Photovoltaic generating capacity prediction method and device and controller

Abstract

The invention discloses a photovoltaic power generation prediction method and device and a controller, and relates to the technical field of photovoltaic power generation prediction, and the method comprises the steps: obtaining a photovoltaic power change curve of a to-be-predicted target photovoltaic module, and obtaining photovoltaic power data observed in a current to-be-predicted time period; according to the photovoltaic power change curve and the photovoltaic power data, the photovoltaic power generation capacity in the remaining time of the current to-be-predicted time period is predicted, the photovoltaic power generation capacity is better fitted and predicted through a large amount of collected historical photovoltaic power generation capacity data, a complex calculation process and a specific AI chip are not needed, the use cost can be remarkably reduced, and the method is suitable for popularization and application. And the problems of frequent shutdown of the load, insufficient photovoltaic utilization rate and the like are solved.

Description

Technical Field

[0001] The present application relates to the technical field of photovoltaic power generation prediction, and in particular, to a photovoltaic power generation prediction method, device and controller. Background Art

[0002] In the photovoltaic-energy storage-load system, existing load equipment cannot adjust its power in a timely manner according to the photovoltaic output, resulting in frequent shutdowns and insufficient photovoltaic utilization. Among related technologies, photovoltaic power generation prediction methods mostly rely on AI technology or meteorological forecast data, with high training and deployment costs and complex processes. Summary of the Invention

[0003] In order to overcome the deficiencies of the prior art, the present application provides a photovoltaic power generation prediction method, device and controller.

[0004] The technical solution adopted by this application to solve its technical problems is:

[0005] In a first aspect, a method for predicting photovoltaic power generation is provided, the method comprising:

[0006] Acquire a photovoltaic power variation curve of a target photovoltaic assembly to be predicted, wherein the photovoltaic power variation curve is used to represent a variation pattern of the photovoltaic power of the target photovoltaic assembly to be predicted over time within a time period;

[0007] Obtain the photovoltaic power data observed during the current forecast period;

[0008] The photovoltaic power generation amount in the remaining time of the current time period to be predicted is predicted according to the photovoltaic power change curve and the photovoltaic power data.

[0009] Furthermore, predicting the photovoltaic power generation in the remaining time of the current time period to be predicted based on the photovoltaic power change curve and the photovoltaic power data includes: obtaining the posterior distribution of the photovoltaic power change curve based on the Bayesian formula and the photovoltaic power data; and predicting the photovoltaic power generation in the remaining time of the current time period to be predicted using the posterior distribution of the photovoltaic power change curve.

[0010] Furthermore, obtaining the posterior distribution of the photovoltaic power change curve according to the Bayesian formula and the photovoltaic power data includes: obtaining the prior Gaussian distribution of the photovoltaic power change curve; determining the likelihood function corresponding to the photovoltaic power data; and obtaining the posterior distribution of the photovoltaic power change curve according to the Bayesian formula, the prior Gaussian distribution and the likelihood function.

[0011] Furthermore, obtaining the prior Gaussian distribution of the photovoltaic power change curve includes: obtaining a parameter mean of the photovoltaic power change curve; calculating a covariance matrix based on the parameter mean; and obtaining the prior Gaussian distribution of the photovoltaic power change curve based on the parameter mean and the covariance matrix.

[0012] Furthermore, determining the likelihood function corresponding to the photovoltaic power data includes: determining a function type corresponding to the photovoltaic power data; and determining the likelihood function corresponding to the photovoltaic power data based on the function type.

[0013] Furthermore, obtaining the posterior distribution of the photovoltaic power change curve according to the Bayesian formula, the prior Gaussian distribution and the likelihood function includes: obtaining the original posterior distribution of the photovoltaic power change curve according to the Bayesian formula, the prior Gaussian distribution and the likelihood function; obtaining the posterior coefficient of the original posterior distribution; judging whether the original posterior distribution coefficient meets the preset error index; if the original posterior distribution coefficient meets the preset error index, using the original posterior distribution as the posterior distribution; if the original posterior distribution coefficient does not meet the preset error index, correcting the original posterior distribution to obtain the posterior distribution.

[0014] Furthermore, correcting the original posterior distribution includes: correcting a parameter mean of the photovoltaic power variation curve; and obtaining a posterior distribution corresponding to the photovoltaic power data according to the parameter mean of the corrected photovoltaic power variation curve.

[0015] Furthermore, the parameter mean of the photovoltaic power variation curve is corrected by the following formula: Among them, ε is the preset correction coefficient, a1, b1, and c1 are the parameters of the photovoltaic power change curve, u0 is the parameter mean of the photovoltaic power change curve before correction, and u1 is the parameter mean of the photovoltaic power change curve after correction.

[0016] Furthermore, obtaining the photovoltaic power variation curve of the target photovoltaic assembly to be predicted includes: obtaining historical photovoltaic power generation data of the target photovoltaic assembly to be predicted; and fitting the photovoltaic power variation curve according to the historical photovoltaic power generation data.

[0017] Furthermore, fitting the photovoltaic power change curve according to the historical photovoltaic power generation data includes: dividing the historical photovoltaic power generation data according to date; sampling the historical photovoltaic power generation data of each date by time period, and taking the maximum power generation in each time period; calculating the first-order difference data of the maximum power generation in each time period; obtaining the upper and lower bounds corresponding to the specified percentage quantile interval of each first-order difference data; truncating each first-order difference according to a fixed interval; and obtaining the photovoltaic power change curve according to the truncation result and the maximum power generation in each time period.

[0018] Furthermore, obtaining the photovoltaic power data observed in the current time period to be predicted includes: determining whether photovoltaic power data has been generated in the current time period to be predicted; if not, delaying one time step and obtaining the photovoltaic power data observed in the current time period to be predicted again until the photovoltaic power data observed in the current time period to be predicted is obtained.

[0019] Furthermore, the posterior distribution of the photovoltaic power variation curve is obtained by the following formula:

[0020] p(θ|y)=p(θ)* p(y|θ) =N(u n , Σ n ),in:

[0021] Among them, p(θ|y) is the posterior distribution of the photovoltaic power change curve, p(θ) is the prior Gaussian distribution of the photovoltaic power change curve, p(y|θ) is the likelihood function corresponding to the photovoltaic power data, and u n is the parameter mean of the photovoltaic power variation curve, Σ n is the covariance matrix of the photovoltaic power change curve, u0 is the mean of the historical photovoltaic power generation data curve, Σ0 is the covariance matrix of the historical photovoltaic power generation data curve, and X is a matrix.

[0022] Furthermore, the function type corresponding to the photovoltaic power data is a quadratic function, and the likelihood function corresponding to the photovoltaic power data is determined by the following formula: Among them, p(y |θ) is the likelihood function corresponding to the photovoltaic power data, y t is the photovoltaic power generation power, x t is time, a, b, c are the parameters of the photovoltaic power change curve, σ 2 is the noise variance.

[0023] Furthermore, it also includes: controlling the power supply device to adjust the power according to the photovoltaic power generation in the remaining time of the current time period to be predicted.

[0024] Further, ε=0.03, or, ε=-0.03.

[0025] In a second aspect, a photovoltaic power generation prediction device is provided, comprising:

[0026] A photovoltaic curve acquisition module is used to obtain a photovoltaic power variation curve of a target photovoltaic assembly to be predicted, wherein the photovoltaic power variation curve is used to represent a variation pattern of the photovoltaic power of the target photovoltaic assembly to be predicted over time within a time period;

[0027] Photovoltaic data acquisition module, used to obtain photovoltaic power data observed during the current forecast period;

[0028] The power generation prediction module is used to predict the photovoltaic power generation in the remaining time of the current time period to be predicted based on the photovoltaic power change curve and the photovoltaic power data.

[0029] In a third aspect, a photovoltaic power generation controller is provided, comprising:

[0030] at least one processor and at least one memory;

[0031] The memory stores executable instructions of the processor;

[0032] The processor is configured to execute any one of the photovoltaic power generation prediction methods described above.

[0033] The technical solution of the present application provides a photovoltaic power generation prediction method, device and controller, which obtain the photovoltaic power change curve of the target photovoltaic module to be predicted, and obtain the photovoltaic power data observed in the current time period to be predicted; the photovoltaic power generation in the remaining time of the current time period to be predicted is predicted based on the photovoltaic power change curve and the photovoltaic power data, and the photovoltaic power generation is better fitted by collecting a large amount of historical photovoltaic power generation data. Without the need for complex calculation processes and specific AI chips, the cost of use can be significantly reduced, and problems such as frequent load shutdowns and insufficient photovoltaic utilization can be solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0035] Figure 1 This is a flow chart of a photovoltaic power generation prediction method provided in an embodiment of the present application;

[0036] Figure 2 This is a flow chart of another photovoltaic power generation prediction method provided in an embodiment of the present application;

[0037] Figure 3 This is a schematic diagram of the photovoltaic energy storage load system structure in an embodiment of the present application;

[0038] Figure 4 This is a schematic diagram of a photovoltaic power generation curve in an embodiment of the present application;

[0039] Figure 5 This is a schematic diagram of a photovoltaic power generation oscillation curve with load in an embodiment of the present application;

[0040] Figure 6 This is a schematic diagram of a curve when photovoltaic power is in excess in an embodiment of the present application;

[0041] Figure 7 1 is a flow chart of the photovoltaic power variation curve fitting process in an embodiment of the present application;

[0042] Figure 8 Schematic diagram of the photovoltaic power generation prediction process in the embodiment of the present application;

[0043] Figure 9 This is a functional structural diagram of a photovoltaic power generation prediction device provided in an embodiment of the present application;

[0044] Figure 10 This is a functional structural diagram of another photovoltaic power generation prediction device provided in an embodiment of the present application;

[0045] Figure 11 This is a functional structure diagram of the photovoltaic power generation controller provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application are described in detail below with reference to the accompanying drawings and examples. Obviously, the described embodiments are only some of the embodiments of this application, rather than all of them. Based on the embodiments in this application, all other implementation methods obtained by ordinary technicians in this field without making any creative work are within the scope of protection of this application.

[0047] To solve the above problems, refer to Figure 1 , an embodiment of the present application provides a photovoltaic power generation prediction method, the method comprising:

[0048] 101. Obtain a photovoltaic power variation curve of a target photovoltaic module to be predicted.

[0049] The photovoltaic power variation curve is used to represent the variation pattern of the photovoltaic power of the target photovoltaic module to be predicted over time within a time period.

[0050] 102. Obtain photovoltaic power data observed during the current prediction time period.

[0051] 103. Predict the photovoltaic power generation in the remaining time of the current predicted time period based on the photovoltaic power change curve and photovoltaic power data.

[0052] The photovoltaic power generation prediction method provided in this embodiment obtains the photovoltaic power change curve of the target photovoltaic module to be predicted and obtains the photovoltaic power data observed in the current time period to be predicted; the photovoltaic power generation in the remaining time of the current time period to be predicted is predicted based on the photovoltaic power change curve and the photovoltaic power data. The photovoltaic power generation is better fitted by the large amount of historical photovoltaic power generation data collected. Without the need for complex calculation processes and specific AI chips, the use cost can be significantly reduced, and problems such as frequent load shutdowns and insufficient photovoltaic utilization can be solved.

[0053] As an improvement to the above embodiment, the present invention provides another photovoltaic power generation prediction method. Figure 2 , the method comprising:

[0054] 201. Obtain a photovoltaic power variation curve of a target photovoltaic module to be predicted.

[0055] Photovoltaic modules are the core units that convert solar energy into electricity. Composed of encapsulated photovoltaic cells, they offer structural stability and efficient power generation. Photovoltaic power is a key indicator of a PV system's power generation capacity, typically expressed in watts (W) or kilowatts (KW).

[0056] Figure 3 This is a schematic diagram of a common non-grid-connected photovoltaic-energy storage-load system, where a controller coordinates photovoltaic output, energy storage charging and discharging, and load power consumption. PV modules are solar panels, energy storage devices are batteries, and loads are electrical devices such as air conditioners and water heaters. Existing load control logic is relatively simple and cannot predict photovoltaic power generation. Therefore, when photovoltaic power is less than load power, the load shuts down, and otherwise restarts. The energy storage battery charges during the day when solar power is sufficient and discharges at night when solar power is scarce to supply power to the load.

[0057] The photovoltaic power variation curve is used to represent the time-dependent variation of the photovoltaic power of the target photovoltaic module to be predicted within a time period. In the embodiment of the present invention, the time period is 6:00-18:00 of a day. In specific engineering practice, it can also be other time periods.

[0058] In some optional embodiments, 201 may be implemented by, but is not limited to, the following process (not shown in the figure):

[0059] 2011. Obtain historical photovoltaic power generation data of the target photovoltaic module to be predicted.

[0060] 2012. Fitting the photovoltaic power change curve based on historical photovoltaic power generation data.

[0061] In some optional embodiments, 2012 may be implemented by, but not limited to, the following process (not shown in the figure):

[0062] 2012-1. The historical photovoltaic power generation data is segmented by date.

[0063] 2012-2. The historical photovoltaic power generation data of each date is sampled by time period, and the maximum power generation in each time period is taken.

[0064] 2012-3. Calculate the first-order difference data of the maximum power generation in each time period.

[0065] 2012-4. Get the upper and lower bounds corresponding to the specified percentage quantile interval of each first-order difference data.

[0066] 2012-5. Each first-order difference is truncated according to a fixed interval.

[0067] 2012-6. The photovoltaic power change curve is obtained based on the truncation results and the maximum power generation in each period.

[0068] Figure 4 This diagram illustrates photovoltaic power generation data collected by a photovoltaic-storage-load system in a region with abundant solar resources. During the summer months of May to August, photovoltaic power is available from approximately 6:00 AM to 6:00 PM, reaching a maximum of approximately 3000W around noon. The system collects data every minute. Circles represent the photovoltaic power at a specific moment, and dashed lines represent the curve-fitted result.

[0069] Figure 4 、 Figure 5 、 Figure 6 Three common photovoltaic power fluctuations in historical data were statistically analyzed.

[0070] Figure 4 This is a typical curve diagram of photovoltaic abundance in summer. Most of the data are real data, and the data fluctuations are slightly large.

[0071] Figure 5In the morning, between 7:00 and 8:00, as photovoltaic power begins to increase, the load detects the presence of photovoltaic power and starts to increase the frequency. However, the photovoltaic power is insufficient to support the load at its maximum frequency, so the load device shuts down. This process repeats after a period of time, resulting in frequent starts and stops during the morning period.

[0072] Figure 6 Starting at 2:00 PM, PV power drops sharply, then stabilizes until around 4:00 PM, forming an L-shaped curve. This occurs because the load power is maintained at a constant value and the energy storage device is fully charged, making it impossible for either to absorb more PV. This results in an L-shaped curve, and the PV power during this period is underutilized and wasted.

[0073] See also Figure 7 , the photovoltaic power variation curve of the target photovoltaic module to be predicted can be fitted through the following process:

[0074] First, the collected historical photovoltaic power generation data is segmented by date. Taking the data for May 1, 2024 as an example, duplicate data is removed and data from 06:00 to 18:00 is filtered.

[0075] Then, some abnormal values are removed and interpolation is performed using an interpolation method. The embodiment of the present invention does not limit the specific interpolation method, and it can be a Newton interpolation method, a Lagrange interpolation method, a cubic spline interpolation method, etc.

[0076] The maximum value of photovoltaic power generation in the sampled data in a specific time period T is used to obtain the sequence Pmax, and its first-order difference sequence Pdiff is calculated. The first difference is filled with 0, where the time period can be 5 minutes.

[0077] To avoid excessive data fluctuations, we use the upper and lower bounds [a, b] corresponding to the X% quantile interval of the first-order difference series. When X is 90, if the corresponding upper and lower bounds are [-5, 10], it means that 90% of the first-order difference data are between -5 and 10, thus avoiding large fluctuations in the first-order difference.

[0078] The truncated differential sequence Pdiff_r is added to the Pmax of the previous moment to obtain the corrected photovoltaic power change curve Previse, that is, Previse[i]=Pdiff_r[i]+Pmax[i-1], where i represents the subscript corresponding to a certain time point. The corrected Previse is a relatively smooth curve that can outline the outline of photovoltaic power generation, such as Figure 4 As shown by the dotted line in .

[0079] Finally, we fit Previse using the quadratic curve equation f(x) = ax² + bx + c to obtain the corresponding parameters a, b, and c. Given this equation and the acquired power generation data, we can use the curve_fit function in the scientific computing library scipy to fit the curve equation and save the corresponding parameters and errors for the next step of photovoltaic power generation prediction.

[0080] It should be noted that the technical solution claimed to be protected by this application is not limited to the use of quadratic curve equations, and other forms of curve equations can also be used. The specific use process refers to the use of quadratic curve equations in this embodiment and will not be repeated here.

[0081] 202. Obtain photovoltaic power data observed during the current prediction time period.

[0082] In a specific implementation process, existing photovoltaic power data can be obtained through observation, so that actual photovoltaic curve parameters can be subsequently determined based on the obtained photovoltaic power data.

[0083] In some optional embodiments, 202 may be implemented by, but not limited to, the following process (not shown):

[0084] 2021. Determine whether photovoltaic power data has been generated within the current prediction time period;

[0085] 2022. If not, after a delay of one time step, the photovoltaic power data observed in the current time period to be predicted is obtained again until the photovoltaic power data observed in the current time period to be predicted is obtained.

[0086] Since the sun rises at different times every day, the starting time of the observed photovoltaic power data is different. If photovoltaic power generation prediction starts when photovoltaic power generation has not yet been generated, the observed photovoltaic power data will be obtained again after a delay of one time period until the observed photovoltaic power data is obtained.

[0087] 203. Predict the photovoltaic power generation in the remaining time of the current predicted time period based on the photovoltaic power change curve and the photovoltaic power data.

[0088] In some optional embodiments, 203 may be implemented by, but not limited to, the following process (not shown in the figure):

[0089] 2031. Obtain the posterior distribution of the photovoltaic power change curve based on the Bayesian formula and photovoltaic power data.

[0090] In Bayesian statistics, the posterior distribution is the conditional probability distribution of an unknown parameter given observed data. It combines the prior distribution with the likelihood of the observed data to update the belief about the parameter. It reflects the latest state of knowledge about the parameter after obtaining new data. This updating process emphasizes the continuous correction of prior knowledge through data.

[0091] In some optional embodiments, 2031 may be implemented by, but not limited to, the following process (not shown in the figure):

[0092] 2031-1. Obtain the prior Gaussian distribution of the photovoltaic power variation curve.

[0093] Specifically, in some optional embodiments, the prior Gaussian distribution of the photovoltaic power variation curve may be obtained through the following process:

[0094] First, the parameter mean of the photovoltaic power variation curve is obtained.

[0095] Next, the covariance matrix is calculated based on the parameter means.

[0096] Finally, the prior Gaussian distribution of the photovoltaic power variation curve is obtained according to the parameter mean and covariance matrix.

[0097] 2031-2. Determine the likelihood function corresponding to the photovoltaic power data.

[0098] In some optional embodiments, the likelihood function corresponding to the photovoltaic power data may be determined by the following process:

[0099] First, determine the function type corresponding to the photovoltaic power data.

[0100] Secondly, the likelihood function corresponding to the photovoltaic power data is determined based on the function type.

[0101] In some optional embodiments, the function type corresponding to the photovoltaic power data is a quadratic function, and the likelihood function corresponding to the photovoltaic power data is determined by the following formula:

[0102] Among them, p(y|θ) is the likelihood function corresponding to the photovoltaic power data, y t is the actual photovoltaic power generation power, x t is time, a, b, c are the parameters of the photovoltaic power change curve, σ 2 is the noise variance.

[0103] 2031-3. Obtain the posterior distribution of the photovoltaic power change curve based on the Bayesian formula, the prior Gaussian distribution and the likelihood function.

[0104] In some optional embodiments, the posterior distribution of the photovoltaic power variation curve may be obtained by the following formula:

[0105] p(θ|y)=p(θ)* p(y|θ) =N(u n , Σ n ),in, Among them, p(θ|y) is the posterior distribution of the photovoltaic power change curve, p(θ) is the prior Gaussian distribution of the photovoltaic power change curve, and p(y|θ ) is the likelihood function corresponding to the photovoltaic power data, u n is the parameter mean of the photovoltaic power variation curve, Σ n is the covariance matrix of the photovoltaic power change curve, u0 is the mean of the historical photovoltaic power generation data curve, Σ0 is the covariance matrix of the historical photovoltaic power generation data curve, and X is a matrix.

[0106] In some optional embodiments, 2031-3 may be implemented by, but is not limited to, the following process:

[0107] First, obtain the posterior coefficients of the original posterior distribution.

[0108] Secondly, determine whether the original posterior distribution coefficient meets the preset error index.

[0109] If the original posterior distribution coefficient meets the preset error index, the original posterior distribution is used as the posterior distribution.

[0110] If the original posterior distribution coefficient does not meet the preset error index, the original posterior distribution is modified to obtain the posterior distribution.

[0111] In some optional embodiments, the original posterior distribution can be modified in the following way:

[0112] First, the parameter mean of the photovoltaic power variation curve is corrected.

[0113] Specifically, the parameter mean of the photovoltaic power change curve can be corrected by the following formula:

[0114] Wherein, ε is a preset correction coefficient, a1, b1, and c1 are parameters of the photovoltaic power change curve, u0 is the parameter mean of the photovoltaic power change curve before correction, and u1 is the parameter mean of the photovoltaic power change curve after correction. Optionally, ε = 0.03, or ε = -0.03.

[0115] Secondly, the posterior distribution corresponding to the photovoltaic power data is obtained according to the parameter mean of the corrected photovoltaic power change curve.

[0116] 2032. Use the posterior distribution of the photovoltaic power change curve to predict the photovoltaic power generation in the remaining time of the current predicted time period.

[0117] In some alternative embodiments, see Figure 8 ,The process of predicting the photovoltaic power generation in the remaining time of the current prediction period is as follows:

[0118] According to the curve equation of the large amount of historical data collected in the previous step, 31 sets of parameters are obtained: {a(k), b(k), c(k)}, k∈[1,31]. Through statistics, the parameters obey the Gaussian distribution p(θ)=N(u0, Σ 0), θ=[a,b,c] T .

[0119] in,

[0120]

[0121] The observed data satisfies the quadratic function:

[0122]

[0123] Among them, x t is time, y t is the photovoltaic power generation power, Indicates compliance with (0,σ 2 ) Gaussian distribution error. Then the likelihood function corresponding to N observation data points is:

[0124]

[0125] According to Bayes' theorem, the posterior distribution p(θ|y)=p(θ)*p(y|θ)=N(u n ,Σ n ),in:

[0126]

[0127] Among them, the design matrix X is:

[0128]

[0129] Calculation example:

[0130] Assume that the prior statistical results of the May parameters are:

[0131]

[0132] The first three observation data obtained on June 1 are: (x1=1, y1=0.1), (x2=2, y2=0.3), (x3=3, y3=0.7).

[0133] Noise variance σ 2 =0.1, design matrix:

[0134]

[0135] Posterior covariance update process: Calculate its inverse matrix using ∑0 and substitute it into Equation 1.3. Substitute the relevant data into Equation 1.4 to obtain the parameters a1, b1, and c1. Use these parameters to fit the remaining data for the current time step, which can be 1 hour. Determine whether the error metric is met. If so, proceed to the next time step and repeat the process. Otherwise, adjust the coefficients until the metrics (such as MAE, RMSE, and R²) are met.

[0136] MAE (Mean Absolute Error) is the average of the absolute differences between the predicted and actual values. It gives the average size of the prediction error, but does not consider the direction of the error (positive or negative). Compared to MSE and RMSE, MAE is less sensitive to outliers and does not amplify the results due to the squares of a few large errors. It is suitable for datasets with many outliers. RMSE is the root mean square error. A smaller RMSE is better. RMSE is a typical metric for regression models, used to indicate the magnitude of the error a model will produce in its predictions, with higher weighting for larger errors. R² is the coefficient of determination, which indicates the quality of the model and ranges from [0 to 1]. The closer the value is to 1, the better the fit, and vice versa.

[0137] The coefficient correction method is:

[0138]

[0139] The mean() function is used to find the average value. t ,y p )>0, ε is 0.03; otherwise, when mean(y t ,y p )<0,ε is -0.03. t ,y p They represent the actual photovoltaic power generation and the predicted photovoltaic power generation of the photovoltaic module to be tested respectively, so that the predicted value of the corrected curve will be closer to the actual value.

[0140] 204. Control the power supply device to adjust power according to the photovoltaic power generation in the remaining time of the current predicted time period.

[0141] The photovoltaic power generation prediction method provided in this embodiment obtains the photovoltaic power change curve of the target photovoltaic module to be predicted and obtains the photovoltaic power data observed in the current time period to be predicted; the photovoltaic power generation in the remaining time of the current time period to be predicted is predicted based on the photovoltaic power change curve and the photovoltaic power data. The photovoltaic power generation is better fitted by the large amount of historical photovoltaic power generation data collected. Without the need for complex calculation processes and specific AI chips, the use cost can be significantly reduced, and problems such as frequent load shutdowns and insufficient photovoltaic utilization can be solved.

[0142] In order to implement the above photovoltaic power generation prediction method, the embodiment of the present invention provides a photovoltaic power generation prediction device, see Figure 9 , the device comprises:

[0143] The photovoltaic curve acquisition module 91 is used to acquire the photovoltaic power variation curve of the target photovoltaic assembly to be predicted. The photovoltaic power variation curve is used to represent the variation pattern of the photovoltaic power of the target photovoltaic assembly to be predicted over time within a time period.

[0144] The photovoltaic data acquisition module 92 is used to acquire photovoltaic power data observed in the current prediction time period.

[0145] The power generation prediction module 93 is used to predict the photovoltaic power generation in the remaining time of the current time period to be predicted based on the photovoltaic power change curve and photovoltaic power data.

[0146] The photovoltaic power generation prediction device provided in this embodiment obtains the photovoltaic power change curve of the target photovoltaic module to be predicted and obtains the photovoltaic power data observed in the current time period to be predicted; the photovoltaic power generation in the remaining time of the current time period to be predicted is predicted based on the photovoltaic power change curve and the photovoltaic power data. The photovoltaic power generation is better fitted by the large amount of historical photovoltaic power generation data collected, and no complex calculation process and specific AI chips are required. It can significantly reduce the cost of use and solve problems such as frequent load shutdowns and insufficient photovoltaic utilization.

[0147] As an improvement to the above embodiment, the present invention provides another photovoltaic power generation prediction device. Figure 10 , the device comprises:

[0148] The photovoltaic curve acquisition module 11 is used to acquire a photovoltaic power variation curve of a target photovoltaic module to be predicted. The photovoltaic power variation curve is used to represent a variation pattern of the photovoltaic power of the target photovoltaic module to be predicted over time within a time period.

[0149] In some optional embodiments, the photovoltaic curve acquisition module 11 includes:

[0150] The historical data acquisition unit 111 is used to acquire historical photovoltaic power generation data of the target photovoltaic module to be predicted.

[0151] The curve fitting unit 112 is used to fit the photovoltaic power variation curve according to the historical photovoltaic power generation data.

[0152] In some optional embodiments, the curve fitting unit 112 includes:

[0153] The data segmentation subunit 1121 is used to segment the historical photovoltaic power generation data according to date.

[0154] The data sampling subunit 1122 is used to sample the historical photovoltaic power generation data of each date by time period and obtain the maximum power generation value in each time period.

[0155] The difference calculation subunit 1123 is used to calculate the first-order difference data of the maximum power generation in each time period.

[0156] The boundary acquisition subunit 1124 is used to obtain the upper bound and the lower bound corresponding to the specified percentage quantile interval of each first-order difference data.

[0157] The truncation subunit 1125 truncates each first-order difference according to a fixed interval.

[0158] The curve acquisition subunit 1126 is used to obtain a photovoltaic power variation curve according to the truncation result and the maximum power generation in each time period.

[0159] The photovoltaic data acquisition module 12 is used to acquire photovoltaic power data observed in the current prediction time period.

[0160] The power generation prediction module 13 is used to predict the photovoltaic power generation in the remaining time of the current time period to be predicted based on the photovoltaic power change curve and photovoltaic power data.

[0161] In some optional embodiments, the power generation prediction module 13 includes:

[0162] The posterior distribution acquisition unit 131 is configured to acquire the posterior distribution of the photovoltaic power variation curve according to the Bayesian formula and the photovoltaic power data.

[0163] In some optional embodiments, the posterior distribution acquisition unit 131 includes:

[0164] The prior Gaussian distribution subunit 1311 is used to obtain the prior Gaussian distribution of the photovoltaic power variation curve.

[0165] In some optional embodiments, the prior Gaussian distribution subunit 1311 obtains the prior Gaussian distribution of the photovoltaic power change curve through the following process: the prior Gaussian distribution subunit 1311 obtains the parameter mean of the photovoltaic power change curve; calculates the covariance matrix based on the parameter mean; obtains the prior Gaussian distribution of the photovoltaic power change curve based on the parameter mean and the covariance matrix.

[0166] The likelihood function determination subunit 1312 is configured to determine a likelihood function corresponding to the photovoltaic power data.

[0167] In some optional embodiments, the likelihood function determination subunit 1312 determines the likelihood function corresponding to the photovoltaic power data through the following process: the likelihood function determination subunit 1312 determines the function type corresponding to the photovoltaic power data; and determines the likelihood function corresponding to the photovoltaic power data based on the function type.

[0168] In some optional embodiments, the function type corresponding to the photovoltaic power data is a quadratic function, and the likelihood function determination subunit 1312 determines the likelihood function corresponding to the photovoltaic power data using the following formula:

[0169] Among them, p(y|θ) is the likelihood function corresponding to the photovoltaic power data, y t is the photovoltaic power generation power, x t is time, a, b, c are the parameters of the photovoltaic power change curve, σ 2 is the noise variance.

[0170] The posterior distribution determination subunit 1313 is configured to obtain the posterior distribution of the photovoltaic power variation curve according to the Bayesian formula, the prior Gaussian distribution, and the likelihood function.

[0171] In some optional embodiments, the photovoltaic data acquisition module 12 determines whether photovoltaic power data has been generated in the current time period to be predicted; if not, it delays one time step and acquires the photovoltaic power data observed in the current time period to be predicted again until the photovoltaic power data observed in the current time period to be predicted is acquired.

[0172] In some optional embodiments, the posterior distribution determining subunit 1313 obtains the posterior distribution of the photovoltaic power variation curve by the following formula:

[0173] p(θ|y)=p(θ)*p(y|θ)=N(u n ,Σ n ),in,

[0174] Among them, p(θ|y) is the posterior distribution of the photovoltaic power change curve, p(θ) is the prior Gaussian distribution of the photovoltaic power change curve, p(y|θ) is the likelihood function corresponding to the photovoltaic power data, and u n is the parameter mean of the photovoltaic power variation curve, Σ n is the covariance matrix of the photovoltaic power change curve, u0 is the mean of the historical photovoltaic power generation data curve, Σ0 is the covariance matrix of the historical photovoltaic power generation data curve, and X is a matrix.

[0175] In some optional embodiments, the posterior distribution determination subunit 1313 obtains the posterior distribution of the photovoltaic power change curve according to the Bayesian formula, the prior Gaussian distribution and the likelihood function through the following process: the posterior distribution determination subunit 1313 obtains the original posterior distribution of the photovoltaic power change curve according to the Bayesian formula, the prior Gaussian distribution and the likelihood function: obtains the posterior coefficient of the original posterior distribution; determines whether the original posterior distribution coefficient meets the preset error index; if the original posterior distribution coefficient meets the preset error index, the original posterior distribution is used as the posterior distribution; if the original posterior distribution coefficient does not meet the preset error index, the original posterior distribution is corrected to obtain the posterior distribution.

[0176] In some optional embodiments, the posterior distribution determination subunit 1313 corrects the original posterior distribution, including: the posterior distribution determination subunit 1313 corrects the parameter mean of the photovoltaic power change curve; and the posterior distribution acquisition unit 131 obtains the posterior distribution corresponding to the photovoltaic power data according to the parameter mean of the corrected photovoltaic power change curve.

[0177] In some optional embodiments, the posterior distribution determination subunit 1313 corrects the parameter mean of the photovoltaic power variation curve using the following formula:

[0178] Wherein, ε is a preset correction coefficient, a1, b1, and c1 are parameters of the photovoltaic power change curve, u0 is the parameter mean of the photovoltaic power change curve before correction, and u1 is the parameter mean of the photovoltaic power change curve after correction. Optionally, ε = 0.03, or ε = -0.03.

[0179] The photovoltaic power generation prediction unit 132 is used to predict the photovoltaic power generation in the remaining time of the current time period to be predicted by using the posterior distribution of the photovoltaic power change curve.

[0180] The power supply device control module 14 is used to control the power supply device to adjust the power according to the photovoltaic power generation in the remaining time of the current time period to be predicted.

[0181] The photovoltaic power generation prediction device provided in this embodiment obtains the photovoltaic power change curve of the target photovoltaic module to be predicted and obtains the photovoltaic power data observed in the current time period to be predicted; the photovoltaic power generation in the remaining time of the current time period to be predicted is predicted based on the photovoltaic power change curve and the photovoltaic power data. The photovoltaic power generation is better fitted by the large amount of historical photovoltaic power generation data collected, and no complex calculation process and specific AI chips are required. It can significantly reduce the cost of use and solve problems such as frequent load shutdowns and insufficient photovoltaic utilization.

[0182] Based on the same inventive concept, Figure 11 As shown, the present application also provides a photovoltaic power generation controller, comprising:

[0183] At least one processor 121 and at least one memory 122 .

[0184] The memory stores executable instructions for the processor.

[0185] The processor is configured to execute the photovoltaic power generation prediction method provided by the above embodiment.

[0186] The photovoltaic power generation controller provided in the embodiment of the present application stores executable instructions of the processor in a memory. When the executable instructions are executed, the processor can obtain the photovoltaic power change curve of the target photovoltaic component to be predicted and obtain the photovoltaic power data observed in the current time period to be predicted; predict the photovoltaic power generation in the remaining time of the current time period to be predicted based on the photovoltaic power change curve and the photovoltaic power data, and better fit the predicted photovoltaic power generation by collecting a large amount of historical photovoltaic power generation data. Without the need for complex calculation processes and specific AI chips, it can significantly reduce the cost of use and solve problems such as frequent load shutdowns and insufficient photovoltaic utilization.

[0187] It should be noted that, in the description of this application, the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0188] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

Claims

1. A photovoltaic power generation prediction method, characterized in that: include: Acquire a photovoltaic power variation curve of a target photovoltaic assembly to be predicted, wherein the photovoltaic power variation curve is used to represent a variation pattern of the photovoltaic power of the target photovoltaic assembly to be predicted over time within a time period; Obtain the photovoltaic power data observed during the current forecast period; The photovoltaic power generation amount in the remaining time of the current time period to be predicted is predicted according to the photovoltaic power change curve and the photovoltaic power data.

2. The method according to claim 1, wherein: Predicting the photovoltaic power generation in the remaining time of the current predicted time period according to the photovoltaic power change curve and the photovoltaic power data includes: Obtaining a posterior distribution of the photovoltaic power change curve according to the Bayesian formula and the photovoltaic power data; The photovoltaic power generation in the remaining time of the current time period to be predicted is predicted using the posterior distribution of the photovoltaic power variation curve.

3. The method according to claim 2, wherein: Obtaining the posterior distribution of the photovoltaic power change curve according to the Bayesian formula and the photovoltaic power data includes: Obtaining a priori Gaussian distribution of the photovoltaic power variation curve; Determining a likelihood function corresponding to the photovoltaic power data; The posterior distribution of the photovoltaic power variation curve is obtained according to the Bayesian formula, the prior Gaussian distribution and the likelihood function.

4. The method according to claim 3, wherein: Obtaining the prior Gaussian distribution of the photovoltaic power variation curve includes: Obtaining a parameter mean of the photovoltaic power variation curve; Calculating a covariance matrix based on the parameter means; A priori Gaussian distribution of the photovoltaic power variation curve is obtained according to the parameter mean and the covariance matrix.

5. The method according to claim 4, characterized in that: Determining the likelihood function corresponding to the photovoltaic power data includes: Determining a function type corresponding to the photovoltaic power data; A likelihood function corresponding to the photovoltaic power data is determined based on the function type.

6. The method according to claim 5, characterized in that: Obtaining the posterior distribution of the photovoltaic power variation curve according to the Bayesian formula, the prior Gaussian distribution, and the likelihood function includes: Obtaining an original posterior distribution of the photovoltaic power variation curve according to the Bayesian formula, the prior Gaussian distribution, and the likelihood function; Obtaining the posterior coefficient of the original posterior distribution; Determining whether the original posterior distribution coefficient meets a preset error index; If the original posterior distribution coefficient meets the preset error index, the original posterior distribution is used as the posterior distribution; If the original posterior distribution coefficient does not meet the preset error index, the original posterior distribution is modified to obtain the posterior distribution.

7. The method according to claim 6, characterized in that: Modifying the original posterior distribution includes: Correcting the parameter mean of the photovoltaic power variation curve; The posterior distribution corresponding to the photovoltaic power data is obtained according to the parameter mean of the corrected photovoltaic power variation curve.

8. The method according to claim 6, wherein: The parameter mean of the photovoltaic power variation curve is corrected by the following formula: Among them, ε is the preset correction coefficient, a1, b1, and c1 are the parameters of the photovoltaic power change curve, u0 is the parameter mean of the photovoltaic power change curve before correction, and u1 is the parameter mean of the photovoltaic power change curve after correction.

9. The method according to any one of claims 1 to 8, characterized in that: Obtaining the photovoltaic power change curve of the target photovoltaic module to be predicted includes: Obtain historical photovoltaic power generation data of the target photovoltaic module to be predicted; The photovoltaic power variation curve is fitted according to the historical photovoltaic power generation data.

10. The method according to claim 9, characterized in that: Fitting the photovoltaic power change curve according to the historical photovoltaic power generation data includes: Segmenting the historical photovoltaic power generation data by date; The historical photovoltaic power generation data of each date is sampled by time period, and the maximum power generation in each time period is taken; Calculate the first-order difference data of the maximum power generation in each period; Obtain the upper bound and lower bound corresponding to the specified percentile interval of each of the first-order difference data; truncating each of the first-order differences according to a fixed interval; The photovoltaic power variation curve is obtained according to the truncation result and the maximum power generation in each time period.

11. The method according to claim 1, wherein: Obtaining the photovoltaic power data observed during the current forecast period includes: Determine whether photovoltaic power data has been generated within the current prediction time period; If not, the photovoltaic power data observed in the current time period to be predicted is obtained again after a delay of one time step, until the photovoltaic power data observed in the current time period to be predicted is obtained.

12. The method according to claim 4, wherein: The posterior distribution of the photovoltaic power variation curve is obtained by the following formula: p(θ|y)=p(θ)*p(y|θ)=N(u n ,S n ), among which: Among them, p(θ|y) is the posterior distribution of the photovoltaic power change curve, p(θ) is the prior Gaussian distribution of the photovoltaic power change curve, p(y|θ) is the likelihood function corresponding to the photovoltaic power data, and u n is the parameter mean of the photovoltaic power variation curve, Σ n is the covariance matrix of the photovoltaic power change curve, u0 is the mean of the historical photovoltaic power generation data curve, Σ0 is the covariance matrix of the historical photovoltaic power generation data curve, and X is a matrix.

13. The method according to claim 5, wherein: The function type corresponding to the photovoltaic power data is a quadratic function, and the likelihood function corresponding to the photovoltaic power data is determined by the following formula: Among them, p(y|θ) is the likelihood function corresponding to the photovoltaic power data, y t is the photovoltaic power generation power, x t is time, a, b, c are the parameters of the photovoltaic power change curve, σ 2 is the noise variance.

14. The method according to claim 12, wherein: Also includes: The power supply device is controlled to adjust power according to the photovoltaic power generation in the remaining time of the current predicted time period.

15. The method according to claim 8, wherein: ε=0.03, or ε=-0.

03.

16. A photovoltaic power generation prediction device, characterized by: include: A photovoltaic curve acquisition module is used to obtain a photovoltaic power variation curve of a target photovoltaic assembly to be predicted, wherein the photovoltaic power variation curve is used to represent a variation pattern of the photovoltaic power of the target photovoltaic assembly to be predicted over time within a time period; Photovoltaic data acquisition module, used to obtain photovoltaic power data observed during the current forecast period; The power generation prediction module is used to predict the photovoltaic power generation in the remaining time of the current time period to be predicted based on the photovoltaic power change curve and the photovoltaic power data.

17. A photovoltaic power generation controller, characterized in that: include: at least one processor and at least one memory; The memory stores executable instructions of the processor; The processor is configured to execute the photovoltaic power generation prediction method according to any one of claims 1 to 15.

Images