Novel photovoltaic micro-grid energy coupling method and system for agricultural and pastoral areas in Tibetan

Abstract

The invention relates to the technical field of photovoltaic micro-grids, in particular to a novel photovoltaic micro-grid energy coupling method and system for agricultural and pastoral areas in Tibet. Comprising the following steps: collecting photovoltaic power generation data, energy storage data and load data by using a big data collection technology, carrying out feature extraction on the collected data, constructing a neural network model by taking the extracted feature data as an input variable and taking whether photovoltaic power generation and energy storage can meet load power utilization as an output variable, the method comprises the following steps: establishing a neural network model, training the neural network model by using previous data, judging whether the working mode of the photovoltaic micro-grid is switched through the neural network model, predicting the generated power of a photovoltaic cell by using a physical model during switching, and maintaining the dynamic balance of the power of the micro-grid by controlling the operation of an energy storage cell and a load. Whether the working mode of the photovoltaic micro-grid is switched or not is judged through the neural network model, the energy utilization rate is improved, power balance is maintained by controlling the operation of the energy storage battery and the load, and the loss of power equipment is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic microgrids, and in particular to a novel photovoltaic microgrid energy coupling method and system in Tibetan farming and pastoral areas. Background Art

[0002] Photovoltaic microgrid energy coupling is an energy utilization method that combines solar photovoltaic power generation systems with microgrid technology. It aims to improve energy utilization efficiency, promote renewable energy access, and enhance the flexibility and reliability of the power system. Photovoltaic microgrids are mainly composed of photovoltaic arrays, energy storage devices, power electronic conversion devices, and control systems. The working mode is that when the daytime irradiation is low, the power consumption is insufficient and supplemented by energy storage. When the irradiation is high, part of the power is directly used by the equipment, and the surplus power is stored in the battery. At night, the solar photovoltaic panels stop generating electricity. At this time, the power stored in the battery is used first. The island working mode is adopted. When the battery power is insufficient, it is supplemented by the power grid. In rainy weather, it is supplemented by energy storage and the power grid. The grid-connected working mode is adopted to ensure the normal power consumption of residents.

[0003] For plateau areas, if photovoltaic microgrid energy coupling is used to utilize solar energy to power residents, the sunshine time will continue to change with the change of seasons, and the working mode of the photovoltaic microgrid cannot be correctly adjusted. If the photovoltaic power generation is simply compared with the user's electricity consumption and the working mode of the photovoltaic microgrid is adjusted, it will lead to energy waste and even affect the normal electricity demand of residents. In order to dynamically adjust the working mode of the photovoltaic microgrid in different seasons, improve energy utilization, and ensure the normal electricity use of residents, we propose a new photovoltaic microgrid energy coupling method and system for Tibetan agricultural and pastoral areas. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that in plateau areas, with the change of seasons, the sunshine time will constantly change, and the working mode of the photovoltaic microgrid cannot be correctly adjusted to improve energy utilization and ensure the normal electricity use of residents. When the photovoltaic microgrid switches the working mode, how to control the energy storage battery and load operation to maintain the dynamic balance of the microgrid power and reduce the loss of power equipment.

[0005] To achieve the above purpose, the present invention provides a novel photovoltaic microgrid energy coupling method in Tibet's agricultural and pastoral areas, comprising the following steps:

[0006] S1. Use big data collection technology to collect photovoltaic power generation data, energy storage data and load data, and extract features from the collected data. At the same time, collect meteorological data such as solar radiation intensity and battery temperature;

[0007] S2. Take the extracted feature data as input variables, and whether photovoltaic power generation and energy storage can meet the load power demand as output variables. If they can meet the demand, switch to island working mode; if not, switch to grid-connected working mode. Construct a neural network model and use previous data to train the neural network model.

[0008] S3. Use the neural network model to determine whether the photovoltaic microgrid switches its working mode. When switching, use the physical model to predict the power generation of the photovoltaic cells, and maintain the dynamic balance of the microgrid power by controlling the energy storage battery and load operation.

[0009] Preferably, the step of collecting data in S1 is as follows:

[0010] S1.1.1. Collect photovoltaic power generation data using solar radiation sensors and temperature sensors;

[0011] S1.1.2. Complete the collection of energy storage data by collecting battery management system data and energy storage converter data;

[0012] S1.1.3, collect load data through smart meters and load sensors;

[0013] S1.1.4. Extract the power generation fluctuation characteristic data, energy storage system SOC characteristic data and load power demand characteristic data based on the collected data.

[0014] Preferably, S1.1.4 uses statistical analysis methods and time series analysis methods to calculate the standard deviation, variance, median, quantile, average power, power peak and valley values ​​of relevant characteristics, and extracts the power generation power fluctuation characteristic data, energy storage system SOC characteristic data and load power demand characteristic data.

[0015] Preferably, the S2 constructs a neural network model, and the method steps are as follows:

[0016] S2.1.1. Determine the input layer nodes: the power standard deviation, coefficient of variation, and autocorrelation coefficient in the power fluctuation characteristic data are used as nodes; the SOC range, change rate, and mean value in the energy storage system SOC characteristic data are used as nodes; the average power, power peak, and power fluctuation rate in the load power demand characteristic data are used as nodes;

[0017] S2.1.2, Determine the number of hidden layers: First try 1 to 3 hidden layers. The number of nodes in each layer can be preliminarily estimated based on the number of input nodes and output nodes. Start with half or two-thirds of the number of input nodes in each layer, and gradually adjust according to the experiment until it is determined;

[0018] S2.1.3. Determine the output layer: The output layer has only one node, which is used to output the result of whether photovoltaic power generation and energy storage can meet the load power demand. The Sigmoid function is used to map the output to the [0,1] interval, where 0 means it cannot be met and 1 means it can be met.

[0019] S2.1.4, Model training: Divide the collected data into training set, validation set and test set, and train the neural network model.

[0020] Preferably, when determining the number of hidden layers in S2.1.2, the ReLU function is used as the activation function to help the model quickly learn the complex nonlinear relationship between these input features and reduce the gradient vanishing problem.

[0021] Preferably, when training the S2.1.4 model, a binary cross entropy loss function is selected to measure the difference between the predicted output and the true label, and the neural network is trained by minimizing this loss function. The formula is:

[0022]

[0023] Where N is the number of samples, y i is the true label, is the predicted output.

[0024] Preferably, S3 maintains the dynamic balance of the microgrid power, and the method steps are as follows:

[0025] S3.1.1. Establish a physical model of photovoltaic power generation and predict photovoltaic power generation. The formula is:

[0026] P = η × A × G × (1-α × (TT ref ))

[0027] Where P is the power generation, η is the photoelectric conversion efficiency of the photovoltaic cell, A is the area of ​​the photovoltaic panel, G is the solar radiation intensity, α is the temperature coefficient, T is the actual temperature of the photovoltaic panel, T ref is the reference temperature;

[0028] S3.1.2. Compare the predicted photovoltaic power generation with the load power, and maintain power balance by controlling the energy storage battery and load operation.

[0029] Preferably, when controlling the energy storage battery to maintain power balance, S3.1.2 sets upper and lower limits of SOC according to the standard performance of the battery, and adjusts the upper and lower limits of the battery SOC regularly as the battery is used.

[0030] Preferably, when controlling the load operation to maintain power balance, S3.1.2 prioritizes the loads according to actual needs of residents and classifies them into critical loads and non-critical loads.

[0031] The second object of the present invention is to provide a new photovoltaic microgrid energy coupling system in Tibet's agricultural and pastoral areas, including any one of the new photovoltaic microgrid energy coupling methods in Tibet's agricultural and pastoral areas described above, characterized in that it includes a data acquisition module, a model building module and a switching adjustment module;

[0032] The data acquisition module uses big data acquisition technology to collect photovoltaic power generation data, energy storage data and load data, and uses statistical analysis methods and time series analysis methods to calculate the standard deviation, variance, median, quantile, average power, power peak and valley values ​​of relevant features, and transmits the feature data to the model construction module;

[0033] The model building module uses the extracted feature data as input variables and whether photovoltaic power generation and energy storage can meet the load power demand as output variables. If it can be met, the isolated island working mode is switched, and if it cannot be met, the grid-connected working mode is switched. A neural network model is constructed, and a binary cross entropy loss function is used to measure the difference between the predicted output and the true label. The neural network is trained by minimizing this loss function.

[0034] The switching adjustment module determines whether the photovoltaic microgrid switches the working mode through a neural network model. When switching, it uses a physical model to predict the power generation of the photovoltaic cell and maintains the dynamic balance of the microgrid power by controlling the energy storage battery and the load operation.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. The new photovoltaic microgrid energy coupling method and system in Tibet's agricultural and pastoral areas uses big data acquisition technology to collect photovoltaic power generation data, energy storage data and load data through the data acquisition module, and extracts relevant feature data for the model construction module to build a neural network model for whether the photovoltaic microgrid switches the working mode. By collecting real-time data and substituting it into the neural network model, when the power generation power of the photovoltaic module can meet the needs of the load equipment, the island working mode is switched, and the excess power is stored in the battery to reduce energy consumption. When the power generation power of the photovoltaic module cannot meet the needs of the load equipment, the grid-connected working mode is switched, and the insufficient power is supplemented by the power grid to maintain the normal electricity demand of residents and improve energy utilization.

[0037] 2. When the photovoltaic microgrid switches its working mode, the physical model is used to predict the power generation of the photovoltaic cells, and the predicted photovoltaic power generation is compared with the load power. The power balance is maintained by controlling the operation of the energy storage battery and the load to prevent the power generation power of the photovoltaic microgrid from differing too much from the load and the power grid during switching, resulting in excessive impact current and damage to the power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is the overall flow chart of the present invention;

[0039] Figure 2 It is a data collection flow chart of the present invention;

[0040] Figure 3 A flowchart of constructing a neural network model of the present invention;

[0041] Figure 4 A flow chart of maintaining dynamic balance of microgrid power according to the present invention;

[0042] Figure 5 It is a system flow chart of the present invention.

[0043] The meaning of each mark in the figure is:

[0044] 100. Data collection module; 200. Model building module; 300. Switching adjustment module. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] At present, in plateau areas, with the change of seasons, the sunshine time will constantly change, and it is impossible to correctly adjust the working mode of photovoltaic microgrids, improve energy utilization, and ensure the normal electricity use of residents. When the photovoltaic microgrid switches the working mode, how to control the energy storage battery and load operation to maintain the dynamic balance of microgrid power and reduce the loss of power equipment.

[0047] Therefore, the present invention proposes to collect photovoltaic power generation data, energy storage data and load data through a data acquisition module, and extract relevant feature data, which is used for the model construction module to construct a neural network model of whether the photovoltaic microgrid switches the working mode, select the working mode of the photovoltaic microgrid, and use the physical model to predict the power generation power of the photovoltaic cell, and maintain power balance by controlling the operation of the energy storage battery and the load.

[0048] Example 1

[0049] The details are as follows:

[0050] like Figure 1 As shown, one of the purposes of the present invention is to provide a novel photovoltaic microgrid energy coupling method in agricultural and pastoral areas, comprising the following steps:

[0051] S1. Use big data collection technology to collect photovoltaic power generation data, energy storage data and load data, and extract features from the collected data. At the same time, collect meteorological data such as solar radiation intensity and battery temperature;

[0052] S2. Take the extracted feature data as input variables, and whether photovoltaic power generation and energy storage can meet the load power demand as output variables. If they can meet the demand, switch to island working mode; if not, switch to grid-connected working mode. Construct a neural network model and use previous data to train the neural network model.

[0053] S3. Use the neural network model to determine whether the photovoltaic microgrid switches its working mode. When switching, use the physical model to predict the power generation of the photovoltaic cells, and maintain the dynamic balance of the microgrid power by controlling the energy storage battery and load operation.

[0054] The present invention uses a data acquisition module 100 to collect photovoltaic power generation data, energy storage data and load data using big data acquisition technology, and extracts relevant feature data for the model building module 200 to build a neural network model of whether the photovoltaic microgrid switches the working mode. By collecting real-time data and substituting it into the neural network model, when the power generation power of the photovoltaic module can meet the needs of the load equipment, the island working mode is switched, and the excess power is stored in the battery to reduce energy consumption. When the power generation power of the photovoltaic module cannot meet the needs of the load equipment, the grid-connected working mode is switched, and the power grid supplements the insufficient power to maintain the normal electricity demand of residents and improve energy utilization. When the photovoltaic microgrid switches the working mode, the physical model is used to predict the power generation power of the photovoltaic cell, and the predicted photovoltaic power generation power is compared with the load power. The power balance is maintained by controlling the operation of the energy storage battery and the load to prevent the power generation power of the photovoltaic microgrid from being too different from the power of the load and the power grid during switching, and the impact current generated is too large, causing damage to the power equipment.

[0055] like Figure 2 As shown, S1 collects data, and the method steps are as follows:

[0056] S1.1.1. Collect photovoltaic power generation data using solar radiation sensors and temperature sensors;

[0057] S1.1.2. Complete the collection of energy storage data by collecting battery management system data and energy storage converter data;

[0058] S1.1.3, collect load data through smart meters and load sensors;

[0059] S1.1.4. Extract the power generation fluctuation characteristic data, energy storage system SOC characteristic data and load power demand characteristic data based on the collected data.

[0060] Among them, solar radiation sensors are installed at different locations of the photovoltaic power station to measure the intensity of solar radiation, and temperature sensors are installed on the back of the photovoltaic panels to monitor the operating temperature of the panels. For the energy storage system, BMS is the core of data acquisition. BMS can monitor the state of charge (SOC), health state (SOH), charge and discharge current, charge and discharge voltage, battery temperature and other parameters of the energy storage battery in real time. The energy storage inverter (PCS) is used to realize the energy conversion between the battery and the power grid or load. The PCS can record its input / output power, power factor, operating frequency and other data. Smart meters are installed at the load access point to measure the load power, current, voltage, power factor and other parameters. Smart meters can distinguish different types of loads (such as single-phase loads, three-phase loads) and can record the real-time power changes of the loads.

[0061] In order to better extract the characteristic data from the collected data, S1.1.4 uses statistical analysis methods and time series analysis methods to calculate the standard deviation, variance, median, quantile, average power, power peak and valley values ​​of relevant characteristics, and extracts the characteristic data of power generation fluctuation, energy storage system SOC characteristic data and load power demand characteristic data;

[0062] Statistical analysis is a method that reveals the inherent laws of data by collecting, organizing, analyzing and interpreting data. It is mainly based on the principles of probability theory and mathematical statistics, and describes and infers the central tendency, degree of dispersion, distribution form, etc. of data. For example, when analyzing power generation data, statistical analysis can be used to calculate the average power generation to understand the overall power generation level of the power generation system, and the fluctuation of power generation can be measured by calculating the standard deviation.

[0063] Time series analysis is a method specifically used to analyze data sequences arranged in chronological order. It focuses on the changing patterns of data over time, including trends, seasonality, periodicity, and irregular fluctuations. For example, photovoltaic power generation has obvious time series characteristics due to the influence of daily and seasonal changes in solar radiation intensity. Time series analysis can better predict power generation.

[0064] like Figure 3 As shown, S2 constructs a neural network model, and the method steps are as follows:

[0065] S2.1.1. Determine the input layer nodes: the power standard deviation, coefficient of variation, and autocorrelation coefficient in the power fluctuation characteristic data are used as nodes; the SOC range, change rate, and mean value in the energy storage system SOC characteristic data are used as nodes; the average power, power peak, and power fluctuation rate in the load power demand characteristic data are used as nodes;

[0066] S2.1.2, Determine the number of hidden layers: First try 1 to 3 hidden layers. The number of nodes in each layer can be preliminarily estimated based on the number of input nodes and output nodes. Start with half or two-thirds of the number of input nodes in each layer, and gradually adjust according to the experiment until it is determined;

[0067] S2.1.3. Determine the output layer: The output layer has only one node, which is used to output the result of whether photovoltaic power generation and energy storage can meet the load power demand. The Sigmoid function is used to map the output to the [0,1] interval, where 0 means it cannot be met and 1 means it can be met.

[0068] S2.1.4, Model training: Divide the collected data into training set, validation set and test set, and train the neural network model;

[0069] When dividing the data set, the training set accounts for 70%-80%, the validation set accounts for 10%-15%, and the test set accounts for 10%-15%. For example, if there are 1,000 sets of data, 700 sets can be used as training sets, 150 sets as validation sets, and 150 sets as test sets. The validation set is used to adjust the hyperparameters of the model during the training process, and the test set is used to evaluate the performance of the final model.

[0070] When training the neural network model, the Adam optimization algorithm is used to update the weights of the neural network. The learning rate is adaptively adjusted according to the historical gradient information of each parameter to update the weights of the neural network. During the training process, the input data (power generation power fluctuation characteristics, energy storage system SOC characteristics and load power demand characteristics) are input into the neural network, and the output is calculated according to the current weights and biases. The error between the predicted output and the actual output is then calculated using the binary cross entropy loss function. The Adam optimization algorithm is used to update the weights and biases of the neural network according to the error back propagation. This process is repeated until the predetermined number of training rounds (such as 1000 rounds) is reached or the loss on the validation set no longer decreases, and the construction of the neural network model is completed.

[0071] In order to determine the number of hidden layers more accurately, S2.1.2 uses the ReLU function as the activation function when determining the number of hidden layers to help the model quickly learn the complex nonlinear relationship between these input features and reduce the gradient vanishing problem;

[0072] The function expression of the ReLU function is: f(x) = max(0, x). When x>0, the function output is equal to the input x. When x<0, the output is 0.

[0073] During the back-propagation process, as long as the input of the neuron is greater than 0, the gradient can be back-propagated smoothly, allowing the neural network to be trained faster. At the same time, the output of the ReLU function is sparse, that is, the output of some neurons is 0, which can reduce the overfitting of the model to a certain extent.

[0074] In order to better train the neural network model, when training the S2.1.4 model, a binary cross entropy loss function is selected to measure the difference between the predicted output and the true label. The neural network is trained by minimizing this loss function. The formula is:

[0075]

[0076] Where N is the number of samples, y i is the true label, is the predicted output;

[0077] For example, suppose we already have a series of sample data, each sample contains the power generation fluctuation characteristic data, energy storage system SOC characteristic data and load power demand characteristic data as input features, and the corresponding real results indicating whether photovoltaic power generation and energy storage can meet the load power demand (0 means it cannot be met, 1 means it can be met) as labels. The sample data is as follows:

[0078]

[0079] From the perspective of information theory, it measures the difference between the model's predicted probability distribution and the true label probability distribution. When the model's prediction is accurate, the loss function value will approach 0; the less accurate the prediction, the larger the loss function value;

[0080] Assume that we have trained the above neural network model for a period of time. Now we predict a new set of data and calculate the loss function value. For example, for a new sample, its input feature data is the power generation power fluctuation feature value of 0.25, the energy storage system SOC feature value of 0.55, and the load power demand feature value of 3.2KW. After the forward propagation calculation of the neural network model, the prediction probability of the output layer is obtained. is 0.7, and the true label y corresponding to this sample i is 1. According to the binary cross entropy loss function formula, the loss contribution of this sample is 0.357. If we have multiple such new samples (assuming 5), we can calculate the loss contribution of each sample separately, add them together according to the formula and divide them by the number of samples 5 to get the overall binary cross entropy loss function value of this group of new samples.

[0081] The calculation results of the other 4 new samples are as follows:

[0082] Sample No. Prediction probability True label Loss Contribution Loss Contribution 1 0.7 1 Log(0.7) 0.357 2 0.4 0 Log(0.6) 0.511 3 0.8 1 Log(0.8) 0.223 4 0.3 0 Log(0.7) 0.357

[0083] The overall binary cross entropy loss function value is 0.392. By calculating the gradient of the binary cross entropy loss function with respect to the model parameters (such as the weights and biases of each layer of the neural network), and then using optimization algorithms such as gradient descent to update these parameters, the loss function value gradually decreases. By continuously adjusting the model parameters to minimize the binary cross entropy loss function value, we can train a neural network model that can more accurately determine whether photovoltaic power generation and energy storage can meet the power consumption of the load.

[0084] like Figure 4 As shown, S3 maintains the dynamic balance of microgrid power, and the method steps are as follows:

[0085] S3.1.1. Establish a physical model of photovoltaic power generation and predict photovoltaic power generation. The formula is:

[0086] P = η × A × G × (1-α × (TT ref ))

[0087] Where P is the power generation, η is the photoelectric conversion efficiency of the photovoltaic cell, A is the area of ​​the photovoltaic panel, G is the solar radiation intensity, α is the temperature coefficient, T is the actual temperature of the photovoltaic panel, T ref is the reference temperature;

[0088] S3.1.2. Compare the predicted photovoltaic power generation with the load power, and maintain power balance by controlling the energy storage battery and load operation;

[0089] During the grid-connected switching process, when it is predicted that the photovoltaic power generation power is higher than the load power, the excess electric energy is stored in the energy storage battery. In the island mode, the supporting role of the power grid disappears, and more accurate power generation power prediction is needed. Since the island operation may face more complex environmental changes (such as the lack of a stable frequency reference from the power grid), the physical model needs to be appropriately adjusted. For example, in the absence of a grid frequency reference, the impact of the inverter's frequency self-regulation ability on the power generation power should be considered. At the same time, for some small distributed photovoltaic systems, it may be necessary to consider the impact of the reflected light of the surrounding environment on the intensity of solar radiation and re-evaluate the power generation power. At the same time, according to the changes in the SOC (state of charge) of the energy storage battery and the load power, the prediction results of the physical model are dynamically updated. If the SOC of the energy storage battery is low, it is necessary to predict the power generation power more cautiously to avoid excessive discharge.

[0090] In order to better protect the battery and extend its service life, S3.1.2 sets the upper and lower limits of SOC according to the standard performance of the battery when controlling the energy storage battery to maintain power balance. As the battery is used, the upper and lower limits of the battery SOC are adjusted regularly;

[0091] After multiple charge and discharge cycles, the capacity of lithium batteries may drop from the initial 100% to 80% or even lower. This means that at the same charge and discharge power, the battery can store or release less electricity. If it continues to be controlled according to the initial SOC upper and lower limits, the battery may reach an overcharge or over-discharge state faster. The battery capacity decay is determined by regular capacity testing. After a certain number of charge and discharge cycles (such as 100 cycles) or every period of time (such as half a year), the battery is fully charged / fully discharged. Test, record the actual amount of electricity that can be charged and discharged, compare it with the initial capacity of the battery, and calculate the capacity decay rate. If it is found that the battery capacity has decayed by a certain proportion (such as 10%), in order to extend the remaining service life of the battery, the SOC upper limit can be appropriately reduced, such as from 0.9 to 0.85, and the lower limit can be appropriately increased, such as from 0.2 to 0.25. This can reduce the working time of the battery in high and low SOC states, slow down the battery aging, and reduce battery loss.

[0092] In order to better regulate load operation without affecting the normal life of residents, S3.1.2, when controlling load operation to maintain power balance, prioritizes loads according to the actual needs of residents and classifies them into critical loads and non-critical loads;

[0093] It is important to prioritize the power supply to critical loads when the power supply is tight. In the event of a grid failure or insufficient power generation, non-critical loads should be cut off first to ensure the stable operation of critical loads.

[0094] The second object of the present invention is to provide a new photovoltaic microgrid energy coupling system for Tibetan agricultural and pastoral areas, including any one of the above-mentioned new photovoltaic microgrid energy coupling methods for Tibetan agricultural and pastoral areas, characterized in that: it includes a data acquisition module 100, a model building module 200 and a switching adjustment module 300;

[0095] The data acquisition module 100 uses big data acquisition technology to collect photovoltaic power generation data, energy storage data and load data, and uses statistical analysis methods and time series analysis methods to calculate the standard deviation, variance, median, quantile, average power, power peak and valley values ​​of relevant features, and transmits the feature data to the model construction module 200;

[0096] The model building module 200 uses the extracted feature data as input variables and whether photovoltaic power generation and energy storage can meet the load power demand as output variables. If it can be met, the isolated island working mode is switched, and if it cannot be met, the grid-connected working mode is switched. A neural network model is constructed, and a binary cross entropy loss function is used to measure the difference between the predicted output and the true label. The neural network is trained by minimizing this loss function.

[0097] The switching adjustment module 300 determines whether the photovoltaic microgrid switches the working mode through the neural network model. When switching, it uses the physical model to predict the power generation of the photovoltaic cell and maintains the dynamic balance of the microgrid power by controlling the energy storage battery and the load operation.

[0098] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A new photovoltaic microgrid energy coupling method for agricultural and pastoral areas in Tibet, characterized in that: The following steps are involved: S1. Use big data collection technology to collect photovoltaic power generation data, energy storage data and load data, and extract features from the collected data. At the same time, collect meteorological data such as solar radiation intensity and battery temperature; S2. Take the extracted feature data as input variables, and whether photovoltaic power generation and energy storage can meet the load power demand as output variables. If they can meet the demand, switch to island working mode; if not, switch to grid-connected working mode. Construct a neural network model and use previous data to train the neural network model. S3. Use the neural network model to determine whether the photovoltaic microgrid switches its working mode. When switching, use the physical model to predict the power generation of the photovoltaic cells, and maintain the dynamic balance of the microgrid power by controlling the energy storage battery and load operation.

2. The new photovoltaic microgrid energy coupling method for agricultural and pastoral areas in Tibet according to claim 1 is characterized by: The S1 data collection method and steps are as follows: S1.1.

1. Collect photovoltaic power generation data using solar radiation sensors and temperature sensors; S1.1.

2. Complete the collection of energy storage data by collecting battery management system data and energy storage converter data; S1.1.3, collect load data through smart meters and load sensors; S1.1.

4. Extract the power generation fluctuation characteristic data, energy storage system SOC characteristic data and load power demand characteristic data based on the collected data.

3. The new photovoltaic microgrid energy coupling method for agricultural and pastoral areas in Tibet according to claim 2 is characterized by: S1.1.4 uses statistical analysis methods and time series analysis methods to calculate the standard deviation, variance, median, quantile, average power, power peak and valley values ​​of relevant characteristics, and extracts power generation fluctuation characteristic data, energy storage system SOC characteristic data and load power demand characteristic data.

4. The novel photovoltaic microgrid energy coupling method for agricultural and pastoral areas in Tibet according to claim 1 is characterized by: The S2 constructs a neural network model, and the method steps are as follows: S2.1.

1. Determine the input layer nodes: the power standard deviation, coefficient of variation, and autocorrelation coefficient in the power fluctuation characteristic data are used as nodes; the SOC range, change rate, and mean value in the energy storage system SOC characteristic data are used as nodes; the average power, power peak, and power fluctuation rate in the load power demand characteristic data are used as nodes; S2.1.2, Determine the number of hidden layers: First try 1 to 3 hidden layers. The number of nodes in each layer can be preliminarily estimated based on the number of input nodes and output nodes. Start with half or two-thirds of the number of input nodes in each layer, and gradually adjust according to the experiment until it is determined; S2.1.

3. Determine the output layer: The output layer has only one node, which is used to output the result of whether photovoltaic power generation and energy storage can meet the load power demand. The Sigmoid function is used to map the output to the [0,1] interval, where 0 means it cannot be met and 1 means it can be met. S2.1.4, Model training: Divide the collected data into training set, validation set and test set, and train the neural network model.

5. The novel photovoltaic microgrid energy coupling method for agricultural and pastoral areas in Tibet according to claim 4 is characterized by: When determining the number of hidden layers in S2.1.2, the ReLU function is used as the activation function to help the model quickly learn the complex nonlinear relationship between these input features and reduce the gradient vanishing problem.

6. The new photovoltaic microgrid energy coupling method for agricultural and pastoral areas in Tibet according to claim 4 is characterized by: When training the S2.1.4 model, a binary cross entropy loss function is selected to measure the difference between the predicted output and the true label. The neural network is trained by minimizing this loss function. The formula is: Where N is the number of samples, y i is the true label, is the predicted output.

7. The new photovoltaic microgrid energy coupling method for Tibet's agricultural and pastoral areas according to claim 1 is characterized by: S3 maintains the dynamic balance of microgrid power, and the method steps are as follows: S3.1.

1. Establish a physical model of photovoltaic power generation and predict photovoltaic power generation. The formula is: P=η×A×G×(1-α×(T-T ref )) Where P is the power generation, η is the photoelectric conversion efficiency of the photovoltaic cell, A is the area of ​​the photovoltaic panel, G is the solar radiation intensity, α is the temperature coefficient, T is the actual temperature of the photovoltaic panel, T ref is the reference temperature; S3.1.

2. Compare the predicted photovoltaic power generation with the load power, and maintain power balance by controlling the energy storage battery and load operation.

8. The novel photovoltaic microgrid energy coupling method for agricultural and pastoral areas in Tibet according to claim 7 is characterized by: When controlling the energy storage battery to maintain power balance, S3.1.2 sets the upper and lower limits of the SOC according to the standard performance of the battery, and adjusts the upper and lower limits of the battery SOC regularly as the battery is used.

9. The novel photovoltaic microgrid energy coupling method for agricultural and pastoral areas in Tibet according to claim 7 is characterized by: S3.1.2, when controlling load operation to maintain power balance, prioritizes loads according to actual needs of residents and classifies them into critical loads and non-critical loads.

10. A new photovoltaic microgrid energy coupling system for Tibetan agricultural and pastoral areas, applied to the new photovoltaic microgrid energy coupling method for Tibetan agricultural and pastoral areas as claimed in any one of claims 1 to 9, characterized in that: It comprises a data acquisition module (100), a model building module (200) and a switching adjustment module (300); The data acquisition module (100) uses big data acquisition technology to collect photovoltaic power generation data, energy storage data and load data, uses statistical analysis methods and time series analysis methods to calculate the standard deviation, variance, median, quantile, average power, power peak and valley values ​​of relevant characteristics, and transmits the characteristic data to the model construction module (200); The model building module (200) uses the extracted feature data as input variables and whether photovoltaic power generation and energy storage can meet the load power demand as output variables, and switches to an isolated island working mode if it can be met, and switches to a grid-connected working mode if it cannot be met, constructs a neural network model, a binary cross entropy loss function, measures the difference between the predicted output and the true label, and trains the neural network by minimizing the loss function; The switching adjustment module (300) determines whether the photovoltaic microgrid switches its working mode through a neural network model, and when switching, predicts the power generation of the photovoltaic cell using a physical model, and maintains the dynamic balance of the microgrid power by controlling the operation of the energy storage battery and the load.

Images