustment is triggered, and the nearest idle mobile charging terminal is assigned to the sorted hot spot section in order, and the terminal is moved to the corresponding hot spot section to wait.

[0020] Furthermore, the photovoltaic power generation prediction curve is generated as follows:

[0021] Multi-source operating data includes power supply side data, load side data, environmental data and grid interaction data. At the beginning of each control cycle, the environmental data time series and photovoltaic data time series within the set historical time window are extracted, normalized, and then input into the pre-trained photovoltaic time series prediction model.

[0022] The photovoltaic time series prediction model extracts local features through a one-dimensional convolutional layer, captures temporal dependencies through a gated recurrent unit layer, and finally outputs the photovoltaic power generation prediction sequence within a complete control cycle through a fully connected layer, i.e., the photovoltaic power generation prediction curve.

[0023] Furthermore, the optimal control instruction set is obtained as follows:

[0024] At the beginning of each control cycle, the mathematical programming solver is invoked to solve the constructed multi-objective optimization function;

[0025] Using the global prediction data package as known parameter input, the mathematical programming solver searches for the set of decision variables that minimizes the value of the multi-objective optimization function while satisfying all operational constraints, and encapsulates it into an optimal control instruction set.

[0026] Furthermore, the set of decision variables includes: the charging and discharging power of the modular energy storage unit at each moment during the control cycle, the real-time output power of each mobile charging terminal, and the connection point interaction power in the grid interaction data;

[0027] Furthermore, the multi-objective optimization function is constructed as follows:

[0028] The stability objective sub-function and the economic objective sub-function are constructed and then weighted and fused using preset weight coefficients to obtain a complete multi-objective optimization function;

[0029] The stability objective sub-function expression is the standard deviation of the interconnected power at the connection points within the control period, plus a penalty term for deviation of the real-time state of charge from the preset ideal center value. The economic objective sub-function expression is the cumulative sum of the products of the interconnected power at the connection points at each time point within the control period and the corresponding market electricity price.

[0030] Furthermore, the operational constraints include: power balance constraints, energy storage operation constraints, charging load constraints, and grid interaction constraints. The power balance constraint requires that at any given time, the absolute value of the deviation between the predicted photovoltaic power generation, the energy storage discharge power, and the grid interaction power (which constitute the input power) and the output power (which constitutes the output power of the mobile charging terminal, the energy storage charging power, and the grid interaction power) does not exceed the preset system loss safety balance margin.

[0031] The beneficial effects of this invention are as follows:

[0032] 1. This invention achieves significant engineering and economic advantages through the innovative integration of modular hardware design and intelligent rail-mounted charging system. Standardized prefabricated components and rapid assembly process support flexible expansion of carport size, greatly reducing initial investment and subsequent expansion costs. Mobile charging terminals are dynamically scheduled according to predicted demand, breaking the rigid resource limitations of fixed charging piles, improving the utilization rate of charging facilities, and significantly enhancing the overall efficiency of parking space and power capacity.

[0033] 2. This invention, based on AI-driven photovoltaic-storage collaborative optimization control, effectively improves the stability and economy of microgrid operation. By integrating photovoltaic output prediction and load probability prediction, the system achieves precise forward scheduling of uncertain source loads. With multi-objective optimization as its core control strategy, it can smooth power fluctuations at the microgrid grid connection point while ensuring that all charging needs are met, and reduce overall operating costs. This not only enhances the system's autonomy and power supply quality, but also creates better economic and environmental benefits by maximizing the local consumption of green electricity and participating in grid demand response. Attached Figure Description

[0034] The invention will now be further described with reference to the accompanying drawings.

[0035] Figure 1 This is a modular architecture diagram of a modular photovoltaic carport microgrid system based on AI-based photovoltaic-storage synergy, as described in an embodiment of the present invention.

[0036] Figure 2 This is a flowchart illustrating the specific steps of a modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy, as described in an embodiment of the present invention. Detailed Implementation

[0037] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0038] Example 1

[0039] Please see Figure 1 and Figure 2 As shown in the embodiment of the present invention, a modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy aims to solve the problems of rigid charging facilities during construction of existing photovoltaic carports, large power fluctuations and high operating costs due to the lack of synergy between photovoltaic, energy storage, and charging, as well as unreasonable allocation of charging resources. By constructing modular photovoltaic carports and deploying a sensor network to collect multi-source operating data, the system generates charging probability distribution values ​​for each track section based on a historical charging database and triggers pre-adjustment of mobile charging terminals. A photovoltaic time-series prediction model is used to generate photovoltaic power generation prediction curves, which are then integrated to obtain a global prediction data package. A multi-objective optimization function aimed at improving stability and reducing economic costs is constructed and solved, outputting the optimal control command set. This achieves dynamic synergy and optimal scheduling of photovoltaic and energy storage resources, comprehensively improving the autonomy, economy, and service efficiency of the microgrid. Specifically, it includes the following modules:

[0040] Hardware construction module: Construct a modular photovoltaic carport composed of standardized hardware components, and combine it with the deployment of a sensor network to collect multi-source operating data;

[0041] Specifically, the standardized hardware components include standardized steel structure canopy, precast concrete foundation and embedded parts, integrated photovoltaic system, modular energy storage unit and intelligent rail charging system;

[0042] The standardized steel structure shed uses factory-prefabricated standardized aluminum alloy profiles as the main beams and secondary beams of the shed. They are quickly assembled and connected by high-strength locking bolts to form a stable support structure. At the same time, the on-site pouring of the precast concrete foundation and the positioning and installation of the embedded parts are carried out. The embedded parts are equipped with standardized connecting flanges on the upper part for precise docking and fixing with the support columns of the upper steel structure shed.

[0043] It should be noted that the use of standardized design and prefabricated construction can significantly improve on-site construction efficiency, ensure structural accuracy and consistency, and support flexible expansion of the carport size.

[0044] The deployment of the integrated photovoltaic system involves laying photovoltaic power generation units on the top of the standardized steel structure shed. The photovoltaic power generation units are monocrystalline silicon photovoltaic modules of uniform specifications. All photovoltaic power generation units are connected in series and parallel in a pre-designed manner to form a photovoltaic array connected to an integrated DC combiner box. The output end is connected to an intelligent photovoltaic inverter installed on the side wall of the shed column. The voltage and current sensors built into the intelligent photovoltaic inverter monitor and calculate the photovoltaic power generation of the photovoltaic array in real time as photovoltaic data.

[0045] The modular energy storage unit is deployed in a dedicated equipment area on one side of the carport. The modular energy storage unit is a lithium-ion battery system with a standardized cabinet design. Each energy storage cabinet integrates a battery module, a battery management system and a bidirectional converter, and collects energy storage data in real time. This includes real-time monitoring of the modular energy storage unit's state of charge through the battery management system, and real-time calculation of the maximum chargeable and dischargeable power by the bidirectional converter through internal sensors.

[0046] Photovoltaic data and energy storage data together constitute power source-side data;

[0047] It should be noted that the role of modular energy storage units is to provide energy buffering and power support for microgrids, and they are key devices for achieving photovoltaic-storage synergy and power smoothing.

[0048] The intelligent track-type charging system includes a track subsystem and a mobile charging terminal. The track subsystem is a rigid sliding contact line track laid parallel to the parking space direction under the main beam of the carport. The rigid sliding contact line track integrates high-voltage electricity and communication bus, and provides a physical sliding track and power transmission channel for the mobile charging terminal. The track subsystem is divided into multiple track sections, and each track section is assigned a number. The mobile charging terminal is an integrated mobile device that integrates a servo-driven trolley, a highly flexible oil-resistant and flame-retardant cable, and an intelligent charging gun. The servo-driven trolley engages with the rigid sliding contact line track through a conductive slider on the top. The highly flexible oil-resistant and flame-retardant cable is wound on an electric drum inside the trolley and automatically retracts as the trolley moves. The intelligent charging gun is mounted under the trolley and integrates a card reader, a billing control unit, and a communication module for identifying the user and performing charging billing.

[0049] Within the intelligent track-based charging system, load-side data is collected in real time. The load-side data includes the track section number where each mobile charging terminal is located, the real-time output power and charging demand, and the charging demand includes the target power and the charging cut-off time.

[0050] It should be noted that the role of the intelligent rail-mounted charging system is to break the spatial limitations of fixed charging piles and realize the dynamic allocation and efficient utilization of charging resources in the carport through mobile services. It is the core of realizing flexible control on the load side.

[0051] Environmental sensing units are deployed on the top and shaded areas of the standardized steel structure shed to collect environmental data in real time, including horizontal light intensity, ambient temperature and photovoltaic panel back temperature. Smart meters and grid information receiving terminals are installed at the connection points between the modular photovoltaic shed and the public power grid to collect grid interaction data in real time, including the interaction power at the connection point and the market electricity price.

[0052] Power supply side data, load side data, environmental data, and grid interaction data together constitute the multi-source operation data of the modular photovoltaic carport;

[0053] It should be noted that the purpose of this step is to integrate the standardized and prefabricated construction concept with the integrated photovoltaic system, modular energy storage unit and mobile intelligent track-type charging system, and to build a physical carrier that integrates power generation, energy storage and dynamic charging, breaking through the traditional form of fixed charging pile carports.

[0054] Load forecasting module: Sets the load forecasting period, builds a historical charging database based on multi-source operation data and performs periodic statistical forecasts, generates the charging probability value distribution of each track section within the load forecasting period, and determines whether to trigger the adjustment of the mobile charging terminal location.

[0055] Specifically, a global historical charging database is defined. Based on multi-source operation data, after each charging task is completed, the charging event log is organized and recorded to the historical charging database. The charging event log is a set of parameters describing a complete charging process, including event ID, user ID, track section number where the servo-driven car stops, charging period, target power, actual charging power and average charging power.

[0056] It should be noted that the role of the historical charging database is to provide a long-term and standardized training and validation data foundation for building a data-driven mobile charging prediction model, which is a prerequisite for achieving accurate load prediction.

[0057] For each charging task, the average charging power in the corresponding charging event log is used as the instantaneous charging load of the track section at each moment during the charging period. If there is no charging task in the track section at any moment, the instantaneous charging load is 0.

[0058] Set the load forecasting period and define time characteristics, including date type and time period;

[0059] For example, the load forecasting cycle is set to 1 hour, the time characteristics include the time period of each hour within a day, and the date type includes weekdays, rest days and holidays;

[0060] Define a historical running segment with the current time as the endpoint and a duration not exceeding the preset standard historical duration. Calculate the mean of all instantaneous charging loads of any track segment within the historical running segment at the same time characteristic, and mark it as a historical load characteristic. Normalize all the calculated historical load characteristics to values ​​in the interval [0,1] as charging probability values. All charging probability values ​​together constitute a charging probability prediction matrix with track segment number as the row and time characteristic as the column.

[0061] The normalization process uses the minimum-maximum normalization method. For each historical load feature, the difference is processed with the minimum value among all corresponding historical load features, and the ratio is processed with the difference between the maximum and minimum values ​​among all corresponding historical load features to obtain the corresponding charging probability value.

[0062] For example, the charging probability value data is shown below:

[0063] Table 1: Examples of typical charging probability values ​​(excerpt from the table);

[0064]

[0065] At the beginning of each load forecast period, based on the charging probability forecast matrix and the time characteristics of the current load forecast period, the charging probability value of each track segment within the load forecast period is extracted. Based on the real-time load-side data, the charging demand of the mobile charging terminal corresponding to each track segment is extracted. If there is a track segment with charging demand, the charging probability value of the corresponding track segment is updated to 1, thus obtaining the distribution of the charging probability value of each track segment within the load forecast period.

[0066] It should be noted that the charging probability value of track sections with actual charging demand is updated to 1 because once there is a charging demand in a track section, the charging event of the track section is a deterministic event, not a probabilistic event. In the subsequent global optimization scheduling, the charging demand of the corresponding track section must be given priority. Therefore, its charging probability value is set to the highest value of 1 to clearly identify the load as a rigid demand in the prediction data packet and ensure that the optimization model can reserve sufficient power resources for it.

[0067] Compare the charging probability value of each track section with the preset probability standard. Track sections with a charging probability value that meets the preset probability standard and where no mobile charging terminal is located are marked as hotspot sections. The hotspot sections are sorted from highest to lowest probability value. If there are currently idle mobile charging terminals that have not been activated, the mobile charging terminal position adjustment is triggered. The nearest idle mobile charging terminal is assigned to the hotspot section in sequence, and the idle mobile charging terminal is moved to the corresponding hotspot section to stand by. The adjustment stops when there are no idle mobile charging terminals or no hotspot sections.

[0068] It should be noted that the preset probability standard is a configurable dynamic threshold, which is calculated based on the probability value of covering 80% of non-zero charging demand by statistically analyzing multi-source time-series data collected and stored within a fixed period of time in the past.

[0069] It should be noted that the role of generating the optimal adjustment path based on the probability prediction matrix is ​​to transform static charging facilities into dynamically allocated resources, enabling mobile charging terminals to be strategically pre-positioned in high-probability charging areas, thereby shortening user waiting time and improving the overall system response speed and service efficiency.

[0070] Photovoltaic power generation prediction module: Based on the load prediction cycle, the control cycle is divided. Based on multi-source operation data, a photovoltaic time series prediction model is constructed and used for prediction to generate the photovoltaic power generation prediction curve within the control cycle. Combined with the charging probability value distribution, the global prediction data package within the control cycle is obtained.

[0071] Specifically, the load forecasting period is evenly divided into control periods, and a photovoltaic time-series forecasting model is constructed and deployed. The photovoltaic time-series forecasting model is a time-series forecasting model based on the fusion of one-dimensional convolution and gated recurrent units. The structure of the photovoltaic time-series forecasting model includes a one-dimensional convolutional layer, a gated recurrent unit layer and a fully connected layer. The one-dimensional convolutional layer extracts patterns at different time scales by sliding multiple convolutional kernels, and the gated recurrent unit layer stacks two layers of gated recurrent units.

[0072] It should be noted that the role of the photovoltaic time series prediction model is to perform deep feature extraction and time series dependency modeling on the time series sequence of photovoltaic data and the corresponding environmental data within the historical operating period, so as to achieve the sequence prediction of photovoltaic power generation within the control period.

[0073] For example, if the load forecasting cycle is 1 hour, the preset control cycle duration is 15 minutes, ensuring that multiple optimized controls can be executed within a load forecasting cycle to adapt to rapid fluctuations in power supply and load.

[0074] Specifically, at the beginning of each control cycle, a historical time window with the current time as the end point and a fixed duration is extracted. The input sequence of the photovoltaic time series prediction model is constructed based on the length of the historical time window, including the environmental data time series and photovoltaic data time series within the historical time window. All input sequences are normalized to eliminate dimensions and accelerate the convergence of the photovoltaic time series prediction model. The normalization process adopts the minimum-maximum normalization method.

[0075] The pre-trained and deployed photovoltaic time-series prediction model is invoked. The normalized input sequence is input into the photovoltaic time-series prediction model. The photovoltaic time-series prediction model extracts local features of the input sequence through a one-dimensional convolutional layer, captures the long-term and short-term temporal dependencies within the input sequence through stacked gated recurrent unit layers, and maps the learned local features and temporal dependencies into a sequence of photovoltaic power generation prediction values ​​for each moment in a complete future control cycle through a fully connected layer. The photovoltaic power generation prediction curve within the control cycle is then obtained.

[0076] It should be noted that the photovoltaic time series prediction model is based on a preset model update cycle, and periodically uses the time series sequence of photovoltaic data and the corresponding environmental data time series as the training set for pre-training.

[0077] Align the photovoltaic power generation prediction curve within the control period with the corresponding charging probability distribution of each track segment, and combine the market electricity price in the grid interaction data within the control period with the real-time state of charge in the energy storage data at the start of the control period to obtain the global prediction data package for the control period.

[0078] Global optimization module: Based on global prediction data packets, it constructs and solves a multi-objective optimization function aimed at improving the stability of microgrid operation and reducing operating costs, outputs the corresponding optimal control command set and executes it;

[0079] Specifically, the set of decision variables is defined to include the charging and discharging power of the modular energy storage unit at each moment during the control cycle, the real-time output power of each mobile charging terminal, and the connection point interaction power in the grid interaction data;

[0080] It should be noted that the purpose of using the charging and discharging power of the modular energy storage unit, the output power of each mobile charging terminal, and the interaction power with the grid as decision variables is to provide the optimization algorithm with complete controllability, enabling it to coordinate the scheduling of power supply, energy storage, and load-side resources to achieve global optimal operation.

[0081] The multi-objective optimization function is composed of a weighted sum of a stability objective sub-function and an economic objective sub-function.

[0082] Specifically, the stability objective sub-function aims to minimize the power fluctuations within the microgrid and enhance its autonomy. Its expression is the standard deviation of the interconnected power at the connection points within the control period, plus a penalty term for deviations of the real-time state of charge from the energy storage data from the preset ideal center value. The economic objective sub-function aims to minimize the overall operating cost of the modular photovoltaic carport microgrid. Its expression is the cumulative sum of the products of the interconnected power at the connection points at each moment within the control period and the corresponding market electricity price. The stability objective sub-function and the economic objective sub-function are weighted and fused through preset weight coefficients to form a complete multi-objective optimization function.

[0083] The operation constraints are set, including power balance constraints, energy storage operation constraints, charging load constraints, and grid interaction constraints. Among them, the power balance constraint is the real-time power balance that the microgrid must meet at any time. That is, the absolute value of the deviation between the input power composed of the photovoltaic power generation prediction, energy storage discharge power and grid interaction power and the output power composed of the mobile charging terminal output power, energy storage charging power and grid interaction power does not exceed the preset system loss safety balance margin.

[0084] Among them, the system loss safety balance margin takes into account the unavoidable energy loss of lines and converters in actual operation. Based on the line impedance, power electronic converter efficiency and the accuracy range of measuring instruments in the microgrid, the allowable deviation threshold is pre-set to accommodate the inherent energy loss and prediction error of the system, and to ensure that the optimal scheduling model always has a feasible solution.

[0085] Among them, the energy storage operation constraints include that the absolute value of the energy storage charging and discharging power of the modular energy storage unit at any sampling time does not exceed the maximum chargeable and dischargeable power, the real-time state of charge of the modular energy storage unit must always be between the preset safety upper and lower limits during the control cycle, the charging load constraint is that for any mobile charging terminal with charging demand, its output power during the control cycle must not exceed the rated maximum power and must meet the charging demand set by the user, and the grid interaction constraint is that the interaction power of the connection point must be within the transmission limit range of the physical line and the contract.

[0086] Based on the constructed multi-objective optimization function and operational constraints, a mathematical programming solver is called at the beginning of each control cycle to solve the problem. The global prediction data packet is used as the known parameter input. Under the condition of satisfying all operational constraints, the mathematical programming solver finds the set of decision variables that minimizes the value of the multi-objective optimization function, encapsulates it into the optimal control instruction set, issues the optimal control instruction set and executes it within the control cycle, thereby realizing the dynamic coordination of optical and energy storage resources.

[0087] It should be noted that the purpose of this step is to construct a multi-objective optimization function with stability and economy as the objectives, and to comprehensively consider all operational constraints such as power balance constraints and energy storage operation constraints. The optimal control instruction set is obtained by using a mathematical programming solver, which ultimately realizes the dynamic coordination of photovoltaic and energy storage resources, and improves the economic efficiency and autonomy of operation while ensuring charging demand.

[0088] The technical solution of this invention is as follows: A modular photovoltaic vehicle shed composed of standardized hardware components is constructed; multi-source operating data is collected by deploying a sensor network; a load prediction cycle is set; a historical charging database is built based on the multi-source operating data and periodic statistical predictions are performed; the charging probability distribution of each track section within the load prediction cycle is generated; it is determined whether the mobile charging terminal position adjustment is triggered; a control cycle is divided based on the load prediction cycle; a photovoltaic time-series prediction model is constructed and used for prediction based on the multi-source operating data; a photovoltaic power generation prediction curve within the control cycle is generated; a global prediction data package within the control cycle is obtained by combining the charging probability distribution; based on the global prediction data package, a multi-objective optimization function aimed at improving the stability of microgrid operation and reducing operating costs is constructed and solved; and the corresponding optimal control instruction set is output and executed.

[0089] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. An AI-light storage coordination-based modular photovoltaic carport micro-grid system, characterized in that: Includes the following modules: Hardware construction module: Construct a modular photovoltaic carport composed of standardized hardware components, and combine it with the deployment of a sensor network to collect multi-source operating data; Load forecasting module: Sets the load forecasting period, builds a historical charging database based on multi-source operation data and performs periodic statistical forecasts, generates the charging probability value distribution of each track section within the load forecasting period, and determines whether to trigger the adjustment of the mobile charging terminal location. Photovoltaic power generation prediction module: Based on the load prediction cycle, the control cycle is divided. Based on multi-source operation data, a photovoltaic time series prediction model is constructed and used for prediction to generate the photovoltaic power generation prediction curve within the control cycle. Combined with the charging probability value distribution, the global prediction data package within the control cycle is obtained. Global optimization module: Based on global prediction data packets, it constructs and solves a multi-objective optimization function aimed at improving the stability of microgrid operation and reducing operating costs, outputs the corresponding optimal control command set and executes it.

2. The AI-based photovoltaic carport microgrid system with light storage coordination according to claim 1, characterized in that: The standardized hardware components include a standardized steel structure canopy, precast concrete foundations and embedded parts, an integrated photovoltaic system, modular energy storage units, and an intelligent track-type charging system. The intelligent track-type charging system includes a track subsystem and a mobile charging terminal. The track subsystem is a rigid sliding contact line track laid under the main beam of the canopy, which integrates high-voltage electricity and communication buses and is divided into multiple numbered track sections. The mobile charging terminal is an integrated mobile device that integrates a servo-driven trolley, a highly flexible oil-resistant and flame-retardant cable, and an intelligent charging gun. 3.The AI-light-storage-collaborative-based modular photovoltaic carport micro-grid system according to claim 2, characterized in that: The charging probability value distribution is generated as follows: At the beginning of each load forecast cycle, the charging probability value of each track segment within the current load forecast cycle is obtained and extracted based on the charging probability forecast matrix. Combined with real-time load-side data, the probability value corresponding to the track segment with charging demand is updated to 1, and finally the distribution of charging probability values ​​of each track segment within the load forecast cycle is obtained.

4. The AI-based photovoltaic carport microgrid system with light storage coordination of claim 3, wherein: The charging probability prediction matrix is ​​obtained as follows: Define time characteristics including date type and time period, and define historical runtime segments with the current time as the end point and a duration not exceeding the preset standard historical duration; Based on the historical charging database, the average value of all instantaneous charging loads of any track segment within the historical operating period at the same time characteristic is used as the historical load characteristic. After normalizing the historical load characteristic, the charging probability value is obtained. All charging probability values ​​constitute a charging probability prediction matrix with track segment number as the row and time characteristic as the column.

5. A modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy, as described in claim 3, characterized in that: The method for adjusting the location of the mobile charging terminal is as follows: Compare the charging probability value of each track segment with the preset probability standard. Track segments with a charging probability value that meets the preset probability standard and where no mobile charging terminal is located are marked as hotspot segments and sorted in descending order of probability value. If there is an idle mobile charging terminal, an adjustment is triggered, and the nearest idle mobile charging terminal is assigned to the sorted hotspot segments in sequence, and the terminal is moved to the corresponding hotspot segment to stand by.

6. A modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy, as described in claim 1, characterized in that: The photovoltaic power generation prediction curve is generated as follows: Multi-source operating data includes power supply side data, load side data, environmental data and grid interaction data. At the beginning of each control cycle, the environmental data time series and photovoltaic data time series within the set historical time window are extracted, normalized, and then input into the pre-trained photovoltaic time series prediction model. The photovoltaic time series prediction model extracts local features through a one-dimensional convolutional layer, captures temporal dependencies through a gated recurrent unit layer, and finally outputs the photovoltaic power generation prediction sequence, i.e., the photovoltaic power generation prediction curve, through a fully connected layer.

7. A modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy, as described in claim 2, characterized in that: The optimal control instruction set is obtained as follows: At the beginning of each control cycle, the mathematical programming solver is invoked to solve the constructed multi-objective optimization function; Using the global prediction data package as known parameter input, the mathematical programming solver searches for the set of decision variables that minimizes the value of the multi-objective optimization function while satisfying all operational constraints, and encapsulates it into an optimal control instruction set.

8. A modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy, as described in claim 7, characterized in that: The set of decision variables includes: the charging and discharging power of the modular energy storage unit at each moment during the control cycle, the real-time output power of each mobile charging terminal, and the connection point interaction power in the grid interaction data.

9. A modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy, as described in claim 8, characterized in that: The multi-objective optimization function is constructed as follows: The stability objective sub-function and the economic objective sub-function are constructed and then weighted and fused using preset weight coefficients to obtain a complete multi-objective optimization function; The stability objective sub-function is expressed as the standard deviation of the interconnected power at the connection points within the control period, plus a penalty term for deviation of the real-time state of charge from the preset ideal center value. The economic objective sub-function is expressed as the cumulative sum of the products of the interconnected power at the connection points at each time point within the control period and the corresponding market electricity price.

10. A modular photovoltaic carport microgrid system based on AI-driven photovoltaic-storage synergy, as described in claim 8, characterized in that: Operational constraints include: power balance constraints, energy storage operation constraints, charging load constraints, and grid interaction constraints. The power balance constraint requires that at any given time, the absolute value of the deviation between the predicted photovoltaic power generation, the energy storage discharge power, and the grid interaction power (which constitute the input power) and the output power (which constitutes the output power of the mobile charging terminal, the energy storage charging power, and the grid interaction power) does not exceed the preset system loss safety balance margin.

Images