A photovoltaic panel cleaning cycle prediction method based on LSTM model
Through the LSTM model-based photovoltaic panel cleaning cycle prediction method, combined with historical and weather data, the cleaning frequency of photovoltaic panels is optimized, and the problem of photovoltaic power generation efficiency and cost balance is solved, achieving more accurate prediction and improving economic benefits.
Patent Information
- Application Number
- CN202410824920.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-25
- Publication Date
- 2025-01-03
- Estimated Expiration
- 2044-06-25
AI Technical Summary
The prior art is difficult to find the best balance between ensuring photovoltaic power generation efficiency and cost, resulting in the problem of photovoltaic panel cleaning too frequently or too sparse.
Using the LSTM model-based photovoltaic panel cleaning cycle prediction method, the training data set and LSTM model are constructed, combined with historical photovoltaic power generation and weather characteristic data, predict the power generation of future photovoltaic panels, and determine the cleaning cycle through simulation calculation model to optimize photovoltaic Plate cleaning frequency.
Effectively capture the long-term dependencies in time series data, dynamically adjust photovoltaic power generation prediction, improve prediction accuracy, reduce unnecessary cleaning costs, and improve the economic benefits of photovoltaic power plants.
Smart Images
Figure CN118822002B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of renewable energy technology, and in particular to a photovoltaic panel cleaning cycle prediction method based on a long short-term memory network (LSTM). Background Art
[0002] As the global demand for renewable energy continues to grow, photovoltaic power generation, as a clean and renewable form of energy, is increasingly accounting for an increasing proportion of the global energy structure. The power generation efficiency of photovoltaic power generation systems is affected by many factors, among which the accumulation of dirt on the surface of photovoltaic panels is a factor that cannot be ignored. Dirt not only reduces the light receiving efficiency of photovoltaic panels, but in severe cases may also cause hot spot effects, which in turn affects the stability and life of the entire photovoltaic system. Cleaning photovoltaic panels is essential to maintaining their power generation efficiency. However, the more frequent the cleaning of photovoltaic panels, the better. Too frequent cleaning increases the operation and maintenance costs of photovoltaic power stations.
[0003] Therefore, how to provide a photovoltaic panel cleaning cycle prediction method based on an LSTM model that can reasonably predict the photovoltaic cleaning cycle is a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention
[0004] In view of the above research status, the present invention provides a photovoltaic panel cleaning cycle prediction method based on the LSTM model, aiming to achieve the optimal balance between the photovoltaic panel cleaning cost and the economic benefits of the photovoltaic power station.
[0005] The present invention provides a photovoltaic panel cleaning cycle prediction method based on an LSTM model, comprising the following steps:
[0006] Construct a training data set: including the historical operation time of photovoltaic panels, as well as the historical photovoltaic power generation and historical weather characteristics data during the corresponding historical operation period;
[0007] Constructing an LSTM model: using the historical operating time and the historical weather characteristic data as input and the historical photovoltaic power generation as output to train the LSTM model, thereby obtaining a trained LSTM model;
[0008] Predicting the actual power generation of photovoltaic panels: obtaining weather characteristic data in the future time period and inputting it into the trained LSTM model, and outputting the predicted value of the actual power generation of photovoltaic panels taking into account the operating time within the prediction period;
[0009] Simulating the theoretical power generation of photovoltaic panels: using a simulation calculation model to simulate and calculate the photovoltaic component model, the photovoltaic component model matches the actual parameters of the photovoltaic panel, and outputs the theoretical power generation of the photovoltaic panel without considering the operating time;
[0010] The cumulative daily dust accumulation profit loss is determined based on the photovoltaic grid-connected electricity price, and the photovoltaic panel cleaning judgment result is output according to the relationship value between the cumulative daily dust accumulation profit loss and the photovoltaic panel cleaning cost in the future time period, and the relationship value between the actual power generation prediction value of the photovoltaic panel and the theoretical power generation of the photovoltaic panel.
[0011] Preferably, the training data set includes:
[0012] Historical run time includes: the cumulative time since the last cleaning procedure was performed;
[0013] Historical PV power generation includes: daily power generation;
[0014] The historical weather characteristic data includes one or more combinations of solar radiation, rainfall, temperature, humidity and wind speed.
[0015] Preferably, the method further includes the steps of performing data analysis and feature selection on the training data set:
[0016] The relationship between historical photovoltaic power generation and historical operating time as well as historical weather characteristic data is analyzed. Through correlation analysis and feature importance evaluation, historical weather characteristics whose correlation with historical photovoltaic power generation during historical operating time meets the requirements are selected.
[0017] Preferably, the LSTM model includes a forget gate, an input gate and an output gate:
[0018] The forget gate is used to determine the weather characteristic data involved in the LSTM model processing, and the formula is as follows:
[0019] f t =σ(W f ×[h t-1 , x t ]+b f ),
[0020] Among them, f t represents the output of the forget gate at time t, σ is the sigmoid function, W f and b f is the weight matrix and bias term of the forget gate, h t-1 is the previous hidden state, x t is the current input;
[0021] The input gate is used to determine the current input x in the current time step. t Filter and store the filtering results in the current cell state. The current input x t Includes historical running time and historical weather characteristic data; consists of a sigmoid layer and a tanh layer, the formula is as follows:
[0022] it =σ(W i ×[h t-1 , x t ]+b i ),
[0023]
[0024] Among them, i t is the output of the input gate, W i and b i is the weight matrix and bias term of the input gate, is the candidate value vector, W C and b C is the weight matrix and bias term of the candidate value vector;
[0025] The output gate is used to determine the output of the current hidden state. The formula is as follows:
[0026] o t =σ(W o ×[h t-1 , x t ]+b o ),
[0027] h t =o t ×tanh(C t ),
[0028] Among them, t is the output of the output gate, W o and b o is the weight matrix and bias term of the output gate, h t is the current hidden state, C t The current unit status.
[0029] Preferably, the LSTM model is trained using a callback function.
[0030] Preferably, the LSTM model evaluation step is also included:
[0031] The normalized root mean square error, normalized mean absolute error and determination coefficient are selected to evaluate the trained LSTM model.
[0032] Preferably, the simulation calculation model performs the following steps:
[0033] Obtaining location information of the photovoltaic panel and meteorological data of the geographical area where the photovoltaic panel is located within a set time period;
[0034] Selecting a photovoltaic module model that matches the actual parameters of the photovoltaic panel and related performance parameters from the PVsyst database;
[0035] Use the Project Design module for simulation and output the theoretical power generation of the photovoltaic panel without considering the running time.
[0036] Preferably, the position information of the photovoltaic panel includes: the geographical longitude and latitude information where the photovoltaic panel is located, the installation tilt angle and azimuth angle of the photovoltaic panel; the meteorological data includes: the total solar irradiance on the horizontal plane, temperature; the relevant performance parameters include: inverter efficiency and system loss.
[0037] Preferably, the calculation steps of the cumulative daily dust accumulation income loss include:
[0038] The cumulative dust accumulation loss is
[0039] The daily dust accumulation income loss S t =(1 - η)×E Price ,
[0040] Relative power generation efficiency
[0041] where, E Price is the photovoltaic grid-connected electricity price, n is the prediction period, W Ash is the predicted value of the actual power generation of the photovoltaic panel, W Clean is the theoretical power generation of the photovoltaic panel.
[0042] Preferably, the step of outputting the cleaning judgment result of the photovoltaic panel according to the relationship value between the cumulative daily dust accumulation income loss and the cleaning cost of the photovoltaic panel within the future time period, and the relationship value between the predicted value of the actual power generation of the photovoltaic panel and the theoretical power generation of the photovoltaic panel includes the following steps:
[0043] When S > S clean , η < α and the rainfall in the R days after the prediction period < D, a cleaning alarm is given;
[0044] where, S clean is the cost of cleaning the photovoltaic panel once, α is the set threshold of the relative power generation efficiency, D is the rainfall threshold, and R ≥ 1.
[0045] The photovoltaic panel cleaning cycle prediction method based on the LSTM model proposed by the present invention has the following beneficial effects compared with the prior art:
[0046] The present invention applies the LSTM model to the prediction of the photovoltaic panel cleaning cycle. Through the internal gating mechanism, it can effectively capture the long-term dependence relationship in the time series data. At the same time, LSTM can remember and forget information, and can dynamically adjust the prediction of the photovoltaic power generation according to the real-time data and historical data to adapt to the change of environmental conditions.
[0047] The present invention obtains the predicted value of actual photovoltaic power generation within the prediction period through LSTM, and couples it with PVsyst simulation calculation to more accurately reflect the power generation efficiency of photovoltaic panels, further improve the accuracy of prediction, and help to more accurately predict the cleaning cycle of photovoltaic panels. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without creative work.
[0049] Figure 1 A flow chart of a photovoltaic panel cleaning cycle prediction method based on an LSTM model provided in an embodiment of the present invention;
[0050] Figure 2 A network structure diagram of the LSTM model provided for an embodiment of the present invention. DETAILED DESCRIPTION
[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0052] The present invention provides a photovoltaic panel cleaning cycle prediction method based on an LSTM model, comprising the following steps:
[0053] Construct a training data set: including the historical operation time of photovoltaic panels, as well as the historical photovoltaic power generation and historical weather characteristics data during the corresponding historical operation period;
[0054] Constructing LSTM model: Using historical operating time and historical weather characteristic data as input and historical photovoltaic power generation as output to train the LSTM model and obtain a trained LSTM model;
[0055] Predict the actual power generation of photovoltaic panels: obtain weather characteristic data in the future time period and input it into the trained LSTM model, and output the predicted value of the actual power generation of photovoltaic panels taking into account the operating time within the prediction period;
[0056] Simulate the theoretical power generation of photovoltaic panels: Use the simulation calculation model to simulate the photovoltaic module model. The photovoltaic module model matches the actual parameters of the photovoltaic panel and outputs the theoretical power generation of the photovoltaic panel without considering the operating time.
[0057] The cumulative daily dust accumulation loss is determined based on the photovoltaic grid-connected electricity price, and the photovoltaic panel cleaning judgment result is output according to the relationship between the cumulative daily dust accumulation loss and the photovoltaic panel cleaning cost in the future time period, as well as the relationship between the actual power generation prediction value of the photovoltaic panel and the theoretical power generation of the photovoltaic panel.
[0058] In one embodiment, the training data set includes:
[0059] Historical run time includes: the cumulative time since the last cleaning procedure was performed;
[0060] Historical PV power generation includes: daily power generation;
[0061] The historical weather characteristic data includes one or more combinations of solar radiation, rainfall, temperature, humidity and wind speed.
[0062] In this embodiment, the historical photovoltaic power generation and operation time are derived from the photovoltaic power stations that have been built in the area, and the weather characteristic data are derived from the existing weather forecast API data.
[0063] In one embodiment, the method further includes cleaning and preprocessing the training data set, including missing value processing, outlier detection and data standardization, and dividing the processed training data set into a training set (80%) and a test set (20%).
[0064] In one embodiment, the step of performing data analysis and feature selection on the training data set is also included:
[0065] The relationship between historical photovoltaic power generation and historical operating time as well as historical weather characteristic data is analyzed. Through correlation analysis and feature importance evaluation, historical weather characteristics whose correlation with historical photovoltaic power generation during historical operating time meets the requirements are selected.
[0066] It should be noted that the operation time and important weather characteristics that have a significant impact on the PV power generation prediction, such as solar radiation and rainfall, are selected through correlation analysis and feature importance assessment.
[0067] In one embodiment, the LSTM network model is used to learn and predict photovoltaic power generation. The LSTM model includes a forget gate, an input gate, and an output gate:
[0068] The forget gate is used to determine the weather feature data involved in the LSTM model processing and determine the key feature parameters (solar radiation, etc.) in the important weather features (solar radiation, rainfall, temperature, etc.). The formula is as follows:
[0069] f t =σ(W f ×[h t-1 , x t ]+bf )
[0070] Among them, f t represents the output of the forget gate at time t, σ is the sigmoid function, W f and b f is the weight matrix and bias term of the forget gate, h t-1 is the previous hidden state, x t is the current input;
[0071] The input gate is used to determine the current input x in the current time step. t Filter and store the filter results in the current cell state. The current input x t Includes historical running time and historical weather characteristic data; consists of a sigmoid layer and a tanh layer, the formula is as follows:
[0072] i t =σ(W i ×[h t-1 ,x t ]+b i ),
[0073]
[0074] Among them, i t is the output of the input gate, W i and b i is the weight matrix and bias term of the input gate, is the candidate value vector, W C and b C is the weight matrix and bias term of the candidate value vector;
[0075] The output gate is used to determine the output of the current hidden state. The formula is as follows:
[0076] o t =σ(W o ×[h t-1 , x t ]+b o ),
[0077] h t =o t ×tanh(C t ),
[0078] Among them, t is the output of the output gate, W o and b o is the weight matrix and bias term of the output gate, h t is the current hidden state, C t The current unit status.
[0079] In one embodiment, the LSTM model is trained using the preprocessed training set, a suitable batch size and number of iterations are selected, and a callback function is used to prevent overfitting, such as an early stopping function.
[0080] In one embodiment, the step of evaluating the prediction performance of the LSTM model using a test set is further included:
[0081] Select the normalized root mean square error (nRMSE), normalized mean absolute error (nMAE), and coefficient of determination (R 2 ) to evaluate the trained LSTM model. The formula is as follows:
[0082]
[0083] in, is the predicted value, y i is the actual value, is the average of the actual values.
[0084] Optimize the model by adjusting hyperparameters, adding or reducing features, changing the network structure, etc.
[0085] In one embodiment, the API interface is used to obtain weather parameters for the future time period and input them into the trained LSTM model to output the predicted value of the actual power generation of the photovoltaic panel considering the operating time within the prediction period (W Ash ).
[0086] In one embodiment, the simulation calculation model adopts the PVsyst simulation calculation method and performs the following steps:
[0087] Obtain the location information of the photovoltaic panels and the meteorological data of the geographical area where the photovoltaic panels are located within a set time period;
[0088] Select a PV module model that matches the actual parameters of the PV panel and related performance parameters from the PVsyst database;
[0089] Use the Project Design module to simulate and output the theoretical power generation of the photovoltaic panels without considering the operating time.
[0090] In this embodiment, the selected photovoltaic module model includes photovoltaic array design (such as the type and quantity of photovoltaic panels), system component configuration (such as selecting system components like inverters, junction boxes, cables, etc.); the position information of the photovoltaic panels includes: the geographical longitude and latitude information where the photovoltaic panels are located, the installation method of the photovoltaic panels (fixed method or tracking method), as well as the installation tilt angle and azimuth angle, etc.; the meteorological data includes: the total solar irradiance on the horizontal plane, temperature; the relevant performance parameters include: inverter efficiency and system losses, and the system losses include: line losses, module temperature losses, etc.
[0091] In one embodiment, the calculation steps of the cumulative daily dust accumulation revenue loss include:
[0092] The cumulative dust accumulation loss is
[0093] The daily dust accumulation revenue loss S t =(1 - η)×E Price ,
[0094] Relative power generation efficiency
[0095] where, E Price is the photovoltaic grid-connected electricity price, n is the prediction period, W Ash is the predicted value of the actual power generation of the photovoltaic panel, and W Clean is the theoretical power generation of the photovoltaic panel.
[0096] In this embodiment, according to the relationship value between the cumulative daily dust accumulation revenue loss and the cleaning cost of the photovoltaic panel in the future time period, and the relationship value between the predicted value of the actual power generation of the photovoltaic panel and the theoretical power generation of the photovoltaic panel, outputting the cleaning judgment result of the photovoltaic panel includes the following steps:
[0097] When S > S clean , η < α and the rainfall in R days after the prediction period < D, output the cleaning judgment result;
[0098] where, S clean is the cost of cleaning the photovoltaic panel once, α is the set threshold of the relative power generation efficiency, D is the rainfall threshold, and R ≥ 1. When specifically executed, R can take the value of 7, and α takes the value of 0 - 1. The closer it is to 1, the cleaner the photovoltaic panel is.
[0099] In this embodiment, S clean = S0×A + S1 + S2, where, S0 is the cleaning price of a group of photovoltaic panels, A is the number of groups of photovoltaic panels, S1 is the labor cost, and S2 is the single consumption cost of the cleaning robot.
[0100] In one embodiment, outputting the cleaning judgment result includes cleaning alarm: The following lists specific calculation examples for executing the cleaning alarm:
[0101] S1: Obtain weather characteristic data within the forecast period through the API interface, and obtain the actual power generation and theoretical power generation of the photovoltaic panels through LSTM model and Pvsyst simulation calculation respectively;
[0102] S2: Assuming that robot cleaning is adopted, the cleaning fee of photovoltaic panels is determined by the installed capacity, local labor costs and the cost of cleaning robots. Assuming that the calculated S clean =5000 yuan;
[0103] S3: Calculate the accumulated dust loss S. When S>5000 yuan and α=0.95, further determine whether there will be rainfall in the next 7 days through the weather parameters obtained from the API interface, and the rainfall is>20mm;
[0104] S4: If it is not raining and the rainfall is less than 20 mm, a cleaning warning will be issued. If the rainfall is greater than 20 mm, no cleaning warning will be issued.
[0105] S5: After the cleaning operation, the accumulated dust loss S=0 is reset, and the accumulated dust loss of the next cycle is performed.
[0106] The above is a detailed introduction to a photovoltaic panel cleaning cycle prediction method based on the LSTM model provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.
[0107] In this article, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
Claims
1. A photovoltaic panel cleaning cycle prediction method based on LSTM model, characterized by: The steps include: Construct a training data set: including the historical operation time of photovoltaic panels, as well as the historical photovoltaic power generation and historical weather characteristics data during the corresponding historical operation period; The training data set includes: Historical run time includes: the cumulative time since the last cleaning procedure was performed; Historical PV power generation includes: daily power generation; The historical weather characteristic data include: a combination of one or more of solar radiation, rainfall, temperature, humidity and wind speed; Performing data analysis and feature selection on the training data set: analyzing the relationship between historical photovoltaic power generation and historical operating time and historical weather characteristic data, and selecting historical weather features that meet the requirements of correlation with historical photovoltaic power generation during the historical operating time through correlation analysis and feature importance evaluation; Constructing an LSTM model: using the historical operating time and the historical weather characteristic data as input and the historical photovoltaic power generation as output to train the LSTM model, thereby obtaining a trained LSTM model; Predicting the actual power generation of photovoltaic panels: obtaining weather characteristic data in the future time period and inputting it into the trained LSTM model, and outputting the predicted value of the actual power generation of photovoltaic panels taking into account the operating time within the prediction period; Simulating the theoretical power generation of photovoltaic panels: using a simulation calculation model to simulate and calculate the photovoltaic component model, the photovoltaic component model matches the actual parameters of the photovoltaic panel, and outputs the theoretical power generation of the photovoltaic panel without considering the operating time; Determine the cumulative daily ash accumulation income loss based on the feed-in tariff for photovoltaic power generation. Output the cleaning judgment result of the photovoltaic panel according to the relationship value between the cumulative daily ash accumulation income loss and the cleaning cost of the photovoltaic panel in the future time period, and the relationship value between the predicted actual power generation of the photovoltaic panel and the theoretical power generation of the photovoltaic panel, including: when S > S clean , η < α and the rainfall in the R days after the prediction period < D, a cleaning alarm is issued; Among them, S clean is the cost of cleaning the photovoltaic panel once, α is the threshold for relative power generation efficiency, D is the rainfall threshold, R≥1; The cumulative daily ash accumulation income loss calculation steps include: The accumulated dust loss is Daily dust accumulation loss S t =(1-η)×E Price , Relative power generation efficiency Among them, E Price is the photovoltaic grid-connected electricity price, n is the forecast period, W Ash is the predicted value of the actual power generation of the photovoltaic panel, W Clean The theoretical power generation of photovoltaic panels.
2. A photovoltaic panel cleaning cycle prediction method based on LSTM model according to claim 1, characterized in that: The LSTM model includes a forget gate, an input gate, and an output gate: The forget gate is used to determine the weather characteristic data involved in the LSTM model processing, and the formula is as follows: f t =σ(W f ×[h t-1 ,x t ]+b f ), Among them, f t represents the output of the forget gate at time t, σ is the sigmoid function, W f and b f is the weight matrix and bias term of the forget gate, h t-1 is the previous hidden state, x t is the current input; The input gate is used to determine the current input x in the current time step. t Filter and store the filtering results in the current cell state. The current input x t Includes historical running time and historical weather characteristic data; consists of a sigmoid layer and a tanh layer, the formula is as follows: i t =σ(W i ×[h t-1 ,x t ]+b i ), Among them, i t is the output of the input gate, W i and b i is the weight matrix and bias term of the input gate, is the candidate value vector, W C and b C is the weight matrix and bias term of the candidate value vector; The output gate is used to determine the output of the current hidden state. The formula is as follows: the t =σ(W o ×[h t-1 ,x t ]·+b o ), h t =o t ×tanh(C t ), Among them, t is the output of the output gate, W o and b o is the weight matrix and bias term of the output gate, h t is the current hidden state, C t The current unit status.
3. A photovoltaic panel cleaning cycle prediction method based on LSTM model according to claim 1, characterized in that: The LSTM model is trained using a callback function.
4. A photovoltaic panel cleaning cycle prediction method based on LSTM model according to claim 1, characterized in that: The evaluation steps of the LSTM model are also included: The normalized root mean square error, normalized mean absolute error and determination coefficient are selected to evaluate the trained LSTM model.
5. A photovoltaic panel cleaning cycle prediction method based on LSTM model according to claim 1, characterized in that: The simulation calculation model performs the following steps: Obtaining location information of the photovoltaic panel and meteorological data of the geographical area where the photovoltaic panel is located within a set time period; Selecting a photovoltaic module model that matches the actual parameters of the photovoltaic panel and related performance parameters from the PVsyst database; Use the Project Design module to simulate and output the theoretical power generation of the photovoltaic panels without considering the operating time.
6. A photovoltaic panel cleaning cycle prediction method based on LSTM model according to claim 5, characterized in that: The location information of the photovoltaic panel includes: the geographical longitude and latitude information of the photovoltaic panel, the installation inclination angle and azimuth of the photovoltaic panel; the meteorological data includes: the total solar radiation and temperature on the horizontal plane; the relevant performance parameters include: inverter efficiency and system loss.