Household Photovoltaic Energy Storage System Energy Scheduling Method and Related Equipment

Abstract

The present invention discloses a household photovoltaic energy storage system energy scheduling method and related devices. The method includes: based on weather data and historical photovoltaic data, using a photovoltaic prediction model pre-constructed based on a convolutional neural network and a long short-term memory network to obtain the photovoltaic power generation; based on weather data and historical load data, using a load prediction model pre-constructed based on a convolutional neural network and a long short-term memory network to obtain the load demand power; creating a strategy matrix, performing power distribution simulation according to the strategy matrix, the photovoltaic power generation, and the load demand power, and iteratively optimizing the simulation result through a particle swarm algorithm to obtain the optimal power distribution result. This application can solve the problems of low accuracy of energy scheduling and low user benefits caused by a single energy scheduling strategy in the existing energy management system.

Description

Technical Field

[0001] The present invention relates to the technical field of energy scheduling, and in particular to an energy scheduling method and related equipment for a household photovoltaic energy storage system. Background Art

[0002] Photovoltaic technology is one of the most mature and widely used renewable energy technologies, with great development potential. In recent years, the global demand for renewable energy has been growing, and renewable energy such as photovoltaics is accelerating the replacement of traditional fossil energy. Household photovoltaic energy storage system is an important form of photovoltaic power generation into the grid, and has gradually become a hot topic of research. In order to maximize the benefits of its energy management, an energy management system is needed to reasonably plan power generation, power consumption and grid connection. The current energy management system usually has a relatively simple strategy, generally storing excess photovoltaic energy in energy storage batteries or directly connecting to the grid. The lack of strategy makes it difficult to achieve different electricity price ranges for power or grid connection when dispatching household photovoltaic energy storage systems. The accuracy of energy dispatch is low, and the benefits obtained by users from household photovoltaic energy storage systems are not high, which reduces the user's experience of household photovoltaic systems, and thus affects the development of the photovoltaic industry. Summary of the invention

[0003] The main purpose of the present invention is to provide an energy scheduling method and related equipment for a household photovoltaic energy storage system, so as to at least solve the problem that the existing energy management strategy is single, which makes it difficult to obtain electricity or connect to the grid in different electricity price ranges when scheduling the energy of the household photovoltaic energy storage system, resulting in low benefits for users from the household photovoltaic energy system.

[0004] According to one aspect of the present invention, there is provided a method for energy scheduling of a household photovoltaic energy storage system, comprising:

[0005] Based on weather data and historical photovoltaic data, the photovoltaic power generation power is obtained using a photovoltaic prediction model pre-built based on a convolutional neural network and a long short-term memory network;

[0006] Based on weather data and historical load data, the load demand power is obtained using a load forecasting model pre-built based on a convolutional neural network and a long short-term memory network;

[0007] A strategy matrix is ​​created, a power allocation simulation is performed according to the strategy matrix, the photovoltaic power generation power and the load demand power, and the simulation results are iteratively optimized by a particle swarm algorithm to obtain an optimal power allocation result.

[0008] Furthermore, based on weather data and historical photovoltaic data, the photovoltaic power generation power is obtained by using a photovoltaic prediction model pre-built based on a convolutional neural network and a long short-term memory network, including:

[0009] Obtain historical photovoltaic data of weather data;

[0010] Analyze the weather data and the historical photovoltaic data respectively through the Pearson correlation analysis method to obtain weather-related features and historical photovoltaic features;

[0011] Process the weather-related features through the convolutional layer of the photovoltaic prediction model to obtain photovoltaic key features;

[0012] Process the photovoltaic key features through the long short-term memory network layer of the photovoltaic prediction model to obtain photovoltaic power time series features;

[0013] Fuse the photovoltaic power time series features and the historical photovoltaic features and process them through the convolutional layer of the photovoltaic prediction model to obtain photovoltaic power generation.

[0014] Further, based on weather data and historical load data, use a load prediction model pre-constructed based on a convolutional neural network and a long short-term memory network to obtain load demand power, including:

[0015] Obtain historical load data;

[0016] Analyze the weather data and the historical load data respectively through the Pearson correlation analysis method to obtain weather-related features and historical load features;

[0017] Process the weather-related features and historical load features through the convolutional layer of the load prediction model to obtain photovoltaic key features and load key features;

[0018] Process the photovoltaic key features and load key features respectively through the long short-term memory network layer of the load prediction model to obtain photovoltaic power time series features and load power time series features;

[0019] Fuse the photovoltaic power time series features and the load power time series features to obtain load demand power.

[0020] Further, create a policy matrix, and perform power distribution simulation according to the policy matrix, the photovoltaic power generation, and the load demand power, including:

[0021] Create a corresponding policy matrix based on power distribution requirements;

[0022] Read the photovoltaic power generation curve corresponding to the photovoltaic power generation, and the load demand power curve corresponding to the load demand power;

[0023] Extract the photovoltaic power generation and load demand power at each time step based on the photovoltaic power generation curve and the load demand power curve;

[0024] Calculate the battery power distribution curve according to the external charging and discharging power and the energy storage battery model parameters, where the battery power distribution curve includes the battery state change curve;

[0025] Based on the photovoltaic power generation power and the load demand power at each time step, combined with the current inverter operation mode, calculate the inverter power distribution curve corresponding to the inverter, where the inverter power distribution curve includes the charging power curve, the discharging power curve, the power purchase curve, and the power selling curve;

[0026] Obtain the power distribution simulation result according to the strategy matrix, the battery distribution power curve, and the inverter power distribution curve.

[0027] Further, the inverter operation mode includes the energy storage priority mode, the grid connection priority mode, the self-consumption mode, the power purchase for energy storage mode, and the energy storage power selling mode.

[0028] Further, perform iterative optimization on the simulation result through the particle swarm algorithm to obtain the optimal power distribution result, including:

[0029] Generate a particle swarm according to the power distribution simulation result;

[0030] Judge the control mode of the large load. If it is the fixed time period mode, change all the large load switch strategies of all particles to the user settings. If it is the fixed duration mode, execute the next step;

[0031] Randomly generate the initial velocity of each particle;

[0032] Based on each particle, perform power distribution simulation to obtain the electricity cost expenditure corresponding to each particle, and judge the control mode of the large load. If it is the fixed time period mode, execute the next step. If it is the fixed duration mode, assign a penalty coefficient to the particle;

[0033] Update the individual optimal electricity cost expenditure and the strategy matrix of each particle, and update the global optimum at the same time;

[0034] Calculate the distance and direction of each particle, and based on the distance and direction of each particle, perform weighted update of the particle velocity based on the learning factor to make each particle move towards the individual and global optimal solutions;

[0035] Judge whether it converges or reaches the maximum number of iterations. If it converges or does not reach the maximum number of iterations, return to the step of randomly generating the initial velocity of each particle. If it reaches the maximum number of iterations, output the global optimal solution matrix and the global optimal solution expenditure.

[0036] Further, updating the individual optimal electricity cost expenditure and the strategy matrix of each particle, and updating the global optimum at the same time, includes:

[0037] For each particle, if the electricity cost expenditure for the corresponding day is less than the optimal electricity cost expenditure it has passed through, then use it as the individual optimal strategy matrix and use the corresponding electricity cost expenditure for the day as the individual optimal electricity cost expenditure;

[0038] For each particle, if the electricity cost expenditure for the corresponding day is less than the optimal electricity cost expenditure passed through by all particles, then use it as the global optimal strategy matrix and use the corresponding electricity cost expenditure for the day as the global optimal electricity cost expenditure.

[0039] According to another aspect of the embodiments of the present invention, there is also provided a household photovoltaic energy storage system energy scheduling device, including:

[0040] A photovoltaic prediction module, configured to obtain the photovoltaic power generation power by using a photovoltaic prediction model pre-constructed based on a convolutional neural network and a long short-term memory network based on weather data and historical photovoltaic data;

[0041] A load prediction module, configured to obtain the load demand power by using a load prediction model pre-constructed based on a convolutional neural network and a long short-term memory network based on weather data and historical load data;

[0042] A simulated photovoltaic energy storage system module, configured to perform power distribution simulation according to the strategy matrix, the photovoltaic power generation power, and the load demand power;

[0043] A strategy calculation module, configured to iteratively optimize the simulation result through a particle swarm algorithm to obtain an optimal power distribution result.

[0044] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including: a processor and a memory storing a program. Further, the program includes instructions that, when executed by the processor, cause the processor to execute the household photovoltaic energy storage system energy scheduling method described in the embodiments of the invention.

[0045] According to another aspect of the embodiments of the present invention, there is also provided a non-transitory machine-readable medium storing computer instructions. Further, the computer instructions are used to cause the computer to execute the household photovoltaic energy storage system energy scheduling method described in the embodiments of the invention.

[0046] In the present invention, a photovoltaic prediction model constructed by combining a convolutional neural network (CNN) and a long short-term memory network (LSTM) can make full use of the time series characteristics and spatial features in weather data and historical photovoltaic data to achieve accurate prediction of photovoltaic power generation, and can better capture the impact of complex and variable weather factors on photovoltaic power generation, thereby improving the accuracy of prediction. The load prediction model constructed by using CNN and LSTM, combined with weather data and historical load data, can more scientifically predict the future load demand power, which helps to optimize energy storage and grid connection strategies to meet the power consumption needs in different time periods. It not only predicts the photovoltaic power generation and load demand power, but also preliminarily determines the power distribution scheme by simulating the power distribution results, considering the dynamic balance between power generation and power consumption. By constructing a strategy matrix and using the particle swarm optimization algorithm to find the optimal solution, the intelligent and refined energy scheduling is realized, and the global optimal solution or approximate optimal solution can be quickly found in the complex multi-dimensional space, so as to ensure that the energy scheduling scheme can maximize the user's benefits while adapting to market changes such as electricity price fluctuations. By optimizing in aspects such as prediction, distribution, and scheduling, the household photovoltaic energy storage system can more accurately respond to market demands, reasonably arrange power generation, power consumption, and grid connection times, thereby maximizing the economic benefits of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The drawings described herein are used to provide a further understanding of the present invention, and constitute a part of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention, and do not constitute an improper limitation to the present invention. In the drawings:

[0048] Figure 1 is a flowchart of the energy scheduling method for a household photovoltaic energy storage system disclosed in an embodiment of the present invention;

[0049] Figure 2 is a schematic structural diagram of the photovoltaic prediction model disclosed in an embodiment of the present invention;

[0050] Figure 3 is a schematic structural diagram of the load prediction model disclosed in an embodiment of the present invention;

[0051] Figure 4 is a schematic diagram of the energy scheduling device for a photovoltaic energy storage system disclosed in an embodiment of the present invention;

[0052] Figure 5 is a schematic structural diagram of the electronic device of this embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0053] The embodiments of the present embodiment will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present embodiment are shown in the drawings, it should be understood that the present embodiment can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present embodiment. It should be understood that the drawings and embodiments of the present embodiment are only for exemplary purposes and are not used to limit the protection scope of the present embodiment.

[0054] In the related art, an energy management system generally stores the excess energy of a photovoltaic system in a storage battery or directly connects to the grid. The strategy is relatively single, and it is difficult to take power or connect to the grid in different electricity price intervals during the energy scheduling of a household photovoltaic energy storage system, resulting in low accuracy of energy scheduling and low benefits for users from the household photovoltaic energy storage system.

[0055] In order to solve the problems in the related art that the energy scheduling strategy of the energy management system is relatively single, resulting in low accuracy of energy scheduling and low user benefits.

[0056] See Figure 1 As shown, according to an embodiment of the present application, there is provided a method for energy scheduling of a household photovoltaic energy storage system, including:

[0057] Step S101, based on weather data and historical photovoltaic data, use a photovoltaic prediction model pre-constructed based on a convolutional neural network and a long short-term memory network to obtain the photovoltaic power generation power;

[0058] Step S102, based on weather data and historical load data, use a load prediction model pre-constructed based on a convolutional neural network and a long short-term memory network to obtain the load demand power;

[0059] Step S103, create a policy matrix, perform power distribution simulation according to the policy matrix, photovoltaic power generation power and load demand power, and iteratively optimize the simulation result through a particle swarm optimization algorithm to obtain the optimal power distribution result.

[0060] In an exemplary embodiment, step S101 includes:

[0061] Obtain weather data and historical photovoltaic data;

[0062] Analyze the weather data and historical photovoltaic data respectively through the Pearson correlation analysis method to obtain weather-related features and historical photovoltaic features;

[0063] Process the weather-related features through the convolutional layer of the photovoltaic prediction model to obtain photovoltaic key features;

[0064] Process the photovoltaic key features through the long short-term memory network layer of the photovoltaic prediction model to obtain photovoltaic power time series features;

[0065] Fuse the photovoltaic power time series characteristics with the historical photovoltaic characteristics and process them through the convolutional layer of the photovoltaic prediction model to obtain the photovoltaic power generation.

[0066] After analyzing the weather data (NWP) and historical photovoltaic data through the Pearson correlation analysis method, weather-related characteristics and historical photovoltaic characteristics are obtained. The weather-related characteristics include temperature, cloud opacity, azimuth angle, solar radiation DHI (diffuse horizontal irradiance), DNI (direct normal irradiance), GHI (global horizontal irradiance), relative humidity, and natural time characteristics. As Figure 2 Shown in the figure is the schematic diagram of the photovoltaic prediction model. First, a convolutional layer is constructed with a convolutional neural network model as the core. The convolutional layer mainly extracts key photovoltaic features from the weather-related characteristics, and can determine the differences in weather changes and their impacts on photovoltaic power output. Then, a long short-term memory network (LSTM) is constructed. Through the long short-term memory network, effective information in the sequence data can be mined to obtain the photovoltaic power time series characteristics, that is, the changing trend of photovoltaic power generation over time. At the same time, the historical photovoltaic characteristics are input into the long short-term memory network for time series processing, and then fused in the Cat (concatenation) layer and processed through the convolutional layer to obtain the photovoltaic prediction result, and the corresponding photovoltaic power generation is obtained. This enhances the model's understanding ability of the input data, thus realizing a highly accurate prediction of the day-ahead photovoltaic power result.

[0067] In an exemplary embodiment, step S102 includes:

[0068] Obtain historical load data;

[0069] Analyze the weather data and historical load data respectively through the Pearson correlation analysis method to obtain weather-related characteristics and historical load characteristics;

[0070] Process the weather-related characteristics and historical load characteristics through the convolutional layer of the load prediction model to obtain key photovoltaic features and key load features;

[0071] Process the key photovoltaic features and key load features respectively through the long short-term memory network layer of the load prediction model to obtain photovoltaic power time series characteristics and load power time series characteristics;

[0072] Fuse the photovoltaic power time series characteristics and the load power time series characteristics to obtain the load demand power.

[0073] In this embodiment, historical load data is the historical power consumption data. The Pearson correlation analysis method is used to analyze the weather data to obtain weather-related features. The weather-related features when constructing the load prediction model in this embodiment include that there is a significant correlation between weather factors such as temperature, solar zenith angle, azimuth angle, etc. and historical power consumption data. Weather information such as temperature, solar zenith angle, azimuth angle, time features, and historical load conditions are selected as input variables of the load prediction model. As Figure 3 shown in the schematic diagram of the load prediction model. The convolutional layer of the load prediction model is built based on a convolutional neural network. The convolutional layer of the load prediction model accurately extracts key photovoltaic features from the main physical factors such as weather information and time, and captures the relationship between weather, time changes, and household loads. Then, the long short-term memory network layer of the load prediction model is constructed. The effective information of features in the sequence direction is mined through the long short-term memory network layer to obtain the photovoltaic power time series features and the load power time series features, that is, the trends of photovoltaic power or load power changing with time. The photovoltaic power time series features and the load power time series features are fused together through the Cat layer, and the predicted daily load result, that is, the load demand power, is further extracted through the convolutional layer.

[0074] In an exemplary embodiment, a policy matrix is created, and power distribution simulation is performed according to the policy matrix, photovoltaic power generation, and load demand power, including:

[0075] Create a corresponding policy matrix based on the power distribution requirements;

[0076] Read the photovoltaic power generation curve corresponding to the photovoltaic power generation and the load demand power curve corresponding to the load demand power;

[0077] Extract the photovoltaic power generation and load demand power at each time step from the photovoltaic power generation curve and the load demand power curve;

[0078] Calculate the battery power distribution curve according to the external charge and discharge power and the energy storage battery model parameters. The battery power distribution curve includes the battery state change curve;

[0079] Calculate the inverter corresponding inverter power distribution curve based on the photovoltaic power generation and load demand power at each time step in combination with the current inverter operation mode. The inverter power distribution curve includes the charging power curve, the discharging power curve, the power purchase curve, and the power selling curve;

[0080] Obtain the power distribution simulation result according to the policy matrix, the battery distribution power curve, and the inverter power distribution curve.

[0081] In an exemplary embodiment, the inverter operation modes include the energy storage priority mode, the grid connection priority mode, the self-use mode, the power purchase for energy storage mode, and the energy storage power selling mode.

[0082] In this embodiment, first, a corresponding policy matrix is created according to the actual power distribution requirements, and then the photovoltaic energy storage system is simulated through a software simulation program. The power distribution of the inverter under different operating modes, different photovoltaic inputs, different load demands, and different battery states (SOC) is simulated. And the electricity cost is calculated based on the obtained power distribution and electricity price. The simulated photovoltaic energy storage system consists of two mathematical models, including a storage battery model and an inverter model. In this embodiment, the parameters of the storage battery model include the maximum battery capacity, battery SOC, maximum charge and discharge power of the battery, charge and discharge efficiency, and charge and discharge power. The storage battery model will calculate the change in SOC according to the input of the external charge and discharge power parameters, and the corresponding calculation formula is as follows:

[0083] ;

[0084] Where: represents t the remaining power state at time represents t the remaining power state at time -1; represents the maximum capacity of the storage battery; represents the unit time; represents the efficiency of the storage battery charging process; represents the efficiency of the storage battery discharging process; and represent the charging power and discharging power respectively.

[0085] In this embodiment, the parameters of the inverter model include the inverter power, grid connection limit power, operating mode, photovoltaic input power, and load demand power. The inverter model calculates the power distribution of the inverter according to the input of the external operating mode, photovoltaic input power, and load demand power parameters, and the output includes a charging power curve, a discharging power curve, a power purchase curve, and a power selling curve. The overall system follows the law of conservation of energy, and the corresponding formula is as follows:

[0086] ;

[0087] Where: , represent the power purchase and power selling of the household system, , represent the photovoltaic power generation and load demand power of the household system.

[0088] Understandably, the inverter in this embodiment includes five operating modes, namely energy storage priority mode, grid connection priority mode, self-consumption mode, power purchase for energy storage mode, and energy storage power selling mode. Each operating mode is actually a priority ranking of energy flow directions, and the specific priority ranking is as follows:

[0089] Mode 1 Energy Storage Priority Mode: Energy storage > Load > Grid connection;

[0090] Mode 2 Grid Connection Priority Mode: Load > Grid connection > Energy storage;

[0091] Mode 3 Self-Consumption Mode: Load > Energy storage > Grid connection;

[0092] Mode 4 Power Purchase for Energy Storage Mode: On the basis of Mode 3, the battery purchases electricity for energy storage;

[0093] Mode 5 Energy Storage Power Selling Mode: On the basis of Mode 3, the battery discharges electricity to the grid.

[0094] The power distribution calculation of the inverter is based on the above priorities. For example, in Mode 3 self-consumption, the inverter energy is preferentially allocated to load demand, the surplus photovoltaic power is stored in the energy storage, and the remaining power is connected to the grid. Based on the above modes, the subsequent algorithm can control the power distribution of the inverter to achieve the effect of saving electricity bills or generating electricity bill revenues. Therefore, the output of the simulated photovoltaic energy storage system is a power distribution result based on the current input power data and under the control of the operating mode. Within the dimension of one day, with any time granularity, the power distribution result of one day is calculated, and based on the result and electricity price, the electricity bill revenue of the user is directly estimated. The simulated photovoltaic energy storage system is used as the adaptation function during strategy calculation, and the index of the adaptation function is the electricity bill expenditure.

[0095] In an exemplary embodiment, the simulated results are iteratively optimized through the particle swarm algorithm to obtain the optimal power distribution result, including:

[0096] Generate a particle swarm according to the power distribution simulation result;

[0097] Judge the control mode of the large load. If it is the fixed time period mode, change the large load switch strategy of all particles to the user setting. If it is the fixed duration mode, proceed to the next step;

[0098] Randomly generate the initial velocity of each particle;

[0099] Based on each particle, perform power distribution simulation to obtain the electricity bill expenditure corresponding to each particle, and judge the control mode of the large load. If it is the fixed time period mode, proceed to the next step. If it is the fixed duration mode, assign a penalty coefficient to the particle;

[0100] Update the individual optimal electricity bill expenditure and strategy matrix of each particle, and at the same time update the global optimum;

[0101] Calculate the distance and direction of each particle, and update the particle velocity based on the learning factor by combining the distance and direction of each particle, so that each particle moves towards the individual and global optimal solutions;

[0102] Judge whether it converges or reaches the maximum number of iterations. If it does not converge or does not reach the maximum number of iterations, return to the step of randomly generating the initial velocity of each particle. If it reaches the maximum number of iterations, output the global optimal solution matrix and the global optimal solution expenditure.

[0103] In an exemplary embodiment, updating the individual optimal electricity cost expenditure and policy matrix of each particle while updating the global optimal includes:

[0104] For each particle, if the corresponding electricity cost expenditure for the current day is less than the optimal electricity cost expenditure it has passed through, then use it as the individual optimal policy matrix and use the corresponding electricity cost expenditure for the current day as the individual optimal electricity cost expenditure;

[0105] For each particle, if the corresponding electricity cost expenditure for the current day is less than the optimal electricity cost expenditure passed through by all particles, then use it as the global optimal policy matrix and use the corresponding electricity cost expenditure for the current day as the global optimal electricity cost expenditure.

[0106] In this embodiment, for example, a day of 24 hours is divided into 96 time periods, that is, the time granularity is 15 minutes, and the policy is updated once at each time point. Taking the inverter, heat pump, charging pile, air conditioner, and washing machine as examples, the policy for a single time point is [inverter operation mode, heat pump switch, charging pile switch, air conditioner switch, washing machine switch], the policy dimension is 1*n, the value range of the inverter operation mode is an integer in [1,5], and the value range of the switch is 0 and 1. Therefore, the policy dimension for a day is 96*n, representing the operation mode of the inverter at each time point and the switch states of each large controllable load, which is the policy matrix. Among them, loads such as heat pumps, charging piles, air conditioners, and washing machines are represented as large loads, and the fixed operation time is freely set by the user, which is the fixed time period mode, or the operation time required for a day can be set and calculated by the algorithm itself, which is the fixed duration mode. When optimizing through the particle swarm algorithm in this embodiment, the policy matrix is represented as a particle, and the specific process includes:

[0107] Step 1: Generate a number of particles , each particle is a 96×n policy matrix, and the values are random within the value range;

[0108] Step 2: Judge the large load control mode: If it is the fixed time period mode, change all the large load switch policies of all particles to the user's settings; if it is the fixed duration mode, continue;

[0109] Step 3: Each particle randomly generates an initial velocity ;

[0110] Step 4: Take each particle as the input of the fitness function of the simulated photovoltaic energy storage system module, and output the electricity cost for the day ; Judge the large load control mode: if it is the fixed time period mode, continue; if it is the fixed duration mode, count the switching situation of each particle of the large load, and give a penalty coefficient to the particle , that is , where is the particle after applying the penalty corresponding to the electricity cost for the day, is a variable coefficient, and the farther away from the user-set value, the larger the penalty coefficient;

[0111] Step 5: For each particle, compare its electricity cost for the day with the optimal electricity cost it has passed through . If it is better, then take it as the individual optimal strategy matrix , and the individual optimal electricity cost ;

[0112] Step 6: For each particle, compare its electricity cost for the day with the optimal electricity cost passed through by all particles . If it is better, then take it as the global optimal strategy matrix , and the global optimal electricity cost ;

[0113] Step 7: Calculate the distance and direction between each particle and , , and update them respectively according to the weighted sum of the learning factors and . The update formula is:

[0114] ;

[0115] ;

[0116] where represents the updated position of particle , represents the velocity of particle at time t + 1, represents the velocity of particle at time t, represents the weight parameter, represents the current optimal strategy matrix of particle , is a random number between (0, 1).

[0117] Step Eight: Determine whether it converges or reaches the maximum number of iterations. If it does not converge or does not reach the maximum number of iterations, jump to Step Three; otherwise, output the global optimal strategy matrix , the global optimal electricity cost .

[0118] In this embodiment, the number of iterations is set to 100 - 10,000. When the update formula converges within the number of iterations, stop the iteration, output the result of the update formula, and obtain the final global optimal strategy matrix and the global optimal electricity cost. When the update formula still does not converge after reaching the maximum number of iterations of 10,000 times, also output the result of the formula, and use the final result as the global optimal strategy matrix and the global optimal electricity cost. In this embodiment, by taking the photovoltaic prediction value, load prediction value, electricity price information, and battery state SOC as the input of the optimization algorithm, with the minimization of electricity cost as the optimization goal, iteratively optimize the strategy matrix to obtain the optimal solution. After obtaining the optimal solution, the EMS can perform operations to achieve the purpose of saving electricity costs or generating electricity revenue.

[0119] The embodiment of the present invention also provides an inverter power control device, as Figure 4 shown, the device includes:

[0120] A photovoltaic prediction module 10, configured to obtain the photovoltaic power generation based on weather data and historical photovoltaic data by using a photovoltaic prediction model pre-constructed based on a convolutional neural network and a long short-term memory network;

[0121] A load prediction module 20, configured to obtain the load demand power based on weather data and historical load data by using a load prediction model pre-constructed based on a convolutional neural network and a long short-term memory network;

[0122] A simulated photovoltaic energy storage system module 30, configured to perform power distribution simulation according to the strategy matrix, photovoltaic power generation, and load demand power;

[0123] A strategy calculation module 40, configured to iteratively optimize the simulation result through a particle swarm algorithm to obtain the optimal power distribution result.

[0124] The embodiment of the present invention also provides an electronic device, including: a processor and a memory storing a program, where the program includes instructions, and the instructions, when executed by the processor, cause the processor to execute the household photovoltaic energy storage system energy scheduling method described in the embodiment of the present invention.

[0125] An embodiment of the present invention also provides a non-transitory machine-readable medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to execute the household photovoltaic energy storage system energy scheduling method described in the embodiments of the present invention.

[0126] An embodiment of the present invention also provides a computer program product, including a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to execute the method described in the embodiments of the present invention.

[0127] Referring to Figure 5 , a block diagram of an electronic device that can be a server or a client according to an embodiment of the present invention will now be described. It is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described herein and / or claimed.

[0128] As Figure 5 shown, the electronic device includes a computing unit 501, which can execute various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the electronic device can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0129] Multiple components in the electronic device are connected to the I / O interface 505, including: an input unit 506, an output unit 507, a storage unit 508, and a communication unit 509. The input unit 506 can be any type of device capable of inputting information to the electronic device. The input unit 506 can receive input digital or character information and generate key signal inputs related to the user settings and / or function controls of the electronic device. The output unit 507 can be any type of device capable of presenting information and can include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 508 can include, but is not limited to, magnetic disks and optical discs. The communication unit 509 allows the electronic device to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks and can include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0130] The computing unit 501 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a CPU, a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 501 executes the various methods and processes described above. For example, in some embodiments, the method embodiments of the present invention can be implemented as a computer program tangibly contained in a machine-readable medium, such as the storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device via the ROM 502 and / or the communication unit 509. In some embodiments, the computing unit 501 can be configured to execute the above-described method in any other suitable manner (e.g., by means of firmware).

[0131] The computer program for implementing the method of the embodiments of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to the processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing devices, such that when the computer programs are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer programs can be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server.

[0132] In the context of embodiments of the present invention, a machine-readable medium may be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable signal medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0133] It should be noted that the term "including" and its variations used in the embodiments of the present invention are open-ended, that is, "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The modifications of "one" and "a plurality" mentioned in the embodiments of the present invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless clearly specified otherwise in the context, it should be understood as "one or more".

[0134] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the embodiments of the present invention are all information and data that have been authorized by the user or fully authorized by all parties. And the collection, use, and processing of the relevant data need to comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation entrances are provided for the user to choose to authorize or refuse.

[0135] The various steps described in the method embodiments provided by the embodiments of the present invention may be executed in a different order and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The protection scope of the present invention is not limited in this regard.

[0136] As used in this specification, the term "embodiment" means that the specific features, structures or characteristics described in connection with an embodiment may be included in at least one embodiment of the present invention. The phrase appears in various positions in the specification and does not necessarily mean the same embodiment, nor does it mean that it is independent or alternative to other embodiments and mutually exclusive. The various embodiments in this specification are described in a related manner, and the same or similar parts between the embodiments are cross-referred to. In particular, for embodiments of devices, equipment, and systems, since they are basically similar to embodiments of methods, the description is relatively simple, and for related parts, reference may be made to the partial description of the method embodiments.

[0137] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of patent protection. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the appended claims.

Claims

1. A household photovoltaic energy storage system energy scheduling method, characterized in that: include: Based on weather data and historical photovoltaic data, the photovoltaic power generation is obtained using a photovoltaic prediction model pre-built based on a convolutional neural network and a long short-term memory network; Based on weather data and historical load data, the load demand power is obtained using a load forecasting model pre-built based on a convolutional neural network and a long short-term memory network; Create a strategy matrix, perform power allocation simulation according to the strategy matrix, the photovoltaic power generation power and the load demand power, and iteratively optimize the simulation results through a particle swarm algorithm to obtain an optimal power allocation result; The creating a strategy matrix and performing a power allocation simulation according to the strategy matrix, the photovoltaic power generation power and the load demand power include: Create a corresponding strategy matrix based on power allocation requirements; Reading a photovoltaic power generation curve corresponding to the photovoltaic power generation power and a load demand power curve corresponding to the load demand power; Extracting the photovoltaic power generation power and the load demand power of each time step based on the photovoltaic power generation power curve and the load demand power curve; Calculate a battery power distribution curve according to the external charge and discharge power and the energy storage battery model parameters, wherein the battery power distribution curve includes a battery state change curve; Calculate the inverter power distribution curve corresponding to the inverter based on the photovoltaic power generation power and the load demand power of each time step in combination with the current inverter operation mode, wherein the inverter power distribution curve includes a charging power curve, a discharging power curve, a buying power curve and a selling power curve; The iterative optimization of the simulation results by the particle swarm algorithm to obtain the optimal power allocation result includes: generating a particle swarm according to the power allocation simulation result; Determine the control mode of the large load. If it is a fixed period mode, change the large load switch strategy of all particles to user settings. If it is a fixed duration mode, execute the next step. Randomly generate the initial velocity of each particle; The power distribution simulation is performed on each particle to obtain the electricity cost corresponding to each particle, and the control mode of the large load is determined. If it is a fixed period mode, the next step is executed; if it is a fixed duration mode, a penalty coefficient is assigned to the particle. Update the individual optimal electricity expenditure and strategy matrix of each particle, and update the global optimal at the same time; Calculate the distance and direction of each particle, and update the particle velocity based on the learning factor based on the distance and direction of each particle, so that each particle moves towards the individual and global optimal solution; Determine whether it converges or reaches the maximum number of iterations. If it converges or does not reach the maximum number of iterations, return to the step of randomly generating the initial velocity of each particle. If the maximum number of iterations is reached, output the global optimal solution matrix and the global optimal solution expenditure.

2. The energy dispatching method for household photovoltaic energy storage system according to claim 1, characterized in that: Based on weather data and historical photovoltaic data, the photovoltaic power generation power is obtained using the photovoltaic prediction model pre-built based on convolutional neural network and long short-term memory network, including: Get weather data and historical photovoltaic data; The weather data and the historical photovoltaic data are analyzed by the Pearson correlation analysis method to obtain weather-related characteristics and historical photovoltaic characteristics; Processing the weather-related features through the convolution layer of the photovoltaic prediction model to obtain photovoltaic key features; The photovoltaic key features are processed through the long short-term memory network layer of the photovoltaic prediction model to obtain photovoltaic power time series features; The photovoltaic power time series characteristics are fused with the historical photovoltaic characteristics and processed through the convolution layer of the photovoltaic prediction model to obtain the photovoltaic power generation power.

3. The energy dispatching method for household photovoltaic energy storage system according to claim 1, characterized in that: Based on weather data and historical load data, the load demand power is obtained using the load forecasting model pre-built based on convolutional neural networks and long short-term memory networks, including: Get historical load data; The weather data and the historical load data are analyzed respectively by a Pearson correlation analysis method to obtain weather-related characteristics and historical load characteristics; Processing the weather-related features and historical load features through the convolution layer of the load prediction model to obtain photovoltaic key features and load key features; The photovoltaic key features and the load key features are processed respectively by the long short-term memory network layer of the load prediction model to obtain photovoltaic power timing features and load power timing features; The photovoltaic power timing characteristics and the load power timing characteristics are combined to obtain the load demand power.

4. The energy dispatching method for household photovoltaic energy storage system according to claim 1, characterized in that: The inverter operation modes include energy storage priority mode, grid connection priority mode, self-generation and self-use mode, electricity purchase and energy storage mode and energy storage and electricity selling mode.

5. The energy dispatching method for household photovoltaic energy storage system according to claim 1, characterized in that: Update the individual optimal electricity expenditure and strategy matrix of each particle, and update the global optimal at the same time, including: For each particle, if its corresponding electricity expenditure on the day is less than its optimal electricity expenditure, it will be used as the individual optimal strategy matrix, and its corresponding electricity expenditure on the day will be used as the individual optimal electricity expenditure; For each particle, if its corresponding electricity expenditure on that day is less than the optimal electricity expenditure experienced by all particles, it will be used as the global optimal strategy matrix, and its corresponding electricity expenditure on that day will be used as the global optimal electricity expenditure.

6. A household photovoltaic energy storage system energy dispatching device, used to execute the steps of the household photovoltaic energy storage system energy dispatching method according to any one of claims 1 to 5, characterized in that: The device comprises: The photovoltaic prediction module is used to obtain photovoltaic power generation based on weather data and historical photovoltaic data using a photovoltaic prediction model pre-built based on a convolutional neural network and a long short-term memory network; A load forecasting module is used to obtain load demand power based on weather data and historical load data using a load forecasting model pre-built based on a convolutional neural network and a long short-term memory network; Simulation module of photovoltaic energy storage system, used to simulate power distribution according to the strategy matrix, photovoltaic power generation and load demand power; The strategy calculation module is used to iteratively optimize the simulation results through the particle swarm algorithm to obtain the optimal power allocation result.

7. An electronic device comprising: A processor, and a memory storing a program, wherein the program includes instructions, and when the instructions are executed by the processor, the processor executes the energy scheduling method for a household photovoltaic energy storage system according to any one of claims 1 to 5.

8. A non-transitory machine-readable medium storing computer instructions, characterized in that: The computer instructions are used to enable the computer to execute the energy scheduling method for a household photovoltaic energy storage system according to any one of claims 1 to 5.

Images