Photovoltaic energy storage system and energy storage method thereof

Abstract

The application provides a photovoltaic energy storage system and an energy storage method thereof, comprising a multi-modal perception module, a battery health optimization unit, a power grid collaborative controller, an intelligent controller and an edge-cloud interaction module; the multi-modal perception module is used for collecting meteorological data, user power load data and historical behavior labels, and outputting a meteorological-load joint prediction curve; the battery health optimization unit is used for monitoring the charge-discharge depth, internal resistance change and temperature coefficient of the energy storage battery pack in real time, and constructing a health degradation model, relating to the new energy technology field, wherein the problems of insufficient prediction accuracy, battery management surface layer and poor collaboration with the power grid existing in the existing photovoltaic energy storage management system are solved, through the innovative system architecture and method, the dynamic distribution of photovoltaic power and the whole life cycle management of the energy storage battery are realized, and the system performance and economic benefits are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy technology, in particular to a photovoltaic energy storage system and an energy storage method thereof. BACKGROUND

[0002] With the growing global demand for clean energy, photovoltaic energy storage systems as an important way of renewable energy utilization have been widely used in the energy field. However, the current photovoltaic energy storage management system has many problems to be solved, which seriously limits its performance and economic benefits.

[0003] Firstly, in terms of prediction accuracy, the existing photovoltaic energy storage management system mostly relies on single-dimensional real-time data, such as simply relying on meteorological data or electricity load data to adjust the strategy. This approach fails to fully consider the periodicity of user electricity behavior, resulting in misalignment between photovoltaic power distribution and actual demand. For example, in some industrial user scenarios, the electricity usage patterns on weekdays and holidays are very different, but the existing system cannot accurately capture this regularity, resulting in over-storage of photovoltaic power during holidays or insufficient storage during peak electricity usage on weekdays, causing energy waste or supply shortages and reducing energy utilization efficiency.

[0004] Secondly, in terms of battery management, the current system often only stays at the level of simple monitoring of battery status, without establishing a dynamic correlation between battery health and charging and discharging strategies. This makes it difficult to avoid damaging operations on the battery during actual operation. For example, when the battery is in a high-temperature environment, if the charging and discharging rate is not adjusted accordingly, it will accelerate the aging of the battery, shorten the service life of the battery, and increase the operation and maintenance cost of the system.

[0005] Furthermore, in terms of interaction with the power grid, the existing photovoltaic energy storage system mostly only passively responds to the instructions of the power grid and cannot actively participate in demand response to improve the economy of the system. For example, during peak load periods of the power grid, the system cannot actively adjust the electricity usage strategy according to real-time electricity prices and its own energy storage conditions to make reasonable electricity purchases or sales, thus failing to fully tap the economic potential of the system in grid interaction.

[0006] In summary, the existing photovoltaic energy storage management system has problems such as insufficient prediction accuracy, superficial battery management, and poor coordination with the power grid, resulting in low energy utilization rate, short battery life, and insufficient release of economic potential. Therefore, there is an urgent need for a new photovoltaic energy storage system and an energy storage method that can overcome these problems to improve the utilization efficiency of photovoltaic energy, prolong the service life of energy storage batteries, and improve the economic benefits of the system in grid interaction. SUMMARY

[0007] In order to overcome the existing problems, the embodiment of the application provides a photovoltaic energy storage system and an energy storage method thereof, which solves the problems of insufficient prediction accuracy, battery management surface layering and poor coordination with the power grid existing in the existing photovoltaic energy storage management system, realizes dynamic allocation of photovoltaic electric energy and full life cycle management of energy storage batteries through innovative system architecture and method, and significantly improves system performance and economic benefits.

[0008] The technical solution adopted by the embodiment of the application to solve the technical problems is:

[0009] A photovoltaic energy storage system and an energy storage method thereof, comprising a multi-modal perception module, a battery health optimization unit, a grid coordination controller, an intelligent controller and an edge-cloud interaction module;

[0010] The multi-modal perception module is used to collect meteorological data, user electricity load data and historical behavior labels, and output a meteorological-load joint prediction curve;

[0011] Among them, more accurate prediction is realized by fusing meteorological data, user electricity load data and historical behavior labels, the meteorological perception submodule collects data such as solar radiation intensity, temperature, and precipitation probability, the prediction period covers 1 hour-7 days, the influence of weather change on photovoltaic power generation can be fully grasped, the load perception submodule collects real-time electricity power, cumulative electricity consumption and user behavior labels, the prediction period is 15 minutes-24 hours, the periodicity of user electricity behavior can be accurately captured, the multi-modal perception module integrates these data, and generates a meteorological-load joint prediction curve through a long short-term memory (LSTM) algorithm, which provides a more accurate basis for subsequent electric energy distribution, and effectively solves the problem of misalignment between photovoltaic electric energy distribution and actual demand;

[0012] The battery health optimization unit is used to monitor the charge-discharge depth, internal resistance change and temperature coefficient of the energy storage battery pack in real time, and construct a health degradation model;

[0013] The battery health optimization unit monitors the charge-discharge depth, internal resistance change and temperature coefficient in real time through sensors installed on each battery monomer or module of the energy storage battery pack, constructs a health degradation model based on the monitoring data, and the health degradation model is

[0014] = ,

[0015] Among them, is the battery health, is the charge-discharge depth, is the internal resistance change, is the temperature coefficient, is a functional relationship determined based on the physical and chemical principles of the battery and experimental data;

[0016] The battery health optimization unit not only monitors the charge and discharge depth, internal resistance change and temperature coefficient of the energy storage battery pack in real time, but also constructs a health degradation model. The model can automatically adjust the charge and discharge strategy according to the different states of the battery health (SOH). When the energy storage battery pack health (SOH) is less than 80%, the charge and discharge depth is automatically limited to within 50% to avoid further damage to the battery caused by excessive charge and discharge. When SOH is greater than or equal to 80%, the charge and discharge rate is dynamically adjusted according to the temperature coefficient. The charge rate is reduced by 10% for every 5℃ increase in temperature, effectively extending the service life of the battery. The closed-loop optimization management of the battery health is realized, and the defects of the existing battery management which only stay at the state monitoring level are overcome.

[0017] The grid coordination controller is used to receive demand response signals and real-time price information of the external grid, and generate a grid coordination strategy.

[0018] The grid coordination controller can receive demand response signals such as peak-valley electricity price, peak shaving instructions and real-time electricity price information of the external grid, generate a grid coordination strategy, and actively interact with the grid. The photovoltaic energy storage system can reasonably arrange the interaction period and power with the grid according to the demand and real-time price of the grid, and actively participate in the demand response of the grid to improve the economy of the system.

[0019] The intelligent controller is built-in long short-term memory network (LSTM) algorithm, based on the prediction data of the multi-modal perception module, the degradation model of the battery health optimization unit and the strategy of the grid coordination controller, to generate a predictive photovoltaic power distribution scheme.

[0020] The intelligent controller is built-in long short-term memory network (LSTM) algorithm, based on the prediction data of the multi-modal perception module, the degradation model of the battery health optimization unit and the strategy of the grid coordination controller, to generate a predictive photovoltaic power distribution scheme. The scheme includes real-time distribution ratio of photovoltaic component power generation, charge and discharge rate and depth limit of energy storage battery pack, and key parameters of interaction period and power with the grid, realizing dynamic and accurate distribution of photovoltaic power.

[0021] The edge-cloud interaction module is used to realize the cooperation of local edge computing and cloud big data analysis, and supports model iteration update.

[0022] The edge-cloud interaction module uploads the operation data of the photovoltaic energy storage system to the cloud, including photovoltaic component power generation, energy storage battery pack charge and discharge state, and interaction data with the grid.

[0023] The edge-cloud interaction module realizes the cooperation of local edge computing and cloud big data analysis. After the operation data is uploaded to the cloud, the powerful computing power of the cloud can be used for big data analysis, which is used for model iteration and update, further improving the performance and adaptability of the system.

[0024] The multi-modal perception module includes a weather perception sub-module and a load perception sub-module.

[0025] The weather perception sub-module collects solar radiation intensity, temperature, and precipitation probability data.

[0026] The weather perception sub-module collects weather data through weather monitoring equipment installed around the photovoltaic power station and transmits the data to the data processing unit of the multi-modal perception module through wired or wireless communication.

[0027] The load perception sub-module collects real-time power consumption, cumulative power consumption, and user behavior tags.

[0028] The load perception sub-module collects power consumption data through smart meters and data acquisition devices installed at user-side distribution boxes or meters, and obtains user behavior tags through user-side devices. The data is transmitted to the data processing unit of the multi-modal perception module.

[0029] Preferably, the attenuation model of the battery health optimization unit satisfies:

[0030] When the state of health (SOH) of the energy storage battery pack is less than 80%, the charging and discharging depth is automatically limited to within 50%.

[0031] When SOH is greater than or equal to 80%, the charging and discharging rate is dynamically adjusted according to the temperature coefficient. For every 5℃ increase in temperature, the charging rate decreases by 10%.

[0032] Preferably, the charging rate adjustment formula of the battery health optimization unit is:

[0033] ,

[0034] wherein, is the adjusted charging rate, is the unadjusted charging rate, is the increase in battery temperature relative to the initial temperature. When the state of health (SOH) of the energy storage battery pack is greater than or equal to 80%, the charging rate is dynamically adjusted according to the temperature coefficient based on this formula.

[0035] Preferably, the data processing unit of the multi-modal perception module generates a joint prediction curve using the following steps:

[0036] The collected meteorological data and user electricity load data are preprocessed, and the preprocessed data is input into a trained LSTM model, and the LSTM model outputs a meteorological-load joint prediction curve, and the training process uses historical meteorological data, user electricity load data and corresponding actual power generation and electricity data as a training set.

[0037] Preferably, the grid coordination controller is connected to an external grid dispatching system through a communication interface to receive demand response signals, and obtains real-time electricity price information through a power market data interface.

[0038] The photovoltaic energy storage method comprises the following steps:

[0039] Step one, the multi-modal perception module collects meteorological data, user electricity load data and historical behavior labels, and generates a meteorological-load joint prediction curve through an LSTM algorithm;

[0040] Step two, the battery health optimization unit monitors the operating parameters of the energy storage battery pack in real time, and outputs a health degree (SOH) value and a damage operation warning;

[0041] Step three, the grid coordination controller receives the demand response signal of the external grid, and generates a grid interaction priority;

[0042] Step four, the intelligent controller inputs the prediction curve of step one, the health data of step two and the grid priority of step three into the LSTM model to generate an energy distribution scheme;

[0043] Step five, the bidirectional inverter executes the operation according to the distribution scheme of the intelligent controller, and the edge-cloud interaction module uploads the operation data to the cloud for model iterative optimization.

[0044] The advantages of the embodiments of the present application are:

[0045] The multi-modal perception module constructs a meteorological-load-battery attenuation three-dimensional prediction model, so that the photovoltaic energy distribution can more accurately match the actual demand of users. Compared with the traditional system which only relies on single-dimensional data, the waste and supply shortage of photovoltaic energy are effectively reduced. Combined with the battery health closed-loop optimization mechanism and the grid coordination response architecture, the long short-term memory network algorithm is used to fuse multi-dimensional data to generate a predictive charging and discharging strategy, and a real-time feedback mechanism of health degree and strategy is established. At the same time, active cooperative scheduling with the grid is supported, realizing dynamic distribution of photovoltaic energy and full life cycle management of energy storage batteries.

[0046] The battery health degree optimization unit constructs a health degree attenuation model and a corresponding charge-discharge strategy adjustment mechanism, which can intelligently adjust the charge-discharge depth and rate according to the real-time health state and operating parameters of the battery, avoid damaging operation of the battery, significantly prolong the service life of the energy storage battery, and the grid cooperative controller enables the photovoltaic energy storage system to actively participate in grid demand response, acquires the demand response signal and electricity price information of the grid in real time, and the system can purchase or sell electricity at appropriate times to maximize economic benefits.

[0047] The predictive power distribution scheme generated by the intelligent controller based on multi-dimensional data and the model iterative update realized by the edge-cloud interaction module enable the entire photovoltaic energy storage system to continuously optimize and adapt to different operating environments and user demands, which not only improves the overall performance of the system, but also provides strong support for the widespread application and development of photovoltaic energy storage technology, and has broad application prospects and promotional value. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 The figure is a schematic diagram of the photovoltaic energy storage system framework of the present application;

[0049] Figure 2 The figure is a schematic diagram of the photovoltaic energy storage method flow of the present application. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. In addition, for the convenience of description, the "up", "down", "left", "right" and the like in the drawings are consistent with the up, down, left and right of the drawings, and "first", "second" and the like in the following are for description and differentiation, and have no other special meanings.

[0051] The embodiments of the present application provide a photovoltaic energy storage system and an energy storage method, solve the problems in the prior art, construct a meteorological-load-battery attenuation three-dimensional prediction model through a multi-modal perception module, so that the photovoltaic power distribution can more accurately match the actual demand of users. Compared with the traditional system which only relies on single-dimensional data, the waste and supply shortage of photovoltaic power are effectively reduced, the battery health degree closed-loop optimization mechanism and the grid cooperative response architecture are combined, the multi-dimensional data is fused through the long short-term memory network algorithm, the predictive charge-discharge strategy is generated, and the real-time feedback mechanism of health degree and strategy is established, while supporting active cooperative scheduling with the grid, realizing dynamic distribution of photovoltaic power and full life cycle management of energy storage batteries.

[0052] The battery health optimization unit constructs a health degradation model and a corresponding charge-discharge strategy adjustment mechanism, which can intelligently adjust the charge-discharge depth and rate according to the real-time health state and operating parameters of the battery, avoid damaging operations on the battery, significantly prolong the service life of the energy storage battery, and enable the photovoltaic energy storage system to actively participate in grid demand response through the grid collaborative controller. By real-time acquisition of the demand response signal and electricity price information of the power grid, the system can purchase or sell electricity at appropriate times to maximize economic benefits.

[0053] The predictive power distribution scheme generated by the intelligent controller based on multi-dimensional data and the model iterative update realized by the edge-cloud interaction module enable the entire photovoltaic energy storage system to continuously optimize and adapt to different operating environments and user demands, which not only improves the overall performance of the system, but also provides strong support for the widespread application and development of photovoltaic energy storage technology, and has broad application prospects and promotional value.

[0054] The technical scheme in the embodiments of the present application is to solve the above problems, and the general idea is as follows:

[0055] Embodiment one

[0056] The present embodiment provides a photovoltaic energy storage system, as shown in Figure 1 The present embodiment provides a photovoltaic energy storage system, as shown in

[0057] The multi-modal perception module is used to collect meteorological data, user electricity load data and historical behavior labels, and output a meteorological-load joint prediction curve;

[0058] The data processing unit of the multi-modal perception module integrates the prediction data of the meteorological perception sub-module and the load perception sub-module, and inputs them into a long short-term memory (LSTM) algorithm model. The LSTM model is trained with a large amount of historical data, and can learn the complex relationship between meteorological factors and electricity load. Through deep analysis and processing of the input data, the LSTM model generates a "meteorological-load" joint prediction curve, which reflects the change trend of photovoltaic power generation and user electricity load in the future, and provides accurate prediction basis for subsequent power distribution;

[0059] The battery health optimization unit is used to monitor the charge-discharge depth, internal resistance change and temperature coefficient of the energy storage battery pack in real time, and construct a health degradation model;

[0060] The battery health optimization unit monitors the charge-discharge depth, internal resistance change and temperature coefficient of each battery monomer or module of the energy storage battery pack in real time through the sensors installed on each battery monomer or module, and constructs a health degradation model based on the monitoring data. The health degradation model is

[0061] = ,

[0062] wherein, SOH is the battery health degree, D is the charge and discharge depth, R is the internal resistance change, T is the temperature coefficient, is a function relationship determined based on the physical and chemical principles of the battery and experimental data;

[0063] The attenuation model of the battery health degree optimization unit satisfies:

[0064] When the energy storage battery pack health degree (SOH) < 80%, the charge and discharge depth is automatically limited to within 50%;

[0065] When SOH≥80%, the charge and discharge rate is dynamically adjusted according to the temperature coefficient, and the charging rate is reduced by 10% for every 5℃ increase in temperature.

[0066] Specific function form

[0067] Health degree attenuation model = The specific function form of the health degree attenuation model is based on the experimental data of lithium iron phosphate batteries, and the expression is:

[0068] = ,

[0069] wherein:

[0070] SOH0 is the initial health degree coefficient (value range 1.0±0.05);

[0071] D is the charge and discharge depth influence coefficient (0.02-0.05 / cycle);

[0072] R is the internal resistance change influence coefficient (0.01-0.03 / Ω);

[0073] T is the temperature coefficient influence coefficient (0.005-0.01 / ℃).

[0074] Parameter value range Experimental data source

[0075] Parameter Value range Source of experimental data 0~100% 500 cycles of charge and discharge experiments (25℃ environment) 1.0~2.5 times of the initial internal resistance Measured by AC impedance method (frequency 1kHz) -20℃~60℃ High and low temperature chamber simulation environment experiment (step 5℃)

[0076] The modeling method is as follows:

[0077] A multiple nonlinear regression combined with a gray prediction model is used:

[0078] First step: Obtain the measured value of SOH under different combinations of DOD, IR and TC through orthogonal experimental design.

[0079] Second step: Obtain the basic function by fitting using the least squares method.

[0080] Third step: Introduce Kalman filter algorithm to correct model error, ensure prediction error ≤3%.

[0081] The charging rate adjustment formula of the battery health optimization unit is:

[0082]

[0083] wherein, is the adjusted charging rate, is the charging rate before adjustment, is the temperature rise relative to the initial temperature, when the state of health (SOH) of the energy storage battery pack ≥80%, the charging rate is dynamically adjusted according to the temperature coefficient based on this formula;

[0084] Install sensors on each battery monomer or battery module of the energy storage battery pack for real-time monitoring of parameters such as charging and discharging depth, internal resistance change and temperature coefficient. These sensors will transmit the monitored data to the control chip of the battery health optimization unit through the internal communication bus. High-precision power monitoring chips are used to monitor the charging and discharging depth, the internal resistance change is measured over time to reflect the aging degree of the battery, and the temperature sensor is used to obtain the battery temperature in real time and calculate the temperature coefficient.

[0085] The control chip of the battery health optimization unit constructs a health degradation model based on the collected parameters. This model is based on the physical and chemical principles of the battery and a large amount of experimental data. It describes the relationship between battery health and charging and discharging depth, internal resistance change and temperature coefficient through mathematical formulas and algorithms. Through the fitting of battery aging experimental data under different temperature and charging and discharging depth conditions, a functional relationship between battery health and these parameters is established to accurately assess the degradation of battery health.

[0086] According to the constructed health degradation model, when the state of health (SOH) of the energy storage battery pack <80%, the control chip automatically sends instructions to limit the charging and discharging depth to within 50%. The specific implementation is to adjust the size and direction of the charging and discharging current by controlling the charging and discharging control circuit in the battery management system (BMS) to ensure that the battery operates within a safe charging and discharging depth range. When SOH ≥80%, the control chip dynamically adjusts the charging and discharging rate according to the temperature coefficient. When the temperature rises by 5℃, the control chip reduces the charging rate by 10% through the BMS, and the adjustment of the charging rate is realized through the power regulation module in the charging circuit.

[0087] ​The power grid coordination controller is used for receiving demand response signals and real-time electricity price information of an external power grid, and generating a power grid coordination strategy;

[0088] The power grid coordination controller is connected with a dispatching system of the external power grid through a communication interface, and receives demand response signals of the power grid in real time, such as peak-valley electricity price information and peak regulation instructions, and obtains real-time electricity price information through a power market data interface. These signals and information are transmitted to a central processing unit of the power grid coordination controller for processing.

[0089] The central processing unit of the power grid coordination controller generates a power grid coordination strategy by using an optimization algorithm according to the received demand response signals and real-time electricity price information. In a peak-valley electricity price mode, when the real-time electricity price is in a low valley period and the power grid has a peak regulation demand, the central processing unit calculates reasonable power purchase time periods and power quantities to store low-price electricity. When the real-time electricity price is in a peak period and the power grid needs load regulation, the central processing unit calculates appropriate power selling time periods and power quantities to maximize economic benefits. In a peak regulation instruction mode, according to the peak regulation requirements of the power grid, the specific strategy of the energy storage system participating in peak regulation is determined, such as releasing electricity during a high peak load period of the power grid and storing electricity during a low valley load period.

[0090] The intelligent controller is built-in with a long short-term memory (LSTM) algorithm, which generates a predictive photovoltaic electricity distribution scheme based on prediction data of a multi-modal perception module, a degradation model of a battery health optimization unit, and a strategy of the power grid coordination controller.

[0091] The intelligent controller obtains a “weather-load” joint prediction curve from the multi-modal perception module, obtains a battery state of health (SOH) value and a damage operation warning information from the battery health optimization unit, and obtains power grid interaction priority information from the power grid coordination controller. The intelligent controller integrates these data and inputs them into the built-in long short-term memory (LSTM) model.

[0092] Based on the input multi-dimensional data, the LSTM model uses a trained algorithm and weight to generate an electricity distribution scheme containing the following parameters: a real-time distribution ratio of photovoltaic component power generation (preferentially supplied to load / storage / grid), a charge and discharge rate and depth limit of the energy storage battery pack, and an interaction time period and power with the power grid. When it is predicted that the electricity load will increase and the battery health is good, the LSTM model calculates to preferentially supply photovoltaic component power generation to the load and appropriately increase the discharge rate of the energy storage battery pack. When it is predicted that the photovoltaic power generation is excessive and the electricity price is favorable, the LSTM model determines the time period and power for selling the excess power to the grid.

[0093] Priority quantification index:

[0094] The power grid coordination controller calculates the interaction priority P by the following formula:

[0095] + + ,

[0096] wherein:

[0097] : real-time electricity price factor (0-1, the higher the electricity price, the greater the value);

[0098] : grid demand factor (0-1, the higher the peak shaving instruction intensity, the greater the value);

[0099] : battery remaining capacity factor (0-1, the lower the capacity, the smaller the value);

[0100] weight = 0.4, = 0.4, = 0.2, determined based on a multi-objective optimization algorithm.

[0101] Strategy selection logic:

[0102] Priority range Collaboration strategy Application scenario example Sell electricity to the grid preferentially, and the depth of discharge is ≤80% When the electricity price is at the peak and the grid needs load reduction Maintain battery capacity balance, and do not interact with the grid When the electricity price is stable and the grid has no peak shaving demand Buy electricity from the grid, and the depth of charge is ≤70% When the electricity price is at the trough and the battery capacity is insufficient

[0103] Data interface and execution mechanism

[0104] The grid coordination controller communicates with the grid dispatching system through the DL / T645 protocol, updates the demand response signal once every second, dynamically adjusts the power flow direction of the bidirectional inverter based on priority, and the response delay is ≤100 ms.

[0105] The edge-cloud interaction module is used to realize the cooperation of local edge computing and cloud big data analysis, and supports model iteration update.

[0106] The edge-cloud interaction module uploads the operation data of the photovoltaic energy storage system to the cloud, including the photovoltaic component power generation, the charge and discharge state of the energy storage battery pack, and the interaction data with the grid.

[0107] The bidirectional inverter receives the power distribution scheme generated by the intelligent controller, and performs specific operations according to the instructions in the scheme. When the scheme requires storing the photovoltaic component power generation to the energy storage battery pack, the bidirectional inverter converts the direct current output by the photovoltaic component into direct current suitable for battery charging, and controls the charging current and voltage; when it is required to supply power to the load or sell electricity to the grid, the bidirectional inverter converts the direct current of the energy storage battery pack into alternating current, and adjusts the output power and frequency according to the demand.

[0108] The edge-cloud interaction module uploads operation data of the photovoltaic energy storage system, such as photovoltaic component power generation, energy storage battery charging and discharging state, and grid interaction data, to the cloud server through the network communication module. The cloud server processes and analyzes these data using big data analysis technology to identify potential problems and optimization space in system operation. Through analysis of a large amount of operation data, the use of electricity mode change law in certain regions in a specific season is found, and the prediction model of the multi-modal perception module is optimized. The optimized model is downloaded to the local edge computing device through the cloud server to realize iterative updating of the model and further improve the performance and adaptability of the system.

[0109] The multi-modal perception module includes a meteorological perception sub-module and a load perception sub-module.

[0110] The meteorological perception sub-module collects solar radiation intensity, temperature, and precipitation probability data.

[0111] Meteorological monitoring equipment such as solar radiation sensors, temperature sensors, and precipitation sensors is installed at appropriate locations around the photovoltaic power station to collect real-time meteorological data such as solar radiation intensity, temperature, and precipitation probability. These data are transmitted to the data processing unit of the multi-modal perception module through wired or wireless communication. The data processing unit uses professional meteorological prediction algorithms and models to predict the weather for the next 1 hour to 7 days based on the collected meteorological data.

[0112] The meteorological perception sub-module collects meteorological data through meteorological monitoring equipment installed around the photovoltaic power station and transmits the data to the data processing unit of the multi-modal perception module through wired or wireless communication.

[0113] The load perception sub-module collects real-time power consumption, cumulative power consumption, and user behavior labels.

[0114] Smart meters and data acquisition devices are installed at the distribution box or meter at the user end to collect real-time power consumption and cumulative power consumption data. At the same time, user behavior labels such as weekday / holiday information are obtained through user end devices. These data are also transmitted to the data processing unit of the multi-modal perception module. The data processing unit uses data mining and machine learning algorithms to analyze the collected load data and behavior labels to predict the power load for the next 15 minutes to 24 hours. By analyzing historical power consumption data, different power consumption patterns on weekdays and holidays are identified, and a power load prediction model based on user behavior is established for rolling prediction combined with real-time data.

[0115] The load perception sub-module collects power consumption data through smart meters and data acquisition devices installed at the distribution box or meter at the user end, and obtains user behavior labels through user end devices. The data are transmitted to the data processing unit of the multi-modal perception module.

[0116] The data processing unit of the multi-modal perception module generates a joint prediction curve using the following steps:

[0117] The collected meteorological data and user electricity load data are preprocessed, and the preprocessed data are input into the trained LSTM model. The LSTM model outputs a meteorological-load joint prediction curve. The training process uses historical meteorological data, user electricity load data, and corresponding actual power generation and electricity consumption data as the training set.

[0118] The grid coordination controller is connected to the external grid dispatching system through a communication interface to receive demand response signals, and obtains real-time electricity price information through a power market data interface.

[0119] Embodiment Two

[0120] This embodiment gives a kind of photovoltaic energy storage method, as shown in Figure 2 The photovoltaic energy storage method includes the following steps:

[0121] Step one, the multi-modal perception module collects meteorological data, user electricity load data and historical behavior labels, and generates a meteorological-load joint prediction curve through an LSTM algorithm;

[0122] The multi-modal perception module fuses meteorological data, user electricity load data and historical behavior labels. The meteorological data covers solar radiation intensity, temperature, and precipitation probability, and can reflect the influence of weather on photovoltaic power generation. The user electricity load data combined with real-time power, cumulative power and behavior labels can grasp the user electricity usage rules. Multi-dimensional data fusion makes the prediction more accurate, solves the problem of misalignment between electricity distribution and actual demand, and uses a long short-term memory network (LSTM) algorithm to process multi-modal data to generate a meteorological-load joint prediction curve. The LSTM algorithm is good at processing long-term dependencies in time series data, and can effectively mine the complex relationship between meteorological data and electricity load, improving the prediction accuracy;

[0123] Step two, the battery health optimization unit monitors the operating parameters of the energy storage battery pack in real time, and outputs the health value (SOH) and damage operation warning;

[0124] The battery health optimization unit monitors the charge and discharge depth, internal resistance change and temperature coefficient of the energy storage battery pack in real time. These parameters comprehensively reflect the battery operating state, provide a basis for accurate assessment of battery health, and construct a health degradation model based on the monitoring parameters to quantify the relationship between battery health and various parameters. Through this model, the system can automatically adjust the charge and discharge strategy according to the battery health, such as limiting the charge and discharge depth when the health value (SOH) is less than 80%, and adjusting the charge and discharge rate according to the temperature coefficient when the SOH is greater than or equal to 80%, to realize closed-loop optimization of battery health and prolong the battery life.

[0125] Step three, the grid coordination controller receives the demand response signal of the external grid, and generates the grid interaction priority;

[0126] The grid coordination controller not only receives the demand response signal of the external grid, peak-valley electricity price, and peak regulation instruction, but also obtains real-time electricity price information. This two-way information interaction enables the system to comprehensively understand the grid state and market price signal. Based on the received information, the grid coordination controller generates the grid coordination strategy, so that the photovoltaic energy storage system changes from passive response to active participation in grid demand response, optimizes the interaction period and power with the grid, and improves the economic efficiency of the system.

[0127] Step four, the intelligent controller inputs the prediction curve of step one, the health degree data of step two, and the grid priority of step three into the LSTM model to generate an electric energy distribution scheme.

[0128] The intelligent controller fuses the prediction data of the multi-modal perception module, the attenuation model of the battery health degree optimization unit, and the strategy of the grid coordination controller, and generates a predictive photovoltaic electric energy distribution scheme through the built-in LSTM algorithm. The scheme considers multiple factors and realizes dynamic distribution of photovoltaic electric energy.

[0129] Step five, the bidirectional inverter executes the operation according to the distribution scheme of the intelligent controller, and the edge-cloud interaction module uploads the running data to the cloud for model iterative optimization.

[0130] The edge-cloud interaction module realizes the cooperation of local edge computing and cloud big data analysis, uploads the running data to the cloud for analysis, supports model iterative update, and enables the system to adapt to different operating environments and user demands, and continuously improves the performance.

[0131] Finally, it should be noted that: obviously, the above embodiments are only examples for clearly illustrating the present application, and are not limitations of the embodiments. For those skilled in the art, other different forms of changes or variations can be made on the basis of the above description. Here, it is not necessary and impossible to enumerate all the embodiments. The obvious changes or variations derived therefrom are still within the protection scope of the present application.

Claims

1. A photovoltaic energy storage system, characterized by, The multi-modal perception module, the battery health optimization unit, the grid coordination controller, the intelligent controller, and the edge-cloud interaction module are included. The multi-modal perception module is used for collecting meteorological data, user power load data, and historical behavior labels, and outputting a meteorological-load joint prediction curve. The multi-modal perception module includes a meteorological perception submodule and a load perception submodule. The meteorological perception submodule collects solar radiation intensity, temperature, and precipitation probability data. The load perception submodule collects real-time power consumption, cumulative power consumption, and user behavior labels. The load perception submodule collects power consumption data through intelligent meters installed at user-side distribution boxes or power meters, and acquires user behavior labels through user-side devices, and transmits the data to a data processing unit of the multi-modal perception module. The data processing unit of the multi-modal perception module generates a joint prediction curve by the following steps: The collected meteorological data and user power load data are preprocessed, and the preprocessed data is input into a trained LSTM model, which outputs a meteorological-load joint prediction curve. The battery health optimization unit is used for real-time monitoring of the charge-discharge depth, internal resistance change, and temperature coefficient of the energy storage battery pack, and constructing a health degradation model. The battery health optimization unit monitors the charge-discharge depth, internal resistance change, and temperature coefficient of each battery cell or module in the energy storage battery pack through sensors installed thereon, and constructs a health degradation model based on the monitoring data. = ， wherein, is the battery health, is the charge and discharge depth, is the internal resistance change, is the temperature coefficient, is a function relationship determined based on the physical and chemical principles of the battery and experimental data; The degradation model of the battery health optimization unit satisfies: When the health of the energy storage battery pack (SOH) is less than 80%, the charge-discharge depth is automatically limited to within 50%; When SOH is greater than or equal to 80%, the charge-discharge rate is dynamically adjusted according to the temperature coefficient, and the charging rate is reduced by 10% for every 5℃ increase in temperature. The grid coordination controller is used for receiving demand response signals and real-time price information of an external power grid, and generating a grid coordination strategy. The intelligent controller is built-in with a long short-term memory (LSTM) algorithm, which generates a predictive photovoltaic power distribution scheme based on the prediction data of the multi-modal perception module, the degradation model of the battery health optimization unit, and the strategy of the grid coordination controller. The edge-cloud interaction module is used for realizing the cooperation of local edge computing and cloud big data analysis, and supporting model iteration and update.

2. A photovoltaic energy storage system according to claim 1, wherein, The charge rate adjustment formula of the battery health optimization unit is: ， wherein, is the adjusted charging rate, is the unadjusted charging rate, is the increase in battery temperature relative to the initial temperature, when the state of health (SOH) of the energy storage battery pack is ≥ 80%, the charging rate is dynamically adjusted according to the temperature coefficient based on this formula.

3. A photovoltaic energy storage system according to claim 1, wherein, The edge-cloud interaction module uploads the operation data of the photovoltaic energy storage system to the cloud, including the power generation of the photovoltaic module, the charge-discharge state of the energy storage battery pack, and the interaction data with the power grid.

4. A photovoltaic energy storage system according to claim 1, wherein, The meteorological perception submodule collects meteorological data through meteorological monitoring devices installed around the photovoltaic power station, and transmits the data to the data processing unit of the multi-modal perception module through wired or wireless communication.

5. A photovoltaic energy storage system according to claim 1, wherein, The power grid coordination controller is connected with an external power grid dispatching system through a communication interface to receive a demand response signal and acquires real-time electricity price information through a power market data interface.

6. A photovoltaic energy storage method based on the photovoltaic energy storage system of any one of claims 1-5, characterized in that, The method comprises the following steps: Step one, the multi-modal perception module collects meteorological data, user electricity load data and historical behavior labels, and generates a meteorological-load joint prediction curve through an LSTM algorithm; Step two, the battery health optimization unit monitors the operating parameters of the energy storage battery pack in real time, and outputs a health degree (SOH) value and a damage operation warning; Step three, the power grid coordination controller receives a demand response signal of an external power grid, and generates a power grid interaction priority; Step four, the intelligent controller inputs the prediction curve of step one, the health degree data of step two and the power grid priority of step three into an LSTM model to generate an electric energy distribution scheme; Step five, the bidirectional inverter executes the operation according to the distribution scheme of the intelligent controller, and the edge-cloud interaction module uploads the operation data to the cloud end for model iterative optimization.

Images