Multifunctional photovoltaic energy storage safety device

Abstract

The invention relates to the technical field of photovoltaic energy storage, and discloses a multifunctional photovoltaic energy storage safety device which comprises a plurality of data acquisition nodes, a first transmission module, a second transmission module, a master control single-chip microcomputer, a photovoltaic output power prediction module, a cloud server, a database and a client module. The plurality of data acquisition nodes are connected with the first transmission module, the main control single-chip microcomputer is respectively connected with the first transmission module, the second transmission module and the photovoltaic output power prediction module, and the cloud server is respectively connected with the database, the second transmission module and the client module. According to the multifunctional photovoltaic energy storage safety device provided by the invention, photovoltaic output power prediction is carried out according to sensor detection data of multiple monitoring nodes so as to better optimize a battery charging and discharging strategy, meanwhile, battery health state evaluation is carried out on the battery, and the service life and safety of the battery are determined; and multifunctional safety monitoring of the photovoltaic energy storage power station is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic energy storage, and particularly relates to a multifunctional photovoltaic energy storage safety device. Background Art

[0002] Photovoltaic power generation is a technology that directly converts light energy into electrical energy by using the photovoltaic effect at the semiconductor interface. It mainly consists of three major parts: solar panels (modules), controllers, and inverters. The main components are composed of electronic components. Solar cells can be encapsulated and protected after being connected in series to form large-area solar cell modules. Together with components such as power controllers, a photovoltaic power generation device is formed. The electrical energy generated by photovoltaic power generation needs to be stored using energy storage devices.

[0003] During the use of photovoltaic power generation energy storage devices, it is necessary to conduct safety monitoring on them to reduce accidents. When conducting safety monitoring on them, a safety monitoring system needs to be used. In the prior art, the safety monitoring effect on photovoltaic power generation energy storage devices is poor, and it is impossible to conduct multifunctional safety monitoring on photovoltaic energy storage power stations, resulting in frequent failures of photovoltaic power generation energy storage devices. If the operation of a photovoltaic energy storage power station cannot be safely monitored through an effective management system, there are significant safety hazards, thus affecting the promotion effect and use safety of photovoltaic energy storage charging stations during actual operation. Summary of the Invention

[0004] The present invention provides a multifunctional photovoltaic energy storage safety device, which predicts the photovoltaic output power based on the sensor detection data of multiple monitoring nodes to better optimize the battery charge and discharge strategy. At the same time, it evaluates the battery health state of the battery to determine the life and safety of the battery, and realizes multifunctional safety monitoring of the photovoltaic energy storage power station.

[0005] The present invention provides a multifunctional photovoltaic energy storage safety device, including multiple data acquisition nodes, a first transmission module, a second transmission module, a main control single-chip microcomputer, a photovoltaic output power prediction module, a cloud server, a database, and a client module. The multiple data acquisition nodes are connected to the first transmission module, and the main control single-chip microcomputer is respectively connected to the first transmission module, the second transmission module, and the photovoltaic output power prediction module. The cloud server is respectively connected to the database, the second transmission module, and the client module;

[0006] The multiple data acquisition nodes are composed of various sensors and STM32 microcontrollers to collect the working environment parameter data and equipment operation status information that need to be monitored in the photovoltaic energy storage power station by using the sensors and transmit them to the main control single-chip microcomputer;

[0007] The transmission module is used to transmit the working environment parameter data and equipment operation status information collected by multiple data collection nodes to the main control microcontroller for photovoltaic power prediction;

[0008] The cloud server receives photovoltaic power prediction results and monitoring data of the photovoltaic energy storage power station battery management system to perform charging and discharging optimization and battery health assessment;

[0009] The database uses MySQL database for data management and storage, and the client module is used for data display and realization of human-computer interaction function.

[0010] Further, each of the plurality of data acquisition nodes includes a light intensity sensor, a component temperature sensor, an ambient temperature sensor, a photovoltaic panel temperature sensor, and an STM32 microcontroller;

[0011] The ambient temperature sensor uses a sensor that can measure temperature and humidity at the same time. The light intensity sensor senses the intensity changes of light in the environment and converts it into an electrical signal for output. The light intensity sensor, component temperature sensor, ambient temperature sensor, and photovoltaic panel temperature sensor are all connected to the STM32 microcontroller via RS485 and UART.

[0012] Furthermore, the first transmission module adopts multiple LoRa communication modules, each LoRa communication unit corresponds to a data acquisition node; the second transmission module adopts a NB-IOT communication module.

[0013] Furthermore, the photovoltaic output power prediction module uses the data collected by the light intensity sensor, the component temperature sensor, the ambient temperature sensor, and the photovoltaic panel temperature sensor to predict the photovoltaic output power. The specific steps are as follows:

[0014] S1. Clean the collected total solar irradiance, direct solar radiation, scattered solar radiation, module temperature, ambient temperature, humidity and photovoltaic power data, remove the full-day data containing invalid data, and normalize the data set using a linear function normalization method;

[0015] S2, split the data set into training set and test set;

[0016] S3. Build a CNN-BiLSTM model and initialize the BiLSTM model parameters. The model uses the CNN algorithm to extract features from the input data and uses the BiLSTM layer to model. Then, the output of the BiLSTM layer is used as the input of the fully connected layer to output the prediction results. A Dropout layer is added after each BiLSTM hidden layer to prevent overfitting.

[0017] S4. Introduce the SCSSA algorithm to optimize the regularization coefficient, dimension, and learning rate of the BiLSTM;

[0018] S5. Use the SCSSA-CNN-BiLSTM model to train the training set samples, and input the test set into the trained SCSSA-CNN-BiLSTM model for testing;

[0019] S6. Perform anti-normalization processing on the prediction results, calculate the model evaluation metrics, and obtain the trained SCSSA-CNN-BiLSTM model;

[0020] S7. Use the trained SCSSA-CNN-BiLSTM model to predict the photovoltaic output power.

[0021] Furthermore, in step S4, the SCSSA optimization algorithm includes:

[0022] S401. Set the population size n, maximum number of iterations, ratio of discoverers to vigilantists, vigilance threshold, and safety threshold;

[0023] S402. Randomly initialize the positions of n sparrows in the optimization range as the initial population positions, generate the refraction reverse population, merge the initial population and the refraction reverse population, sort them according to the rise and fall of the fitness value, and select the top n sparrow individuals with the fitness value as the initial population to initialize the population;

[0024] S403. Calculate the fitness value of each sparrow and sort them, determine the current optimal and worst fitness individuals, and update the positions of the discoverers, followers, and vigilantists;

[0025] S404. Determine whether the current number of iterations reaches the end condition;

[0026] S405. If the end condition is not reached, return to step S403 for loop;

[0027] S406. If the end condition is reached, output the optimal fitness value and the best position.

[0028] Furthermore, the cloud server includes a charge and discharge optimization module and a battery health assessment module; the charge and discharge optimization module formulates a charge and discharge strategy based on the prediction result of the photovoltaic output power, combined with the energy storage state and grid demand; the battery health assessment module conducts battery health assessment based on the battery charge and discharge data monitored by the battery management system.

[0029] Furthermore, the battery health assessment module evaluates the health status of a single battery in the battery array, that is, calculates the ratio of the current rated capacity to the initial rated capacity of the energy storage unit:

[0030]

[0031] Among them, C t is the current capacity of the energy storage unit, and C0 is the initial capacity; the number of battery cycles is determined according to the battery charge and discharge data monitored by the battery management system, and the following is calculated:

[0032]

[0033] Among them, R EOL is the internal resistance value at the end of the battery life, R a is the current internal resistance value, R r is the specified internal resistance value when the battery leaves the factory, N rem is the current number of cycles, N tol is the total number of cycles;

[0034] Finally, the battery health state is calculated as: SOH = A×S OH1 +B×S OH2 +C×S OH3 , where A + B + C = 1, and A, B, and C are all weight values.

[0035] Furthermore, the client module includes a computer, a tablet, and a mobile phone. Staff can view the working conditions and historical data of the photovoltaic system in real time through the computer browser, tablet, and mobile phone, realizing remote safety monitoring of the photovoltaic energy storage power station.

[0036] The beneficial effects of the present invention are as follows:

[0037] In the present invention, multiple data acquisition nodes use sensors to collect the working environment parameter data and equipment operation status information that need to be monitored in the photovoltaic energy storage power station and transmit them to the main control single-chip microcomputer for photovoltaic power prediction; the cloud server receives the photovoltaic power prediction results and the monitoring data of the battery management system of the photovoltaic energy storage power station for charge and discharge optimization and battery health assessment. The photovoltaic power prediction uses the SCSSA-CNN-BiLSTM model to improve the prediction accuracy, so as to better optimize the battery charge and discharge strategy, and then evaluate the battery health state based on the detection data of the battery management system to determine the life and safety of the battery, realizing multi-functional safety monitoring of the photovoltaic energy storage power station. Description of the Drawings

[0038] Figure 1 It is a schematic structural diagram of the photovoltaic energy storage safety device with multiple functions of the present invention.

[0039] The realization, functional characteristics, and advantages of the object of the present invention will be further described with reference to the embodiments and the drawings. Detailed Embodiments

[0040] It should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0041] As Figure 1 shown, the present invention provides a photovoltaic energy storage safety device with multiple functions, including a plurality of data acquisition nodes, a first transmission module, a second transmission module, a main control single-chip microcomputer, a photovoltaic output power prediction module, a cloud server, a database, and a client module. The plurality of data acquisition nodes are connected to the first transmission module, and the main control single-chip microcomputer is respectively connected to the first transmission module, the second transmission module, and the photovoltaic output power prediction module. The cloud server is respectively connected to the database, the second transmission module, and the client module.

[0042] The client module is used to obtain monitoring data from the cloud server, display the data, and implement a human-computer interaction function. The client module includes a computer, a tablet, and a mobile phone. Staff can view the working conditions and historical data of the photovoltaic system in real time through a computer browser, a tablet, and a mobile phone, realizing remote safety monitoring of the photovoltaic energy storage power station.

[0043] (1) Data acquisition node:

[0044] The plurality of data acquisition nodes are composed of various sensors and an STM32 microcontroller, so as to collect the working environment parameter data and equipment operation status information that need to be monitored in the photovoltaic energy storage power station by using the sensors and transmit them to the main control single-chip microcomputer.

[0045] The sensors can monitor relevant information such as the climate environment around the photovoltaic modules in real time. Each data acquisition node in the plurality of data acquisition nodes includes a light intensity sensor, a component temperature sensor, an environmental temperature sensor, a photovoltaic panel temperature sensor, and an STM32 microcontroller.

[0046] The environmental temperature sensor uses a sensor that can measure temperature and humidity simultaneously. The light intensity sensor senses the change in the intensity of light in the environment and converts it into an electrical signal for output. The light intensity sensor, the component temperature sensor, the environmental temperature sensor, and the photovoltaic panel temperature sensor are all connected to the STM32 microcontroller through RS485 and UART.

[0047] 1> Environmental temperature sensor

[0048] Environmental temperature and humidity are an important part of environmental parameters, and accurate measurement of temperature and humidity is particularly important. The present invention selects an AM2302 (also known as DHT22) type sensor that can measure temperature and humidity simultaneously. It accesses the main control through a single-wire communication method and only requires one I / O port to realize real-time acquisition of the environmental temperature and humidity, making the system integration simple and fast. This sensor is reliable and stable, and the signal transmission distance can reach more than 20 meters, enabling it to be used in relatively harsh application scenarios.

[0049] 2> Component temperature sensor

[0050] To meet the requirements of photovoltaic module temperature monitoring, the DS18B20 temperature sensor is selected. The sensor can communicate with the microcontroller through a single digital pin, and the acquisition of multi-point temperature only requires a single bus, which saves pins very much. To improve the anti-interference ability of the system and facilitate on-site temperature measurement of the temperature sensor in a harsh environment, the probe-type DS18B20 sensor is selected. Each temperature sensor has a unique number, and this number can be used to determine the specific sensor.

[0051] 3> Light intensity sensor

[0052] The illuminance sensor can sense the change in the intensity of light in the environment and convert it into an electrical signal for output. The YGC-TBQ type illuminance sensor is selected in the present invention, and its key technical parameters include: the measurement resolution is 1W / m 2 , and the measurement range is 0 to 2000W / m 2 , the measurement accuracy is ±3%, and the power supply voltage is DC 9~30V.

[0053] (2) Transmission module:

[0054] The transmission module is used to transmit the working environment parameter data and device operation status information collected by multiple data acquisition nodes to the main control single-chip microcomputer for photovoltaic power prediction.

[0055] The first transmission module uses multiple LoRa communication modules, and each LoRa communication unit corresponds to a data acquisition node; the second transmission module uses an NB-IOT communication module.

[0056] LoRa is a wireless technology based on spread-spectrum modulation technology. It has strong anti-interference ability, long transmission distance, low rate, and is non-connected, communicating through broadcasting. Currently, LoRa technology mainly includes first-generation LoRa (SX1276 / 78) and second-generation LoRa (SX1262 / 68) modulation technologies. The second-generation SX126X greatly reduces the receiving current, which is only 4.6 mA, nearly 50% less than that of SX127X. At the same time, SX126X also integrates a 22 dBm high-efficiency power amplifier inside, which expands the link budget and has a wider communication distance coverage. In addition, SX126X has also upgraded the spreading factor, and the maximum air rate has increased by nearly twice compared with SX127X. This invention uses the ATK-MW1268D module as the LoRa communication module, and this module uses the high-efficiency ISM (Industrial, Scientific, Medical) band SX1268 radio frequency chip. ATK-MW1268D includes 6 working modes, namely the general mode of transparent and directional data transmission, the wake-up mode that can wake up the receiving party working in the power-saving mode, the power-saving mode with the serial port receiving closed, the signal strength mode for querying the signal strength of both communication parties, the sleep mode that cannot send and receive data and the serial port is closed, and the relay mode for relaying wireless data. The working mode of the module can be switched by sending AT commands.

[0057] The advantages of NB-IoT are large connection, wide coverage, low power consumption, and low cost. Through NB-IoT devices, the data collected by the sensor module can be summarized and encoded and then uploaded to the cloud server. This invention selects the NB-IoT communication module that accesses the OneNET cloud platform as the China Mobile communication module M5311. M5311 meets the 3GPP Release13 standard, and the chip integrates multiple transmission protocols such as LwM2M, MQTT, HTTP, and TCP. The main functions of M5311 include power management, radio frequency, and some peripheral interfaces. The peripheral interfaces include power supply, power-on and power-off interfaces, SIM card interfaces, radio frequency interfaces, and serial ports. Its working voltage is 3 - 3.6V, and the power supply interface of STM32 can be used to supply power to the chip.

[0058] (3) Photovoltaic output power prediction module:

[0059] The photovoltaic output power prediction module uses the data collected by the light intensity sensor, component temperature sensor, ambient temperature sensor, and photovoltaic panel temperature sensor to predict the photovoltaic output power. The specific steps are as follows:

[0060] S1. The total solar irradiance (W / m 2 ), direct solar radiation (W / m 2 ), and diffuse solar radiation (W / m 2) Clean the data of component temperature (°C), ambient temperature (°C), humidity (%), and photovoltaic power (MW), eliminate the full-day data where invalid data is located, and normalize the data set using the linear function normalization method.

[0061] Accurate data is crucial for improving the accuracy of model prediction. Whether the data is recorded manually or automatically by software, data anomalies may occur. If these abnormal data are not processed, it will affect the accuracy of the algorithm. During the actual operation of a photovoltaic power station, due to reasons such as communication interruption, equipment failure, and low sensitivity of monitoring equipment, missing data recorded as "-99" may occur. This information will affect the overall prediction accuracy and cannot be used to build a model. Therefore, the present invention first performs data cleaning on the sampled data, eliminates the full-day data where the invalid data recorded as "-99" is located, and finally obtains the monitoring experiment data.

[0062] Using the normalized data as the input of the model can improve the accuracy of the prediction model, accelerate the convergence speed of the loss function, and prevent the problems of gradient disappearance and gradient explosion in deep learning. The present invention uses the linear function normalization method to process the environmental factor data and historical photovoltaic data in the original data set, normalizing each value to between [0,1]. The normalization calculation method is as follows:

[0063]

[0064] Among them, X is the value before normalization; Y is the value after normalization; X max represents the maximum value in the data; X min represents the minimum value.

[0065] S2. Split the data set into a training set and a test set.

[0066] S3. Build a CNN-BiLSTM model and initialize the BiLSTM model parameters. This model extracts features from the input data through the CNN algorithm, uses the BiLSTM layer for modeling, and then takes the output of the BiLSTM layer as the input of the fully connected layer to output the prediction result. Add a Dropout layer after each BiLSTM hidden layer to prevent overfitting.

[0067] BiLSTM is developed based on the Long Short Term Memory (LSTM) neural network. The LSTM internally includes an input layer, a hidden layer, a recurrent layer, and an output layer. To solve the problems of gradient vanishing, gradient explosion, and insufficient long-term memory during the training process of long sequences, the LSTM network adds memory cell states in the hidden layer and establishes three gating mechanisms, namely input, forget, and output, to filter information. LSTM can only rely on the sequence information at past moments to predict the output at the next moment. However, the photovoltaic power sequence has the characteristic of information correlation before and after. In order to reasonably utilize meteorological factors and historical photovoltaic power data and improve the photovoltaic power prediction ability, the model of the present invention adopts the BiLSTM neural network. In the BiLSTM neural network, the hidden state output by the forward LSTM at time t is denoted as h tf , and the output value of the hidden layer of the backward LSTM is h tb . Then the hidden state output by BiLSTM is:

[0068]

[0069] BiLSTM is formed by combining two LSTMs, forward and backward, in its structure. It can mine the features of input data in the time series and implement two LSTM trainings, forward and backward, fully considering past and future information, effectively reducing the useful information lost when extracting the time series features of data, further improving the globality and integrity of feature extraction, and enhancing the prediction accuracy for the current moment. Because the BiLSTM neural network can process time series data bidirectionally, its prediction accuracy is higher than that of the LSTM neural network.

[0070] CNN can effectively reduce the number of parameters during the algorithm training process, improve the model training speed, and better extract features from the input information. CNN includes a convolutional layer, a pooling layer, and a fully connected layer. The core layer of CNN is the convolutional layer, which efficiently extracts data features through convolutional calculations. The pooling layer will perform optimal feature selection through a certain strategy to improve the generalization ability. There are the average method and the maximum method. The present invention selects the maximum method to perform pooling operations on the output features after convolution to further retain important features. The fully connected layer is responsible for integrating and classifying the extracted features by learning weights.

[0071] That is, the CNN-BiLSTM model sequentially includes a convolutional layer, a pooling layer, a BiLSTM layer, a Dropout layer, a BiLSTM layer, a Dropout layer, a BiLSTM layer, a Dropout layer, and a fully connected layer.

[0072] S4. Introduce the SCSSA algorithm to optimize the regularization coefficient of BiLSTM, the dimension and learning rate of the neural network. Introduce the Convolution Neural Network (CNN) and the Sine-cosine and Cauchy mutation Sparrow Search Algorithm (SCSSA) to optimize the BiLSTM neural network and enhance the prediction ability of the model.

[0073] The Sparrow Search Algorithm (SSA) simulates the foraging process of sparrows and obtains the optimal solution of some parameters within a certain range by continuously updating the individual positions. SSA divides individuals into finders who are responsible for finding food in a specific area, followers who follow the finders to find the best food, and sentinels who are alert to whether there are predators around. By continuously updating the positions of the three, resources can be obtained during the search process. In this algorithm, finders with higher fitness values ​​are more likely to be selected as the basic population for the next round of search, thereby accumulating excellent solutions and improving search efficiency.

[0074] In order to improve the search ability of the prediction algorithm and its ability to jump out of the local optimum, the SCSSA algorithm is used to improve the performance of the sparrow search algorithm in global search for the best solution. The SCSSA algorithm can be used to optimize the learning rate, number of hidden layer nodes, and regularization parameters of the BiLSTM model.

[0075] 1> Sin and cosine mechanism

[0076] When the finder is in the local optimal position, the followers will gather there, causing the entire population to stagnate, reducing position diversity and increasing the risk of falling into a local extreme value. In order to ensure the diversity of individual finders and improve the global search capability of SSA, the sine and cosine strategy adjustment formula is used to cleverly use the vibration change properties of the sine and cosine functions to dynamically adjust the finder's position.

[0077] In the sine-cosine strategy, the step search factor r1 is used to control the size of the search step and balance the search speed and search accuracy. The calculation formula is as follows:

[0078]

[0079] Where: a is a constant; η is an adjustment coefficient with a value not less than 1. In addition, in order to reduce the unnecessary influence of the individual position at the moment on the discoverer position update during the search process of the SSA algorithm, a nonlinear weight factor w is added to the discoverer position update formula:

[0080]

[0081] Then the new discoverer position update formula is as follows:

[0082]

[0083] Where r2 and r3 are random numbers ranging from 0 to 2. r2 describes the distance that the current solution moves when updating to the current optimal solution. r3 gives a random weight to the optimal solution in order to control the influence of the optimal solution in defining the moving distance of the candidate solutions.

[0084] 2> Cauchy mutation mechanism

[0085] During the foraging process, followers usually search around the most ideal discoverer. However, this behavior may lead to a fight for food resources, which in turn prompts the followers themselves to become new discoverers. In order to prevent the algorithm from falling into the dilemma of local optimal solutions and enhance its ability to explore global optimal solutions, the Cauchy mutation mechanism is incorporated into the follower position update formula. This mechanism disturbs the current optimal individual in the follower position update formula, thereby broadening the search space. In this way, the probability of the algorithm jumping out of the local extreme value is increased, thereby improving the overall search effect. The standard Cauchy distribution function is denoted as cauchy(0,1), and the new follower position update formula is as follows:

[0086]

[0087] in, Represents multiplication calculation.

[0088] Specifically, the setting of BiLSTM model parameters is crucial to achieve high prediction accuracy. When establishing a BiLSTM network for photovoltaic output power prediction, the following model parameters need to be determined: input layer dimension, output layer dimension, number of model layers, and number of hidden units in each BiLSTM layer. Among them, six environmental sequence data including total radiation, direct radiation, scattered radiation, component temperature, ambient temperature, and humidity and historical photovoltaic power data are selected as the input layer, and the input layer dimension is 7; photovoltaic power is used as the output layer, and the output layer dimension is 1; the model adopts a three hidden layer structure, and the dimension of each layer is optimized by SCSSA.

[0089] In the SCSSA algorithm, the sparrow population size is set to 30, and the maximum number of iterations is 20. The discoverers account for 30% of the population, and the safety threshold is 0.7. When the warning value is less than 0.7, it indicates that no predators are present and the population is relatively safe. Conversely, if the warning value is greater than or equal to 0.7, it means that predators are present and the population is in danger, and they need to leave the current location to forage elsewhere to ensure safety. The SCSSA algorithm optimizes the BiLSTM parameters. The ranges of the number of neurons in the three hidden layers are [10, 500], [10, 30], and [10, 30] respectively, the range of the regularization coefficient is [0.0001, 0.01], and the range of the initial learning rate is

[0090] [0.0001, 0.01]. After optimization, the best initial learning rate is 0.01, the best regularization coefficient is 0.00011436, and the number of neurons in the three hidden layers are 500, 30, and 30 respectively. The total number of training samples is 8160 groups of data. The last 864 groups of data are selected as test samples, and the remaining data are used as training samples.

[0091] Based on the deep learning network structure of the sparrow search algorithm integrating sine-cosine and Cauchy mutation, convolutional neural network and bidirectional long short-term memory neural network (SCSSA-CNN-BiLSTM), the convolutional layer and pooling layer of the CNN network automatically extract the internal features of the input data. The convolutional layer is used for effective non-linear local feature extraction, and the pooling layer compresses the extracted features through the max-pooling method and generates more critical feature information. BiLSTM is responsible for modeling the non-linear relationship between photovoltaic power and its influencing factors. In the network parameter optimization part, the SCSSA optimization algorithm is used to optimize the network parameters of the model, and finally the prediction of photovoltaic power generation is completed.

[0092] The SCSSA optimization algorithm includes:

[0093] S401. Set the population size n, the maximum number of iterations, the ratio of discoverers to vigilant ones, the vigilance threshold, and the safety threshold;

[0094] S402. Randomly initialize the positions of n sparrows in the optimization range as the initial population positions, generate the refraction reverse population, merge the initial population and the refraction reverse population, sort them according to the rise and fall of the fitness value, and select the top n sparrow individuals with the fitness value as the initial population to initialize the population;

[0095] S403. Calculate the fitness value of each sparrow and sort them, determine the current optimal and worst fitness individuals, and update the positions of the discoverers, followers, and vigilant ones;

[0096] S404. Judge whether the current number of iterations reaches the end condition;

[0097] S405: If the end condition is not met, return to step S403 to loop;

[0098] S406: If the end condition is met, the optimal fitness value and the optimal position are output. The optimized learning rate, number of hidden layer nodes and regularization parameters are obtained, and the SCSSA-CNN-BiLSTM model is trained after obtaining the optimal parameters.

[0099] S5. Use the SCSSA-CNN-BiLSTM model to train the training set samples, and input the test set into the trained SCSSA-CNN-Bi LSTM model for testing;

[0100] S6. Denormalize the prediction results and calculate the model evaluation index to obtain the trained SCSSA-CNN-BiLSTM model.

[0101] S7. Use the trained SCSSA-CNN-BiLSTM model to predict photovoltaic output power.

[0102] (4) Cloud Server:

[0103] The cloud server receives photovoltaic power prediction results and monitoring data of the photovoltaic energy storage power station battery management system to perform charging and discharging optimization and battery health assessment.

[0104] The cloud server includes a charge and discharge optimization module and a battery health assessment module; the charge and discharge optimization module formulates a charge and discharge strategy based on the prediction results of photovoltaic output power, combined with the energy storage status and grid demand, to improve the utilization efficiency and life of the energy storage system; for example, when it is predicted that the photovoltaic power is about to drop, the energy storage system is adjusted to a discharge state in advance to supplement the power that is about to be lost. A reasonable charge and discharge strategy can avoid overcharging, over-discharging and frequent high-current charging and discharging of energy storage batteries, reduce battery aging losses, extend the service life of the energy storage system, and reduce operation and maintenance costs.

[0105] In addition, by predicting the photovoltaic output power, the energy storage system can prepare in advance and quickly adjust the charge and discharge when the photovoltaic power fluctuates, smooth the power output, reduce the impact on the power grid, and ensure the stability of the power grid and the quality of power. At the same time, accurate predictions can enable photovoltaic energy storage power stations to know the output of photovoltaic power generation in advance, and then reasonably adjust the charge and discharge status of the energy storage system according to the load demand of the power grid, ensure the power balance between the power station and the power grid, prevent the occurrence of excess or insufficient power, and maintain the stable operation of the power system. Through the long-term prediction of photovoltaic output power, the power fluctuation characteristics and energy demand of photovoltaic power generation can be analyzed, so as to more accurately determine the capacity and power level of the energy storage system, and avoid excessive energy storage capacity causing waste of investment, or too small capacity to meet the regulation needs.

[0106] The battery health assessment module conducts battery health assessment based on the battery charge and discharge data monitored by the battery management system.

[0107] The battery health assessment module assesses the health status of a single battery in the battery array, that is, calculates the ratio of the current rated capacity to the initial rated capacity of the energy storage unit:

[0108]

[0109] Among them, C t is the current capacity of the energy storage unit, and C0 is the initial capacity; the battery cycle count is determined based on the battery charge and discharge data monitored by the battery management system, and the following is calculated:

[0110]

[0111] Among them, R EOL is the internal resistance value at the end of the battery life, R a is the current internal resistance value, R r is the specified internal resistance value when the battery leaves the factory, N rem is the current cycle count, N tol is the total cycle count;

[0112] Finally, the calculated battery health status is: SOH = A × S OH1 + B × S OH2 + C × S OH3 , A + B + C = 1, where A, B, and C are all weight values.

[0113] During the continuous charge and discharge process, the SOH of the battery decreases accordingly. When the SOH of the battery continuously decreases to a limit value, the risk also increases continuously. Therefore, it is considered that the battery fails at this time and needs to be replaced.

[0114] (5) Database:

[0115] The database uses the MySQL database for data management and storage. In addition to requiring real-time display of the operation information of the photovoltaic power station, the photovoltaic power station monitoring system also needs to save this data. After a long time of operation of the power station data, a large amount of data information is formed. Therefore, it is required to organize and manage this data by certain means. The database organizes, describes, stores, and manages data sets with a professional data model. Users can easily perform operations such as data modification, search, addition, and deletion using the database. The present invention uses the MySQL database to store and manage data in the form of tables, and graphically manages the database through the MySQL Workbench software, giving full play to the advantages of high flexibility and fast retrieval speed of MySQL.

[0116] In the present invention, multiple data acquisition nodes collect the working environment parameter data and equipment operation status information that need to be monitored in the photovoltaic energy storage power station by using sensors and transmit them to the main control single-chip microcomputer for photovoltaic power prediction; the cloud server receives the photovoltaic power prediction results and the monitoring data of the battery management system of the photovoltaic energy storage power station to perform charge and discharge optimization and battery health assessment. The photovoltaic power prediction adopts the SCSSA-CNN-BiLSTM model to improve the prediction accuracy, so as to better optimize the battery charge and discharge strategy, and then evaluate the battery health status based on the detection data of the battery management system to determine the life and safety of the battery, realizing the multi-functional safety monitoring of the photovoltaic energy storage power station.

[0117] It should be noted that in this article, the terms "including", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the phrase "including a..." does not exclude the existence of additional identical elements in the process, device, article or method including that element.

[0118] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A multifunctional photovoltaic energy storage safety device, characterized in that: It includes multiple data acquisition nodes, a first transmission module, a second transmission module, a main control microcontroller, a photovoltaic output power prediction module, a cloud server, a database and a client module, wherein the multiple data acquisition nodes are connected to the first transmission module, the main control microcontroller is respectively connected to the first transmission module, the second transmission module and the photovoltaic output power prediction module, and the cloud server is respectively connected to the database, the second transmission module and the client module; Multiple data acquisition nodes are composed of various sensors and STM32 microcontrollers. Sensors are used to collect working environment parameter data and equipment operation status information that need to be monitored in photovoltaic energy storage power stations and transmit them to the main control microcontroller; The transmission module is used to transmit the working environment parameter data and equipment operation status information collected by multiple data collection nodes to the main control microcontroller for photovoltaic power prediction; The cloud server receives photovoltaic power prediction results and monitoring data of the photovoltaic energy storage power station battery management system to perform charging and discharging optimization and battery health assessment; The database uses MySQL database for data management and storage, and the client module is used for data display and realization of human-computer interaction function.

2. The multifunctional photovoltaic energy storage safety device according to claim 1, characterized in that: Each of the multiple data acquisition nodes includes a light intensity sensor, a component temperature sensor, an ambient temperature sensor, a photovoltaic panel temperature sensor and an STM32 microcontroller; The ambient temperature sensor uses a sensor that can measure temperature and humidity at the same time. The light intensity sensor senses the intensity changes of light in the environment and converts it into an electrical signal for output. The light intensity sensor, component temperature sensor, ambient temperature sensor, and photovoltaic panel temperature sensor are all connected to the STM32 microcontroller via RS485 and UART.

3. The multifunctional photovoltaic energy storage safety device according to claim 1, characterized in that: The first transmission module adopts multiple LoRa communication modules, and each LoRa communication unit corresponds to a data acquisition node; the second transmission module adopts a NB-IOT communication module.

4. The multifunctional photovoltaic energy storage safety device according to claim 2, characterized in that: The photovoltaic output power prediction module uses the data collected by the light intensity sensor, component temperature sensor, ambient temperature sensor, and photovoltaic panel temperature sensor to predict the photovoltaic output power. The specific steps are as follows: S1. Clean the collected total solar irradiance, direct solar radiation, scattered solar radiation, module temperature, ambient temperature, humidity and photovoltaic power data, remove the full-day data containing invalid data, and normalize the data set using a linear function normalization method; S2, split the data set into training set and test set; S3. Build a CNN-BiLSTM model and initialize the BiLSTM model parameters. The model uses the CNN algorithm to extract features from the input data and uses the BiLSTM layer to model. Then, the output of the BiLSTM layer is used as the input of the fully connected layer to output the prediction results. A Dropout layer is added after each BiLSTM hidden layer to prevent overfitting. S4. Introduce the SCSSA algorithm to optimize the regularization coefficient of BiLSTM, the dimension of the neural network and the learning rate; S5. Use the SCSSA-CNN-BiLSTM model to train the training set samples, and input the test set into the trained SCSSA-CNN-Bi LSTM model for testing; S6. Denormalize the prediction results and calculate the model evaluation index to obtain the trained SCSSA-CNN-BiLSTM model. S7. Use the trained SCSSA-CNN-BiLSTM model to predict photovoltaic output power.

5. The multifunctional photovoltaic energy storage safety device according to claim 4, characterized in that: In step S4, the SCSSA optimization algorithm includes: S401, setting the population size n, the maximum number of iterations, the ratio of discoverers to alerters, the alert threshold, and the safety threshold; S402, randomly initialize n sparrow positions in the optimization range as initial population positions, generate a refracted reverse population, merge the initial population and the refracted reverse population, sort them according to the increase and decrease of fitness values, and select the first n sparrow individuals in fitness values ​​as the initial population to initialize the population; S403, calculate the fitness value of each sparrow and sort them, determine the current best and worst fitness individuals, and update the positions of the discoverer, follower, and sentinel; S404, determining whether the current number of iterations reaches the end condition; S405: If the end condition is not met, return to step S403 to loop; S406: If the end condition is reached, the optimal fitness value and the optimal position are output.

6. The multifunctional photovoltaic energy storage safety device according to claim 1, characterized in that: The cloud server includes a charge and discharge optimization module and a battery health assessment module; the charge and discharge optimization module formulates a charge and discharge strategy based on the prediction results of photovoltaic output power, combined with energy storage status and grid demand; the battery health assessment module performs battery health assessment based on battery charge and discharge data monitored by the battery management system.

7. The multifunctional photovoltaic energy storage safety device according to claim 6, characterized in that: The battery health assessment module assesses the health status of individual batteries in the battery array, that is, calculates the ratio of the current rated capacity of the energy storage unit to the initial rated capacity: Among them, C t is the current capacity of the energy storage unit, C0 is the initial capacity; the number of battery cycles is determined based on the battery charge and discharge data monitored by the battery management system, and calculated: Among them, R EOL is the internal resistance value at the end of battery life, R a is the current internal resistance value, R r The internal resistance value of the battery specified at the factory, N rem is the current cycle number, N tol is the total number of cycles; The final calculated battery health status is: SOH = A × S OH1 +B×S OH2 +C×S OH3 ,A+B+C=1, where A, B, and C are all weight values.

8. The multifunctional photovoltaic energy storage safety device according to claim 1, characterized in that: The client modules include computers, tablets, and mobile phones. Workers can use computer browsers, tablets, and collections to view the working status and historical data of the photovoltaic system in real time, thereby achieving remote safety monitoring of photovoltaic energy storage power stations.

Images