Photovoltaic power prediction method, device and equipment based on Beidou short message

Abstract

The present invention relates to the technical field of Internet of Things, and discloses a photovoltaic power generation prediction method, device and equipment based on Beidou short message. The method includes: obtaining weather environment data and photovoltaic power generation power information within a preset time; determining a plurality of historical photovoltaic power generation data according to the photovoltaic power generation power information; performing normalization processing on the weather environment data to obtain photovoltaic influence factor data; and predicting the photovoltaic power generation power through a preset photovoltaic power generation prediction model according to the photovoltaic influence factor data and the plurality of historical photovoltaic power generation data. Compared with the prior art, in which the photovoltaic power generation power is predicted only by modeling meteorological factors or physical components, resulting in low accuracy of photovoltaic power generation power prediction, in the present invention, the photovoltaic power generation power is predicted through a preset photovoltaic power generation prediction model according to the photovoltaic influence factor data and the plurality of historical photovoltaic power generation data, thereby improving the accuracy of photovoltaic power generation power prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and particularly to a photovoltaic power generation prediction method, device and equipment based on Beidou short message. Background Art

[0002] Due to the influence of various environmental factors, the output power of solar photovoltaic power generation exhibits strong randomness, volatility and uncertainty. For the internal inherent properties of a photovoltaic power station, the factors to be considered are very complex, and there are huge differences in different geographical environments, making many performance parameters not advisable in engineering practical applications. Therefore, in the prior art, only considering meteorological factors or physical component modeling results in low accuracy of photovoltaic power generation prediction.

[0003] The above content is only used to assist in understanding the technical solution of the present invention, and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main object of the present invention is to provide a photovoltaic power generation prediction method, device and equipment based on Beidou short message, aiming to solve the technical problem of how to improve the accuracy of photovoltaic power generation prediction.

[0005] To achieve the above object, the present invention provides a photovoltaic power generation prediction method based on Beidou short message, and the photovoltaic power generation prediction method based on Beidou short message includes:

[0006] Obtain weather environment data and photovoltaic power generation power information within a preset time;

[0007] Determine a plurality of historical photovoltaic power generation data according to the photovoltaic power generation power information;

[0008] Perform normalization processing on the weather environment data to obtain photovoltaic influence factor data;

[0009] Perform photovoltaic power generation prediction through a preset photovoltaic power generation prediction model according to the photovoltaic influence factor data and the plurality of historical photovoltaic power generation data.

[0010] Optionally, before the step of obtaining weather environment data and photovoltaic power generation power information within a preset time, it further includes:

[0011] Obtain a plurality of weather environment influence training data and a plurality of photovoltaic power generation training data according to photovoltaic power generation influence factors;

[0012] Perform normalization processing on the plurality of weather environment influence training data to obtain a plurality of weather environment influence sample data;

[0013] Preprocess multiple pieces of the photovoltaic power generation training data to obtain multiple pieces of photovoltaic power generation sample data;

[0014] Train an initial network model based on multiple pieces of the weather environment impact sample data and multiple pieces of the photovoltaic power generation sample data to obtain a preset photovoltaic power generation prediction model.

[0015] Optionally, the step of training an initial network model based on multiple pieces of the weather environment impact sample data and multiple pieces of the photovoltaic power generation sample data to obtain a preset photovoltaic power generation prediction model includes:

[0016] Generate a photovoltaic power generation power prediction vector based on multiple pieces of the weather environment impact sample data and multiple pieces of the photovoltaic power generation sample data;

[0017] Train the initial network model based on the photovoltaic power generation power prediction vector to obtain a preset photovoltaic power generation prediction model.

[0018] Optionally, the step of generating a photovoltaic power generation power prediction vector based on multiple pieces of the weather environment impact sample data and multiple pieces of the photovoltaic power generation sample data includes:

[0019] Analyze multiple pieces of the weather environment impact sample data and multiple pieces of the photovoltaic power generation sample data to obtain photovoltaic power generation impact feature information;

[0020] Generate a photovoltaic power generation power prediction vector based on the photovoltaic power generation impact feature information.

[0021] Optionally, the step of training the initial network model based on the photovoltaic power generation power prediction vector to obtain a preset photovoltaic power generation prediction model includes:

[0022] Integrate the photovoltaic power generation power prediction vector through a non-linear mapping strategy to obtain a photovoltaic power generation integration vector;

[0023] Perform dimensionality reduction processing on the photovoltaic power generation integration vector to obtain a photovoltaic power generation dimensionality reduction vector;

[0024] Train the initial network model based on the photovoltaic power generation dimensionality reduction vector to obtain a preset photovoltaic power generation prediction model.

[0025] Optionally, after the step of predicting the photovoltaic power generation power through a preset photovoltaic power generation prediction model based on the photovoltaic impact factor data and multiple pieces of the historical photovoltaic power generation data, it further includes:

[0026] Obtain the photovoltaic power generation power prediction information corresponding to the weather environment data and the data acquisition device;

[0027] Determine the energy consumption prediction information of the data acquisition device;

[0028] Determine a data collection mode based on the photovoltaic power generation prediction information and the energy consumption prediction information, so that the data collection device collects data according to the data collection mode.

[0029] Optionally, the step of determining the energy consumption prediction information of the data collection device includes:

[0030] Determine the terminal energy consumption information of the data collection device;

[0031] Input the terminal energy consumption information and the photovoltaic power generation prediction information into a preset energy consumption model to obtain the energy consumption prediction information of the data collection device.

[0032] Optionally, the step of determining multiple historical photovoltaic power generation data according to the photovoltaic power generation information includes:

[0033] Extract multiple original photovoltaic power generation data from the photovoltaic power generation information;

[0034] Generate a photovoltaic power generation curve graph based on the multiple original photovoltaic power generation data;

[0035] Judge whether the photovoltaic power generation curve graph meets the preset data conditions;

[0036] When the photovoltaic power generation curve graph does not meet the preset data conditions, process the multiple original photovoltaic power generation data according to the preset nearest neighbor completion rule to obtain multiple historical photovoltaic power generation data.

[0037] In addition, to achieve the above object, the present invention also proposes a photovoltaic power generation prediction device based on Beidou short message. The photovoltaic power generation prediction device based on Beidou short message includes:

[0038] An acquisition module, configured to acquire weather environment data and photovoltaic power generation information within a preset time;

[0039] A determination module, configured to determine multiple historical photovoltaic power generation data according to the photovoltaic power generation information;

[0040] A processing module, configured to perform normalization processing on the weather environment data to obtain photovoltaic influence factor data;

[0041] A prediction module, configured to perform photovoltaic power generation prediction through a preset photovoltaic power generation prediction model according to the photovoltaic influence factor data and the multiple historical photovoltaic power generation data.

[0042] In addition, to achieve the above object, the present invention further provides a photovoltaic power generation prediction device based on Beidou short message. The device includes: a memory, a processor, and a Beidou short message-based photovoltaic power generation prediction program stored on the memory and operable on the processor. The Beidou short message-based photovoltaic power generation prediction program is configured to implement the steps of the Beidou short message-based photovoltaic power generation prediction method as described above.

[0043] The present invention first obtains weather environment data and photovoltaic power generation information within a preset time, then determines a plurality of historical photovoltaic power generation data based on the photovoltaic power generation information, then normalizes the weather environment data to obtain photovoltaic influence factor data, and finally predicts the photovoltaic power generation power through a preset photovoltaic power generation prediction model according to the photovoltaic influence factor data and the plurality of historical photovoltaic power generation data. Compared with the prior art in which only meteorological factors or physical component modeling are used to predict the photovoltaic power generation power, resulting in low accuracy of photovoltaic power generation power prediction, in the present invention, the photovoltaic power generation power is predicted through a preset photovoltaic power generation prediction model according to the photovoltaic influence factor data and the plurality of historical photovoltaic power generation data, thereby improving the accuracy of photovoltaic power generation power prediction. Description of the Drawings

[0044] Figure 1 is a schematic structural diagram of a photovoltaic power generation prediction device based on Beidou short message in the hardware operating environment involved in the embodiment solution of the present invention;

[0045] Figure 2 is a schematic flowchart of the first embodiment of the Beidou short message-based photovoltaic power generation prediction method of the present invention;

[0046] Figure 3 is a schematic diagram of the principle of the preset power prediction model in the first embodiment of the Beidou short message-based photovoltaic power generation prediction method of the present invention;

[0047] Figure 4 is a flowchart of SSAE feature extraction in the first embodiment of the Beidou short message-based photovoltaic power generation prediction method of the present invention;

[0048] Figure 5 is a schematic flowchart of the second embodiment of the Beidou short message-based photovoltaic power generation prediction method of the present invention;

[0049] Figure 6 is a comparison chart of the prediction performance evaluation of three SSAE-BILSTM prediction models in the second embodiment of the Beidou short message-based photovoltaic power generation prediction method of the present invention;

[0050] Figure 7 is a structural block diagram of the first embodiment of the Beidou short message-based photovoltaic power generation prediction device of the present invention.

[0051] The realization, functional features, and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiments

[0052] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0053] Refer to Figure 1 , Figure 1 which is a schematic structural diagram of a photovoltaic power generation prediction device based on Beidou short message for the hardware operating environment involved in the embodiment solution of the present invention.

[0054] As Figure 1 shown, the photovoltaic power generation prediction device based on Beidou short message may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard). Optionally, the user interface 1003 may further include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless-fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM) or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0055] Those skilled in the art can understand that Figure 1 the structure shown in

[0056] does not constitute a limitation on the photovoltaic power generation prediction device based on Beidou short message, and may include more or fewer components than shown in the figure, or combine certain components, or arrange different components. Figure 1 As

[0057] shown, the memory 1005, as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and a photovoltaic power generation prediction program based on Beidou short message. Figure 1In the photovoltaic power generation prediction device based on Beidou short message shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the photovoltaic power generation prediction device based on Beidou short message of the present invention can be arranged in the photovoltaic power generation prediction device based on Beidou short message. The photovoltaic power generation prediction device based on Beidou short message calls the photovoltaic power generation prediction program stored in the memory 1005 through the processor 1001 and executes the photovoltaic power generation prediction method provided by the embodiment of the present invention.

[0058] An embodiment of the present invention provides a photovoltaic power generation prediction method based on Beidou short message, referring to Figure 2 , Figure 2 which is a schematic flowchart of the first embodiment of the photovoltaic power generation prediction method based on Beidou short message of the present invention.

[0059] In this embodiment, the photovoltaic power generation prediction method based on Beidou short message includes the following steps:

[0060] Step S10: Obtain weather environment data and photovoltaic power generation information within a preset time.

[0061] It is easy to understand that the execution subject of this embodiment can be a photovoltaic power generation prediction device based on Beidou short message with functions such as image processing, data processing, network communication, and program operation, or other computer devices with similar functions. This embodiment does not impose any restrictions. Among them, the photovoltaic power generation prediction device based on Beidou short message can adjust the terminal energy consumption information based on the energy consumption prediction information, etc. This embodiment does not impose any restrictions.

[0062] In this embodiment, communication can also be carried out in the form of Beidou short message during photovoltaic power generation prediction. The Beidou short message function is a unique function of the Beidou system. It can perform two-way communication using Beidou satellites without relying on mobile communication signals, and can provide all-weather and non-blind area services throughout the Asia-Pacific region. Its communication coverage is large and is not restricted by geographical location. Each Beidou short message communication device independently has an ID number, and a cryptographic protocol is used for encryption during communication. The information transmission needs to be forwarded through the ground center communication base station.

[0063] In a specific implementation, the specific process of Beidou short message communication is as follows: (1) The Beidou short message sender encrypts the communication request containing the recipient's ID number and content, and relays it through Beidou satellites to be received by the ground central service station; (2) After receiving the satellite communication application, the ground central service station adds the information to the content of the central station broadcast signal and sends it to the recipient by means of broadcast through communication satellites; (3) After receiving the signal from the central station, the Beidou short message recipient decrypts the telegram according to the protocol to obtain the sent information.

[0064] It should also be noted that Beidou short message communication has a very short transmission delay. The transmission time required in two-way Beidou short message communication is about 0.5 s, and the communication is carried out at the fastest frequency of 1 Hz. Beidou short message communication forwards information through satellites and is basically not affected by natural disasters on the ground. In addition, in remote areas or extreme terrains, the equipment and time costs of building wireless communication base stations or wired lines such as optical fibers are relatively high, while Beidou short messages can complete communication quickly and the equipment is inexpensive.

[0065] It can be understood that the preset time can be custom-set for users, such as 3 days or 16 h, etc.; the weather environment data is the real-time data of the solar energy center, and the weather environment data includes timestamp information, received active power information, current information, power information, wind speed information, temperature information, humidity information, global horizontal irradiance, diffuse horizontal irradiance, wind direction information, rainfall information, etc. The photovoltaic power generation power information includes multiple historical photovoltaic power generation power information, etc.

[0066] Step S20: Determine a plurality of historical photovoltaic power generation data according to the photovoltaic power generation power information.

[0067] It should also be noted that the historical photovoltaic power generation data can be the photovoltaic power generation data corresponding to a certain moment and a certain weather environment data before, etc.

[0068] Furthermore, in order to accurately obtain historical photovoltaic power generation data, the processing method of determining a plurality of historical photovoltaic power generation data according to the photovoltaic power generation power information can be to extract a plurality of original photovoltaic power generation data from the photovoltaic power generation power information, and then generate a photovoltaic power generation power curve graph according to the plurality of original photovoltaic power generation data, and judge whether the photovoltaic power generation power curve graph meets the preset data conditions. When the photovoltaic power generation power curve graph does not meet the preset data conditions, process the plurality of original photovoltaic power generation data according to the preset nearest neighbor completion rule to obtain a plurality of historical photovoltaic power generation data.

[0069] For the errors such as missing and abnormal values in the original data, the data needs to be completed and corrected. After the data is corrected, the original data recorded every 5 minutes is used to generate various records every hour according to the arithmetic mean. The methods for handling missing or abnormal data are as follows: (1) For missing values, the method of filling with 0 values is adopted; (2) For negative numbers, which indicate reverse voltage, the method of replacing with 0 can be adopted; (3) For numerical mutations, when a number is particularly large or small compared to the numbers before and after it, the K-nearest neighbor completion algorithm is used to complete the mutated value. The K-nearest neighbor completion algorithm obtains the average value of the k nearest neighbor data near the missing value, and replaces the missing value with this average value. The K-nearest neighbor completion algorithm is shown in the formula:

[0070]

[0071] In the formula, X i is the position of the missing value in the data sample, and x i-k is the k-th data before the missing value, and x i+k is the k-th data after the missing value.

[0072] Step S30: Normalize the weather environment data to obtain photovoltaic influence factor data.

[0073] It should also be noted that when predicting the photovoltaic power, the data involved mainly includes solar radiation intensity, temperature, humidity, and historical photovoltaic power generation data, and there are certain differences in the order of magnitude of each data item. Therefore, it is particularly important to normalize the original data. In the present invention, the photovoltaic influence factor data such as light intensity and temperature are uniformly normalized to the interval [0.1, 0.9]. The calculation expression is:

[0074]

[0075] It should be understood that the photovoltaic influence factor data can be understood as the data after normalization processing.

[0076] Step S40: Predict the photovoltaic power generation power through a preset photovoltaic power generation prediction model according to the photovoltaic influence factor data and multiple pieces of the historical photovoltaic power generation data.

[0077] Furthermore, in order to accurately predict the photovoltaic power generation power, it is also necessary to obtain multiple weather environment influence training data and multiple photovoltaic power generation training data according to the photovoltaic power generation power influence factors, then normalize the multiple weather environment influence training data to obtain multiple weather environment influence sample numbers, then preprocess the multiple photovoltaic power generation training data to obtain multiple photovoltaic power generation sample numbers, and train the initial network model according to the multiple weather environment influence sample data and multiple photovoltaic power generation sample data to obtain a preset photovoltaic power generation prediction model.

[0078] In this embodiment, multiple weather environment impact sample data and multiple photovoltaic power generation sample data can also be analyzed to obtain photovoltaic power generation impact characteristic information. According to the photovoltaic power generation impact characteristic information, a photovoltaic power generation prediction vector is generated. Then, through a non-linear mapping strategy, the photovoltaic power generation prediction vector is integrated to obtain a photovoltaic power generation integration vector. The photovoltaic power generation integration vector is subjected to dimensionality reduction processing to obtain a photovoltaic power generation dimensionality reduction vector. The initial network model is trained according to the photovoltaic power generation dimensionality reduction vector to obtain a preset photovoltaic power generation prediction model.

[0079] It should be noted that the environmental impact factors corresponding to multiple weather environment impact training data include meteorological variables and time characteristic variables. Among them, the meteorological variables include global horizontal irradiance, temperature, wind speed, wind direction, air humidity, rainfall, and diffuse horizontal irradiance; the time characteristic variables include season, month, day, and hour within a day. Then, corresponding multiple weather environment impact training data need to be obtained according to the meteorological variables and time characteristic variables.

[0080] In this embodiment, in order to obtain a preset power prediction model for accurately predicting the photovoltaic power generation, it is necessary to construct a preset power prediction model based on SSAE feature learning and BILSTM network. The meteorological and time factors affecting the photovoltaic power generation can be used as inputs. The SSAE network automatically compresses the data from the input sequence and extracts low-dimensional abstract features, and then the abstract features are used as the inputs of the BILSTM network to achieve hourly prediction of the photovoltaic power generation.

[0081] Furthermore, for the convenience of understanding the composition process of the model, refer to Figure 3 , Figure 3This is the schematic diagram of the preset power prediction model for the first embodiment of the photovoltaic power generation prediction method based on Beidou short message. The model mainly consists of four parts: data input layer, SSAE feature extraction layer, BILSTM data prediction layer, and output layer. Input data layer: Select various factors that affect photovoltaic power generation, perform attribute expansion on the original data sequence of photovoltaic power generation, and input the extended sequence {At-N,..., At-2, At-1} of N known historical moments in chronological order into the SSAE feature extraction layer. SSAE feature extraction layer: Use the SSAE network unit to extract low-dimensional abstract features from the data at each moment in the extended sequence {At-N,..., At-2, At-1}. This process will compress the data to obtain the feature sequence {Bt-P,..., Bt-2, Bt-1}, where P < N. BILSTM data prediction layer: Input the abstract features {Bt-P,..., Bt-2, Bt-1} extracted by the SSAE network unit into the BILSTM memory unit for model training to obtain the photovoltaic power generation prediction vector. Data output layer: Decode the prediction vector output by the BILSTM model with a fully connected neural network to output the photovoltaic power generation prediction value sequence.

[0082] It should also be understood that the various factors affecting photovoltaic power generation in the input layer data are meteorological variables and time feature variables respectively. Among them, the meteorological variables include global horizontal irradiance, temperature, wind speed, wind direction, air humidity, rainfall, and diffuse horizontal irradiance, a total of 7 variables; the time feature variables include season, month, day, and intra-day hour moment, a total of 4 variables.

[0083] It should also be noted that in order to be able to understand the SSAE feature learning layer in detail, refer to Figure 4 , Figure 4 This is the SSAE feature extraction flowchart for the first embodiment of the photovoltaic power generation prediction method based on Beidou short message. In the figure, the data in the dataset is used as the input of the SSAE feature extraction layer after preprocessing and normalization, and an unsupervised method is used to extract the abstract features of the input data, so as to realize the compression and dimensionality reduction of the features. SSAE can automatically extract the low-dimensional abstract influence features of photovoltaic power generation, etc.

[0084] Furthermore, based on the excellent characteristics of the BILSTM, a BILSTM prediction layer is added in the present invention to establish a time series prediction model for photovoltaic power generation. The BILSTM prediction layer takes the low-dimensional abstract features extracted by the SSAE layer as input, further learns the long-term characteristics existing in the historical data, and outputs a predicted vector. It is known from relevant research that the photovoltaic power generation has strong autocorrelation. Through statistical analysis of the existing photovoltaic power generation data, it is found that the photovoltaic power generation at the current moment has a strong correlation not only with the historical data of the 6 days closest to this moment, that is, the historical data of 72 moments. Therefore, the input time series step size N of the BILSTM prediction layer in the present invention is set to 72, and the low-dimensional abstract influence features of 72 moments such as t-72, t-71... t-1 after SSAE feature learning are used as the input data of the BILSTM prediction layer, and the predicted vectors of the photovoltaic power generation for the next 24 moments (2 days) are output, etc.

[0085] It should also be noted that the SSAE feature extraction layer and the BILSTM prediction layer have completed the most important feature extraction process and time series prediction process in the SSAE-BILSTM prediction model. The data output layer is a fully connected network, which takes the predicted vector output by the prediction layer as input, and realizes the integration and dimensionality reduction of the predicted vector through the non-linear mapping function of the fully connected network, and then outputs the predicted value of the photovoltaic power generation. In the output layer, the calculation formula of the network node is as follows:

[0086]

[0087] In the formula, is the value of node j in the kth layer, W ij is the connection weight from node i in the (k-1)th layer to node j in the kth layer, S k-1 is the number of nodes in the (k-1)th layer, f(.) is the activation function, and the Relu activation function is adopted in this paper. is the bias.

[0088] It can be understood that there is no fixed algorithm for selecting the number of nodes in the hidden layer of the neural network. Therefore, 3 SSAE-BILSTM models with different network structures are designed, and the SSAE-BILSTM model with higher prediction accuracy (i.e., SSAE3-BILSTM) is selected from the 3 different network structures as the preset power prediction model.

[0089] In specific implementation, it is also necessary to obtain the photovoltaic power generation prediction information and data acquisition equipment corresponding to the weather environment data, then determine the terminal energy consumption information of the data acquisition equipment, input the terminal energy consumption information and the photovoltaic power generation prediction information into the preset energy consumption model to obtain the energy consumption prediction information of the data acquisition equipment, and determine the data acquisition mode based on the photovoltaic power generation prediction information and the energy consumption prediction information, so that the data acquisition equipment can perform data acquisition according to the data acquisition mode.

[0090] Further, in order to obtain an accurate preset energy consumption model, it is necessary to obtain multiple terminal device identifiers in different environments. Then, when the multiple terminal device identifiers meet the preset identifier conditions, a sensing device, a control device, and a communication device are determined according to the multiple terminal device identifiers. The total sensing energy consumption information corresponding to the sensing device, the total control energy consumption information corresponding to the control device, and the total communication energy consumption information corresponding to the communication device are determined. Finally, a preset energy consumption model is constructed according to the total sensing energy consumption information, the total control energy consumption information, and the total communication energy consumption information.

[0091] Further, in order to accurately obtain the total sensing energy consumption information, the processing method for determining the total sensing energy consumption information corresponding to the sensing device includes obtaining the sensing sleep power consumption information and the sensing sleep duration corresponding to the sensing device, determining the sensing sleep energy consumption information according to the sensing sleep power consumption information and the sensing sleep duration, obtaining the sensing system wake-up energy consumption information, the sensing data measurement energy consumption information, the sensing data processing energy consumption information, and the sensing data transmission energy consumption information corresponding to the sensing device, determining the sensing working energy consumption information according to the sensing system wake-up energy consumption information, the sensing data measurement energy consumption information, the sensing data processing energy consumption information, and the sensing data transmission energy consumption information, and determining the total sensing energy consumption information according to the sensing sleep energy consumption information and the sensing working energy consumption information.

[0092] In this embodiment, to accurately analyze the energy consumption of the sensor, first, different working states of the sensor are divided. Then, the energy consumption of each state is calculated according to the characteristics of different working states. Finally, a complete energy consumption model of the sensor node is obtained. The working states of the sensor node are now divided into the following five items: (1) Sleep state: The time period required for the sensor to measure physical quantities. Therefore, it can be in a low-power mode for most of the acquisition cycle and enter the sleep state; (2) System wake-up: The sensor system enters the working state from the sleep state; (3) Measuring data: The sensor converts the physical quantity of environmental data into a digital quantity; (4) Processing data: The sensor performs preliminary processing on the obtained data volume for data transmission; (5) Data transmission: The sensor transmits the data to the controller, completes the data measurement, and then enters the sleep state.

[0093] In specific implementation, in environmental data acquisition, the time for the sensor to collect environmental data is very short, and it is in the sleep state for most of the time. The energy consumption of the sensor in the sleep state is relatively low, but when calculating the energy consumption in one cycle, the energy consumption in the long-term sleep state cannot be ignored. Therefore, the energy consumption in the sleep state must be considered. According to the division of the working states of the sensor, the total energy consumption E of the sensor node working in one cycle Total1 =E Sleep +E Active , where ESleep is the energy consumption of the node in the sleep mode (i.e., sensing sleep energy consumption information), E Active is the total energy consumption of the sensor in the working state. E Sleep = P Sleep * T Sleep , P Sleep is the power consumption in the sleep mode (i.e., sensing sleep power consumption information), T Sleep is the duration (i.e., sensing sleep duration); E Active = E wark + E measure + E process + E transport , E wark is the sensing system wake-up energy consumption information, E measure is the sensing data measurement energy consumption information, E process is the sensing data processing energy consumption information, E transport is the sensing data transmission energy consumption information. Among them, E wark , E measure , E process , E transport can be calculated using the formula E = P(f MCU ) * T. In the formula, P(f MCU ) is the corresponding power consumption (depending on the frequency f MCU ) of the microcontroller, and T is the duration.

[0094] Furthermore, in order to accurately obtain the total control energy consumption information, it is necessary to obtain the sleep power consumption information and sleep duration corresponding to the control device, determine the sleep energy consumption information of the control device according to the sleep power consumption information and sleep duration, obtain the system wake-up power information and system wake-up duration corresponding to the control device, determine the system wake-up energy consumption information according to the system wake-up power information and system wake-up duration, obtain the sending instruction energy consumption information, waiting for feedback energy consumption information, and data processing energy consumption information corresponding to the control device, determine the working energy consumption information according to the instruction energy consumption information, waiting for feedback energy consumption information, and data processing energy consumption information, and determine the total control energy consumption information of the control device according to the sleep energy consumption information, system wake-up energy consumption information, and working energy consumption information.

[0095] In this embodiment, the controller coordinates and commands the operation of the entire system, performs calculations under a given program, and controls the system and its components to work properly under the logic that conforms to the decision-making. In the environmental data acquisition terminal, the controller is responsible for receiving, processing, packing all sensor data, and sending the data to the server through the communication module. The working state of the acquisition terminal controller can be divided into the following three: (1) System wake-up: When the sleep time ends, the system wakes up and starts to work. (2) Working state: The MCU is responsible for the normal operation of the entire system's functions. It performs logical control through calculations and controls the operation of peripheral devices by sending instructions. ① Sending instructions. The MCU sends instructions or requests to the peripheral devices, causing the peripheral devices to start working; ② Waiting for feedback. After sending the instructions, the MCU enters the waiting state and waits for the response of the peripheral devices; ③ Data processing. After receiving the feedback or data from the peripheral devices, the MCU performs data processing and then sends instructions to the peripheral devices or enters the sleep state. (3) Sleep state: The system enables the low-frequency model, enters the sleep mode, and reduces the energy consumption of the system operation.

[0096] In a specific implementation, the total energy consumption E of the controller for one working cycle Total2 =E Wake +E Work +E Sleep where E Wake is the energy consumption information for system wake-up, E Work is the energy consumption information for working, and E Sleep is the energy consumption information for sleep. E Wake =P Wake *T Wake where P Wake is the system wake-up power information and T Wake is the duration of system wake-up; E Sleep =P Sleep (f MCU )*T Sleep where P Sleep (f MCU ) is the power of the controller in the system sleep state (depending on the frequency (f MCU )) of the microcontroller, and T Sleep is the duration of the controller in the system sleep state; E Work =E send +E Wait +E proc where E send is the energy consumption of the MCU for sending instructions in the working state, E Wait is the energy consumption of the MCU for waiting for feedback in the working state, and E proc is the energy consumption of the MCU for data processing in the working state. When the MCU acquires data from multiple sensors, the above three working time sequences overlap, and it is impossible to calculate the time of each time sequence separately. E Woke=P Woke *T Woke ,P Woke is the power of the controller in the working state, T Woke is the duration of the controller in the working state.

[0097] Furthermore, in order to accurately obtain the total communication energy consumption information of the communication device, it is necessary to obtain the communication transmission signal energy consumption information, communication reception signal energy consumption information, and communication standby energy consumption information corresponding to the communication device; determine the total communication energy consumption information corresponding to the communication device according to the communication transmission signal energy consumption information, communication reception signal energy consumption information, and communication standby energy consumption information.

[0098] In specific implementation, for the communication blind area in remote rural areas, Beidou short message is used for communication. The Beidou short message communication module adopts a transceiver antenna design. Since when transmitting Beidou short messages, the signal needs to be sent to the Beidou satellite 35786 kilometers away from the earth's surface, the energy consumption required for transmission is relatively large, while the energy consumption required for the Beidou short message communication module to receive signals is relatively small. The energy consumption of the Beidou short message communication module (i.e., the communication device) consists of communication transmission signal energy consumption information, communication reception signal energy consumption information, and communication standby energy consumption information. E Total3 =E s +E r +E t ,E Total3 is the total communication energy consumption information, E s is the communication transmission signal energy consumption information, E r is the communication reception signal energy consumption information, E t is the communication standby energy consumption information. The Beidou short message communication module sends lbit data to the satellite, and the distance is d, E s =E el l+ , is the energy consumption of the signal amplification circuit for sending or receiving data, is the energy consumption of signal transmission, E el is the energy consumption per unit data of the signal transmitting or receiving circuit module for sending or receiving. E r =E el l,E r is the communication reception signal energy consumption information, E t =P(f MCU )*T, where P(f MCU ) is the corresponding power consumption (depending on the frequency f of the microcontroller MCU ), and T is the duration.

[0099] It should be noted that a sensor energy consumption model is constructed based on the total sensor energy consumption information, a controller energy consumption model is constructed based on the total controller energy consumption information, and a communication module energy consumption model is constructed based on the total communication energy consumption information. The sensor energy consumption model, the controller energy consumption model, and the communication module energy consumption model are combined to obtain a preset energy consumption model.

[0100] In this embodiment, the currently widely used fixed sampling interval for agricultural environment data terminals cannot respond to environmental changes. For example, when the solar power generation is large and the power supply is sufficient, the sampling interval cannot be reduced to improve the sampling quality. Similarly, especially when the solar power generation is small and the power supply is insufficient, the sampling interval cannot be increased to reduce power consumption and keep the device online for a longer time. To address the above problems, in the present invention, an intelligent acquisition mode that changes according to the solar power generation situation is proposed based on the preset energy consumption model and the preset photovoltaic power generation prediction model. The intelligent acquisition mode calculates the power generation in a future period based on the predicted power generation, adjusts the sampling interval of the terminal, so that the power consumption of the terminal is less than the power generation in the future period, and keeps the device online for a longer time.

[0101] The present invention first obtains weather environment data and photovoltaic power generation information within a preset time, then determines multiple historical photovoltaic power generation data based on the photovoltaic power generation information, then normalizes the weather environment data to obtain photovoltaic impact factor data, and finally predicts the photovoltaic power generation based on the photovoltaic impact factor data and multiple historical photovoltaic power generation data through a preset photovoltaic power generation prediction model. Compared with the prior art where the photovoltaic power generation is predicted only by modeling with meteorological factors or physical components, resulting in a low accuracy of photovoltaic power generation prediction, in the present invention, the photovoltaic power generation is predicted based on the photovoltaic impact factor data and multiple historical photovoltaic power generation data through a preset photovoltaic power generation prediction model, thereby improving the accuracy of photovoltaic power generation prediction.

[0102] Reference Figure 5 , Figure 5 is a schematic flow chart of the second embodiment of the photovoltaic power generation prediction method based on Beidou short messages of the present invention.

[0103] Based on the above first embodiment, in this embodiment, before step S10, it further includes:

[0104] Step S01: Obtain multiple weather environment impact training data and multiple photovoltaic power generation training data according to the photovoltaic power generation impact factors.

[0105] It should be noted that the environmental impact factors corresponding to multiple weather environment impact training data include meteorological variables and time characteristic variables. Among them, the meteorological variables include global horizontal irradiance, temperature, wind speed, wind direction, air humidity, rainfall, and diffuse horizontal irradiance; the time characteristic variables include season, month, day, and hour within a day. Then, it is necessary to obtain multiple weather environment impact training data corresponding to the meteorological variables and time characteristic variables.

[0106] Step S02: Normalize the multiple weather environment impact training data to obtain multiple weather environment impact sample data.

[0107] It should also be noted that since multiple weather environment impact training data include solar radiation intensity, temperature, humidity, and historical photovoltaic power generation data, there are certain differences in the order of magnitude of each data item. It is necessary to uniformly normalize the data of photovoltaic impact factors such as light intensity and temperature to obtain weather environment impact sample data.

[0108] Step S03: Preprocess the multiple photovoltaic power generation training data to obtain multiple photovoltaic power generation sample data.

[0109] It should also be noted that multiple photovoltaic power generation training data can be photovoltaic power generation training data corresponding to a certain moment and a certain weather environment data before.

[0110] Furthermore, in order to accurately obtain photovoltaic power generation sample data, a photovoltaic power curve graph can be generated according to the multiple photovoltaic power generation training data, and it is judged whether the photovoltaic power curve graph meets the preset data conditions. When the photovoltaic power curve graph does not meet the preset data conditions, the multiple photovoltaic power generation training data are processed according to the preset nearest neighbor completion rule to obtain multiple photovoltaic power generation sample data.

[0111] In this embodiment, the environmental data and equipment operation parameters of the photovoltaic power generation station can be recorded by the solar energy center every 5 minutes. Assume that the data from November 20, 2019 to November 19, 2020 of this solar panel are used as the training data set. The data from November 20, 2020 to December 2, 2020 are used as the test data set. The training data set and the test data set are respectively composed of weather variables and PV power output time series of 365 days and 378 days. The data items included in the data set are shown in Table 1 below.

[0112] Table 1

[0113]

[0114] Suppose there are many gaps in the original data collected according to the data items. Through comparison and analysis, it is found that the missing values are 0 values. And the effective moments of the photovoltaic luminous power are from 7:00 to 18:00, a total of 12 moments. At other moments, the photovoltaic luminous power is 0 or very small and can be ignored. Refer to Figure 3 , Figure 3 is the photovoltaic power generation curve diagram of the first embodiment of the photovoltaic power generation prediction method based on Beidou short message of the present invention. In the figure, the power generation power from 7:00 to 18:00 in a day is positive, and the power generation power at other times is zero. At the same time, there are anomalies with negative power and anomalies with numerical mutations. For the errors such as missing and abnormal in the original data, the data needs to be completed and corrected. After the data is corrected, the original data recorded once every 5 minutes is averaged arithmetically to generate each hourly record. The methods for dealing with missing or abnormal data are: (1) For missing values, adopt the method of filling with 0 values; (2) For negative numbers, which are caused by reverse voltage, they can be replaced with 0; (3) For numerical mutations, when a number is particularly large or particularly small compared with the previous and subsequent numbers, use the K-nearest neighbor completion algorithm to complete the mutated value. The K-nearest neighbor completion algorithm obtains the average value of these k data according to the k nearest neighbor data near the missing value, and replaces the missing value with this average value. The K-nearest neighbor completion algorithm is as shown in the formula:

[0115]

[0116] In the formula, X i is the position of the missing value of the data sample, and x i-k is the kth data before the missing value, and x i+k is the kth data after the missing value.

[0117] Step S04: Train the initial network model according to the multiple weather environment influence sample data and the multiple photovoltaic power generation sample data to obtain a preset photovoltaic power generation prediction model.

[0118] Furthermore, in order to accurately predict the photovoltaic power generation power, it is also necessary to obtain multiple weather environment influence training data and multiple photovoltaic power generation training data according to the photovoltaic power generation power influence factors, then normalize the multiple weather environment influence training data to obtain multiple weather environment influence sample data, and then preprocess the multiple photovoltaic power generation training data to obtain multiple photovoltaic power generation sample data, and train the initial network model according to the multiple weather environment influence sample data and the multiple photovoltaic power generation sample data to obtain a preset photovoltaic power generation prediction model.

[0119] In this embodiment, multiple weather environment impact sample data and multiple photovoltaic power generation sample data can also be analyzed to obtain photovoltaic power generation impact characteristic information. According to the photovoltaic power generation impact characteristic information, a photovoltaic power generation prediction vector is generated. Then, through a non-linear mapping strategy, the photovoltaic power generation prediction vector is integrated to obtain a photovoltaic power generation integration vector. The photovoltaic power generation integration vector is subjected to dimensionality reduction processing to obtain a photovoltaic power generation dimensionality reduction vector. The initial network model is trained according to the photovoltaic power generation dimensionality reduction vector to obtain a preset photovoltaic power generation prediction model.

[0120] In a specific implementation, in order to obtain a preset power prediction model for accurately predicting the photovoltaic power generation, it is necessary to construct a preset power prediction model based on SSAE feature learning and BILSTM network. The meteorological and time factors affecting the photovoltaic power generation can be used as inputs. The SSAE network automatically compresses the data from the input sequence and extracts low-dimensional abstract features, and then uses the abstract features as the input of the BILSTM network to achieve hourly prediction of the photovoltaic power generation.

[0121] Furthermore, for the convenience of understanding the composition process of the model, refer to Figure 4 , Figure 4 FIG. is the schematic diagram of the preset power prediction model of the first embodiment of the photovoltaic power generation prediction method based on Beidou short message of the present invention. The model mainly consists of four parts: a data input layer, an SSAE feature extraction layer, a BILSTM data prediction layer, and an output layer. Input data layer: Select various factors affecting the photovoltaic power generation, perform attribute expansion on the original data sequence of the photovoltaic power generation, and input the extended sequence {At-N,..., At-2, At-1} of N known historical moments in chronological order into the SSAE feature extraction layer. SSAE feature extraction layer: Use the SSAE network unit to extract low-dimensional abstract features from the data at each moment in the extended sequence {At-N,..., At-2, At-1}. This process compresses the data to obtain a feature sequence {Bt-P,..., Bt-2, Bt-1}, where P < N. BILSTM data prediction layer: Input the abstract features {Bt-P,..., Bt-2, Bt-1} extracted by the SSAE network unit into the BILSTM memory unit for model training to obtain a photovoltaic power generation prediction vector. Data output layer: Decode the prediction vector output by the BILSTM model with a fully connected neural network to output a sequence of photovoltaic power generation prediction values.

[0122] It should also be understood that the various factors affecting the photovoltaic power generation in the input layer data are respectively meteorological variables and time feature variables. Among them, the meteorological variables include global horizontal irradiance, temperature, wind speed, wind direction, air humidity, rainfall, and diffuse horizontal irradiance, a total of 7 variables; the time feature variables include season, month, day, and hour moment within a day, a total of 4 variables.

[0123] It should also be noted that, in order to understand the SSAE feature learning layer in detail, refer to Figure 4 , Figure 4 which is the SSAE feature extraction flowchart of the first embodiment of the photovoltaic power prediction method based on Beidou short message in the present invention. In the figure, the data in the dataset is preprocessed and normalized and then used as the input of the SSAE feature extraction layer, and an unsupervised method is used to extract the abstract features of the input data, so as to realize the compression and dimensionality reduction of the features. The SSAE can automatically extract the low-dimensional abstract influence features of photovoltaic power generation, etc.

[0124] Furthermore, based on the excellent characteristics of BILSTM, a BILSTM prediction layer is added in the present invention to establish a photovoltaic power time series prediction model. The BILSTM prediction layer takes the low-dimensional abstract features extracted by the SSAE layer as the input, further learns the long-term characteristics existing in the historical data, and outputs the predicted vector. It can be known from relevant research that the photovoltaic power has strong autocorrelation. Through the statistical analysis of the existing photovoltaic power data, it is found that the photovoltaic power at the current moment has a strong correlation not only with the historical data of the 6 days closest to this moment, that is, the historical data of 72 moments. Therefore, the input time series step length N of the BILSTM prediction layer in the present invention is set to 72, and the low-dimensional abstract influence features of 72 moments such as t-72, t-71... t-1 after SSAE feature learning are used as the input data of the BILSTM prediction layer, and the predicted vector of the photovoltaic power for the next 24 moments (2 days) is output, etc.

[0125] It should also be noted that the SSAE feature extraction layer and the BILSTM prediction layer have completed the most important feature extraction process and time series prediction process in the SSAE-BILSTM prediction model. The data output layer is a fully connected network, which takes the predicted vector output by the prediction layer as the input, and realizes the integration and dimensionality reduction of the predicted vector through the nonlinear mapping function of the fully connected network, and then outputs the predicted value of the photovoltaic power. In the output layer, the network node calculation formula is as follows:

[0126]

[0127] In the formula, is the value of node j in the kth layer, W ij is the connection weight from node i in the (k-1)th layer to node j in the kth layer, S k-1 is the number of nodes in the (k-1)th layer, f(.) is the activation function, and the Relu activation function is adopted in this article, is the bias.

[0128] It is understandable that there is no fixed algorithm for selecting the number of hidden layer nodes in a neural network. Therefore, three SSAE-BILSTM models with different network structures were designed, and the SSAE-BILSTM model with higher prediction accuracy (i.e., SSAE3-BILSTM) was selected from the three different network structures as the preset power prediction model.

[0129] In this embodiment, a 3-layer SSAE network (ssae_2, ssae_3, ssae_4), a 1-layer BILSTM network, and a 1-layer fully connected network (fc) can be designed. Then, the SSAE networks with different numbers of layers are combined with the BILSTM network to obtain a total of three SSAE-BILSTM models. The specific structures of the three SSAE-BILSTM network models are shown in Table 1, where the numbers in the table represent the number of neurons in the corresponding network layer, and "none" indicates that there is no such layer in the model.

[0130] Table 1

[0131]

[0132] It should also be noted that when predicting the photovoltaic power, the data involved mainly include solar radiation intensity, temperature, humidity, and historical photovoltaic power generation data, and there are certain differences in the order of magnitude of each data item. The data of photovoltaic influence factors such as light intensity and temperature need to be uniformly normalized. Then, the normalized data of photovoltaic influence factors and the multiple historical photovoltaic power generation data are used to predict the photovoltaic power generation through the preset photovoltaic power generation prediction model to obtain the photovoltaic power generation prediction information, where the photovoltaic power generation prediction information can be the power generation in a future period of time, etc.

[0133] It should also be noted that after predicting the photovoltaic power through the model, the root mean square error RMSE and the mean absolute error MAE are usually used as two indicators to quantitatively analyze the rationality of the prediction. The preprocessed sample set SSAE-BILSTM model is trained. In the network of the model, there are a large number of hidden layer nodes and weights to be trained. If the gradient descent method is directly used to train the model, problems such as gradient dispersion or gradient disappearance are likely to occur, making the model unable to converge. To avoid the above problems, in this paper, the SSAE network is pre-trained layer by layer in an unsupervised manner and the SSAE-BILSTM model is trained in a supervised manner. When pre-training the SSAE network layer by layer, the parameters of the SSAE network model are initialized, and the steps are as follows: (1) The first-layer encoder SAE1 in the SSAE network is trained with the loss function as the objective function to generate the weight matrix w1 and the hidden layer vector h1; (2) h1 is used as the input of the second-layer encoder SAE2 and trained in the same way to generate the weight matrix w2 and the hidden layer vector h2; (3) And so on, each subsequent layer of SAE is trained layer by layer, and finally a set of weights w1, w2, w3,... wn are obtained.

[0134] It should be understood that when training the SSAE-BILSTM model, first set the weights w1, w2, w3,... wn pre-trained by the SSAE as the initial parameters of the SSAE network in the SSAE-BILSTM model, and then use the gradient descent method to learn the overall weights of the SSAE-BILSTM model. Let the predicted photovoltaic power sequence output by the SSAE-BILSTM in each batch (batch) be Y = { }, and the measured photovoltaic power data sequence is , where n is the number of samples in each batch, then the objective function for training the SSAE-BILSTM model is:

[0135]

[0136] The appropriate parameters of the SSAE are crucial for the effectiveness of feature extraction, especially the number of hidden neurons and the choice of activation function. For the SSAE, the size cannot be reduced prematurely because it may lead to a large reconstruction error, indicating poor feature representation ability. On the contrary, the SSAE with more hidden neurons always produces a lower reconstruction error. However, from the perspective of feature extraction, high-dimensional features are meaningless. Therefore, the different sizes of the final features are inestimable because the reconstruction error and the size of the hidden representation should be considered comprehensively.

[0137] In this embodiment, referring to Figure 6 , Figure 6This is the prediction performance evaluation and comparison chart of the second embodiment of the photovoltaic power generation prediction method based on Beidou short message in the present invention. The prediction accuracy of different structured SSAE-BILSTM models on the test set is shown in the figure. It can be seen that the SSAE3-BILSTM model performs the best, which is reflected in that the root mean square error (RMSE) and mean absolute error (MAE) of the SSAE3-BILSTM model are the lowest. The RMSE value of the SSAE3-BILSTM model has dropped to 16.98, which is 2.36 lower than that of the SSAE2-BILSTM model and 3.14 lower than that of the SSAE4-BILSTM model; the MAE has dropped to 13.73, which is 1.08 lower than that of the SSAE2-BILSTM model and 1.50 lower than that of the SSAE4-BILSTM model, indicating that the SSAE3-BILSTM composed of a three-layer SSAE encoder and a BILSTM network has higher prediction accuracy.

[0138] It can be known through experiments that the SSAE3-BILSTM has higher prediction accuracy. The difference compared with the other two models is only the number of layers of the SSAE network. If the number of layers of the SSAE network is small, it will make the model have weak learning ability and low fault tolerance; while when the number of network layers is too high, overfitting is likely to occur, reducing the generalization ability of the model. Through experimental verification, the SSAE3-BILSTM model with a moderate number of network layers has better prediction performance and robustness.

[0139] In this embodiment, multiple weather environment impact training data and multiple photovoltaic power generation training data are obtained according to the influencing factors of photovoltaic power generation. Then, the multiple weather environment impact training data are normalized to obtain multiple weather environment impact sample numbers. After that, the multiple photovoltaic power generation training data are preprocessed to obtain multiple photovoltaic power generation sample numbers, and the initial network model is trained according to the multiple weather environment impact sample data and multiple photovoltaic power generation sample data to obtain a preset photovoltaic power generation prediction model. Compared with the prior art, only modeling through meteorological factors or physical components results in a low-precision model. In this embodiment, the initial network model is trained with multiple weather environment impact sample data and multiple photovoltaic power generation sample data to obtain a preset photovoltaic power generation prediction model, thereby improving the precision of the preset photovoltaic power generation prediction model.

[0140] Refer to Figure 7 , Figure 7 This is the structural block diagram of the first embodiment of the photovoltaic power generation prediction device based on Beidou short message in the present invention.

[0141] As Figure 7 shown, the photovoltaic power generation prediction device based on Beidou short message proposed in the embodiment of the present invention includes:

[0142] An acquisition module 7001 for acquiring weather environment data and photovoltaic power generation information within a preset time;

[0143] A determination module 7002 for determining a plurality of historical photovoltaic power generation data according to the photovoltaic power generation information;

[0144] A processing module 7003 for performing normalization processing on the weather environment data to obtain photovoltaic influence factor data;

[0145] A prediction module 7004 for predicting the photovoltaic power generation according to the photovoltaic influence factor data and a plurality of the historical photovoltaic power generation data through a preset photovoltaic power generation prediction model.

[0146] In this embodiment, first, weather environment data and photovoltaic power generation information within a preset time are acquired, then a plurality of historical photovoltaic power generation data are determined according to the photovoltaic power generation information, then the weather environment data are subjected to normalization processing to obtain photovoltaic influence factor data, and finally, the photovoltaic power generation is predicted according to the photovoltaic influence factor data and a plurality of historical photovoltaic power generation data through a preset photovoltaic power generation prediction model. Compared with the prior art in which the photovoltaic power generation is predicted only by modeling with meteorological factors or physical components, resulting in a low accuracy of photovoltaic power generation prediction, in this embodiment, the photovoltaic power generation is predicted according to the photovoltaic influence factor data and a plurality of historical photovoltaic power generation data through a preset photovoltaic power generation prediction model, thereby improving the accuracy of photovoltaic power generation prediction.

[0147] Further, the photovoltaic power generation prediction device based on Beidou short message also includes:

[0148] A building module for acquiring a plurality of weather environment influence training data and a plurality of photovoltaic power generation training data according to the photovoltaic power generation influence factors;

[0149] The building module is further configured to perform normalization processing on a plurality of the weather environment influence training data to obtain a plurality of weather environment influence sample data;

[0150] The building module is further configured to perform preprocessing on a plurality of the photovoltaic power generation training data to obtain a plurality of photovoltaic power generation sample data;

[0151] The building module is further configured to train an initial network model according to a plurality of the weather environment influence sample data and a plurality of the photovoltaic power generation sample data to obtain a preset photovoltaic power generation prediction model.

[0152] Further, the building module is further configured to generate a photovoltaic power generation prediction vector according to a plurality of the weather environment influence sample data and a plurality of the photovoltaic power generation sample data;

[0153] The establishing module is further configured to train an initial network model according to the photovoltaic power generation prediction vector to obtain a preset photovoltaic power generation prediction model.

[0154] Further, the establishing module is further configured to analyze a plurality of the weather environment impact sample data and a plurality of the photovoltaic power generation sample data to obtain photovoltaic power generation impact feature information;

[0155] The establishing module is further configured to generate a photovoltaic power generation prediction vector according to the photovoltaic power generation impact feature information.

[0156] Further, the establishing module is further configured to integrate the photovoltaic power generation prediction vector through a non-linear mapping strategy to obtain a photovoltaic power generation integration vector;

[0157] The establishing module is further configured to perform dimensionality reduction processing on the photovoltaic power generation integration vector to obtain a photovoltaic power generation dimensionality reduction vector;

[0158] The establishing module is further configured to train an initial network model according to the photovoltaic power generation dimensionality reduction vector to obtain a preset photovoltaic power generation prediction model.

[0159] Further, the photovoltaic power generation prediction device based on Beidou short message also includes:

[0160] An acquisition module, configured to acquire the photovoltaic power generation prediction information corresponding to the weather environment data and a data acquisition device;

[0161] The acquisition module is further configured to determine the energy consumption prediction information of the data acquisition device;

[0162] The acquisition module is further configured to determine a data acquisition mode based on the photovoltaic power generation prediction information and the energy consumption prediction information, so that the data acquisition device performs data acquisition according to the data acquisition mode.

[0163] Further, the acquisition module is further configured to determine the terminal energy consumption information of the data acquisition device;

[0164] The acquisition module is further configured to input the terminal energy consumption information and the photovoltaic power generation prediction information into a preset energy consumption model to obtain the energy consumption prediction information of the data acquisition device.

[0165] Further, the determining module is further configured to extract a plurality of original photovoltaic power generation data from the photovoltaic power generation information;

[0166] The determining module is further configured to generate a photovoltaic power generation power curve according to the plurality of original photovoltaic power generation data;

[0167] The determining module is further configured to determine whether the photovoltaic power generation curve satisfies preset data conditions;

[0168] The determining module is further configured to, when the photovoltaic power generation curve does not satisfy the preset data conditions, process a plurality of original photovoltaic power generation data according to a preset nearest neighbor completion rule to obtain a plurality of historical photovoltaic power generation data.

[0169] Other embodiments or specific implementation manners of the photovoltaic power generation prediction device based on Beidou short message of the present invention may refer to the above method embodiments, and will not be described herein again.

[0170] It should be noted that, in this article, the terms "include", "comprise" or any other variation thereof are intended to cover a non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements but also other elements not expressly listed, or further includes elements inherent to such process, method, article or system. Without further limitation, an element defined by the phrase "including a..." does not exclude the existence of additional identical elements in the process, method, article or system including the element.

[0171] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

[0172] Through the description of the above embodiments, those skilled in the art can clearly understand that the above method of the embodiment can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as a read-only memory / random access memory, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0173] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the description and drawings of the present invention, or directly or indirectly applied to other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A photovoltaic power generation prediction method based on Beidou short message, characterized in that, The photovoltaic power prediction method based on Beidou short message includes: Obtain weather environment data and photovoltaic power generation information within a preset time; Determine a plurality of historical photovoltaic power generation data according to the photovoltaic power generation information and a preset nearest neighbor completion rule; Perform normalization processing on the weather environment data to obtain photovoltaic influence factor data; Integrate the photovoltaic power prediction vector through a non-linear mapping strategy to obtain a photovoltaic power integration vector; Perform dimensionality reduction processing on the photovoltaic power integration vector to obtain a photovoltaic power dimensionality reduction vector; Train an initial network model according to the photovoltaic power dimensionality reduction vector to obtain a preset photovoltaic power prediction model; Predict the photovoltaic power generation according to the photovoltaic influence factor data and a plurality of the historical photovoltaic power generation data through a preset photovoltaic power prediction model, wherein the preset photovoltaic power prediction model is a model constructed based on SSAE feature learning and BILSTM network; Determine the terminal energy consumption information of the data acquisition device corresponding to the weather environment data; Obtain a preset energy consumption model by constructing a sensor energy consumption model, a controller energy consumption model and a communication module energy consumption model of the data acquisition device; Input the terminal energy consumption information and the photovoltaic power generation prediction information corresponding to the weather environment data into the preset energy consumption model to obtain the energy consumption prediction information of the data acquisition device.

2. The method according to claim 1, characterized in that, Before the step of obtaining the weather environment data and the photovoltaic power generation information within a preset time, it further includes: Obtain a plurality of weather environment impact training data and a plurality of photovoltaic power generation training data according to the photovoltaic power generation influence factors; Perform normalization processing on the plurality of weather environment impact training data to obtain a plurality of weather environment impact sample data; Perform preprocessing on the plurality of photovoltaic power generation training data to obtain a plurality of photovoltaic power generation sample data; Train an initial network model according to the plurality of weather environment impact sample data and the plurality of photovoltaic power generation sample data to obtain a preset photovoltaic power prediction model.

3. The method according to claim 2, characterized in that, The step of training an initial network model according to the plurality of weather environment impact sample data and the plurality of photovoltaic power generation sample data to obtain a preset photovoltaic power prediction model includes: Generate a photovoltaic power prediction vector according to the plurality of weather environment impact sample data and the plurality of photovoltaic power generation sample data; Train an initial network model according to the photovoltaic power prediction vector to obtain a preset photovoltaic power prediction model.

4. The method according to claim 3, wherein The step of generating a photovoltaic power prediction vector according to the plurality of weather environment impact sample data and the plurality of photovoltaic power generation sample data includes: Analyze the plurality of weather environment impact sample data and the plurality of photovoltaic power generation sample data to obtain photovoltaic power generation influence characteristic information; Generate a photovoltaic power prediction vector according to the photovoltaic power generation influence characteristic information.

5. The method according to any one of claims 1 to 4, characterized in that After the step of predicting the photovoltaic power generation according to the photovoltaic influence factor data and a plurality of the historical photovoltaic power generation data through a preset photovoltaic power prediction model, it further includes: Obtain the photovoltaic power generation prediction information corresponding to the weather environment data and the data acquisition device; Determine the energy consumption prediction information of the data acquisition device; Determine a data acquisition mode based on the photovoltaic power generation prediction information and the energy consumption prediction information, so that the data acquisition device performs data acquisition according to the data acquisition mode.

6. The method according to any one of claims 1 to 4, characterized in that The step of determining multiple historical photovoltaic power generation data according to the photovoltaic power generation information and the preset nearest neighbor completion rule includes: Extract multiple original photovoltaic power generation data from the photovoltaic power generation information; Generate a photovoltaic power generation power curve graph based on the multiple original photovoltaic power generation data; Determine whether the photovoltaic power generation power curve graph meets the preset data conditions; When the photovoltaic power generation power curve graph does not meet the preset data conditions, process the multiple original photovoltaic power generation data according to the preset nearest neighbor completion rule to obtain multiple historical photovoltaic power generation data.

7. A photovoltaic power generation prediction device based on Beidou short message, characterized in that, The photovoltaic power generation prediction device based on Beidou short message includes: An acquisition module, configured to acquire weather environment data and photovoltaic power generation information within a preset time; A determination module, configured to determine multiple historical photovoltaic power generation data according to the photovoltaic power generation information and the preset nearest neighbor completion rule; A processing module, configured to perform normalization processing on the weather environment data to obtain photovoltaic influence factor data; A construction module, configured to integrate the photovoltaic power generation prediction vector through a non-linear mapping strategy to obtain a photovoltaic power generation integration vector; perform dimensionality reduction processing on the photovoltaic power generation integration vector to obtain a photovoltaic power generation dimensionality reduction vector; train an initial network model according to the photovoltaic power generation dimensionality reduction vector to obtain a preset photovoltaic power generation prediction model; A prediction module, configured to perform photovoltaic power generation prediction through a preset photovoltaic power generation prediction model according to the photovoltaic influence factor data and the multiple historical photovoltaic power generation data, wherein the preset photovoltaic power generation prediction model is a model constructed based on SSAE feature learning and BILSTM network; The construction module is further configured to determine the terminal energy consumption information of the data acquisition device corresponding to the weather environment data; obtain a preset energy consumption model by constructing a sensor energy consumption model, a controller energy consumption model, and a communication module energy consumption model of the data acquisition device; The prediction module is further configured to input the terminal energy consumption information and the photovoltaic power generation prediction information corresponding to the weather environment data into the preset energy consumption model to obtain the energy consumption prediction information of the data acquisition device.

8. A photovoltaic power generation power prediction device based on Beidou short message, characterized in that The device includes: a memory, a processor, and a Beidou short message-based photovoltaic power generation prediction program stored on the memory and executable on the processor, and the Beidou short message-based photovoltaic power generation prediction program is configured to implement the steps of the Beidou short message-based photovoltaic power generation prediction method according to any one of claims 1 to 6.

Images