A method, device, equipment and storage medium for predicting output probability of photovoltaic power station

Abstract

The present application discloses a method, device, equipment and storage medium for predicting the output probability of a photovoltaic power station, and relates to the field of electrical digital data processing. The method utilizes the most important factors affecting the output of a photovoltaic array and directly matches and benchmarks cloud parameters with existing outputs. Compared with the prior art, the intermediate steps of model establishment, conversion, iteration and the like are omitted, and the actual cloud layer results are directly benchmarked and matched with the actual processing results, and the subsequent prediction process is completed in the form of results to results. There is no need to first analyze the cause of the previous result and then derive the second result from the cause, and the intermediate links are omitted, thereby avoiding errors caused by the intermediate links. At the same time, the present embodiment also interpolates and fills the missing output probability coefficients, so that subsequent predictions can be directly searched without the need for model calculations, so that the prediction of the output probability is accurate and concise.

Description

Technical Field

[0001] The present application relates to the technical field of electrical digital data processing, and in particular to a method, device, equipment and storage medium for predicting output probability of a photovoltaic power station. Background Art

[0002] Photovoltaic power station refers to a photovoltaic power generation system that uses solar energy and special materials such as crystalline silicon panels, inverters and other electronic components to connect to the power grid and transmit electricity to the power grid. A photovoltaic power station with a complete set of functions usually has one or more photovoltaic arrays for receiving solar energy, transformers for converting the electricity generated by the photovoltaic arrays, and energy storage devices electrically connected to the transformers. The energy storage devices store the electricity that reaches the required use after conversion, and transmit it to external power-consuming equipment through the municipal power grid. Due to the random characteristics of photovoltaic light, it is easy to cause fluctuations in the power grid of the power station, which has a certain impact on the generation, collection and transmission of electricity. Therefore, it is necessary to predict the output probability of the photovoltaic array in the photovoltaic power station to prepare countermeasures in advance for possible future power grid fluctuations.

[0003] At present, most of the main output probability forecasts use artificial intelligence models based on big data. For example, the quantile regression averaging method is used to integrate a set of independent deterministic predictions of long and short-term memory neural networks to obtain a probabilistic prediction model for photovoltaic output; the correlation of photovoltaic output at adjacent moments is considered to form a static Bayesian network, which is expanded to different time series to form a dynamic Bayesian network, and the probability of photovoltaic output status at each time node is calculated by the maximum likelihood estimation method; Bayesian models, Markov chain models and quantile regression models are established respectively, and a combined prediction method is proposed to predict photovoltaic output probability; probabilistic neural networks are used to divide weather types, and then the principal component analysis method is used to convert multiple meteorological factors affecting photovoltaic output into input vectors, and finally a short-term probability prediction model for photovoltaic power generation output based on support vector machines is established.

[0004] Some of the above models use historical output data for training and prediction, without considering the impact of meteorological factors on photovoltaic power generation output. Or, although some models consider the impact of meteorological factors on photovoltaic power generation output during the prediction process, the expression of the obtained probability prediction results is relatively complex and inconvenient for practical application. Summary of the invention

[0005] The main purpose of the present application is to provide a method, device, equipment and storage medium for predicting the output probability of a photovoltaic power station, so as to solve the problems in the prior art of inaccurate data input and output of the existing output probability prediction model and overly complicated prediction process.

[0006] To achieve the above objectives, this application provides the following technical solutions:

[0007] A method for predicting the output probability of a photovoltaic power station, wherein the photovoltaic power station includes a photovoltaic array installed in a preset area, and the method for predicting the output probability includes:

[0008] Step S1, obtaining a plurality of historical cloud thicknesses of the preset area at intervals of the preset time period through a preset strategy;

[0009] Step S2, based on the photovoltaic array, obtaining a first historical output under the historical maximum cloud thickness, a second historical output under unobstructed illumination, and a plurality of third historical outputs under any cloud thickness;

[0010] Step S3, constructing a historical output interval with the first historical output as the minimum value and the second historical output as the maximum value, and placing all third historical outputs into the historical output interval;

[0011] Step S4, normalizing the historical output interval to obtain a normalized interval between [0, 1], and defining all data in the normalized interval as an output probability coefficient;

[0012] Step S5, interpolating and filling all output probability coefficients in the normalized interval through a machine learning model to obtain a filled normalized interval;

[0013] Step S6, performing linear fitting with the cloud thickness as an independent variable and all output probability coefficients in the filled normalized interval as dependent variables to obtain an output probability function;

[0014] Step S7, obtaining the real-time cloud thickness or future cloud thickness of the preset area and substituting it into the output probability function to obtain the real-time output probability or future output probability.

[0015] As a further improvement of the present application, in step S4, the historical output interval is normalized to obtain a normalized interval between [0, 1], and all data in the normalized interval are defined as an output probability coefficient, including:

[0016] Step S41, normalize all historical outputs according to formula (1):

[0017] (1);

[0018] in, is the normalized value of all historical outputs, The value of all historical contributions, is the average of all historical outputs, The standard deviation of all historical contributions;

[0019] Step S42, integrating all normalized values ​​of historical output to obtain a normalized interval between [0, 1];

[0020] Step S43: defining all data in the normalized interval as the output probability coefficient.

[0021] As a further improvement of the present application, in step S5, all output probability coefficients in the normalized interval are interpolated and filled by a machine learning model to obtain a filled normalized interval, including:

[0022] Step S51, obtaining data missing points of all output probability coefficients in the normalized interval;

[0023] Step S52, taking the data missing point as the midpoint and at least one preset unit thickness as the cutting length to cut the normalized interval, so as to form at least one sample to be interpolated;

[0024] Step S53, taking two adjacent data missing points as end points and at least two preset unit thicknesses as cut-off lengths to cut off the normalized interval to form at least one training sample, wherein all training samples do not contain data missing points;

[0025] Step S54, defining an extreme learning model according to the data dimension of the training samples and training the extreme learning model through the training samples to obtain an interpolation model;

[0026] Step S55, outputting the sample to be interpolated to the interpolation model to obtain the historical output interpolation at the data missing point;

[0027] Step S56, filling all historical output interpolation values ​​into the normalized interval to obtain the filled normalized interval.

[0028] As a further improvement of the present application, step S51, obtaining data missing points of all output probability coefficients in the normalized interval, includes:

[0029] Step S511, generating a coordinate point in a preset rectangular coordinate system with the cloud thickness as the abscissa and the output probability coefficient as the ordinate;

[0030] Step S512, taking each coordinate point as a starting point, extending in the positive direction or the negative direction of the horizontal axis of the preset rectangular coordinate system by at most one preset unit thickness to form a unilateral extension area;

[0031] Step S513, determining whether there are other coordinate points in the one-side extension area of ​​the current coordinate point, if there are no other coordinate points in the one-side extension area of ​​the current coordinate point, executing step S514;

[0032] Step S514: define the farthest end of the one-side extension area of ​​the current coordinate point from the current coordinate point as the data missing point.

[0033] As a further improvement of the present application, step S54, defining an extreme learning model according to the data dimension of the training sample and training the extreme learning model through the training sample to obtain an interpolation model, includes:

[0034] Step S541, defining the training sample as , the number of training samples is ;

[0035] in, , For the The thickness of the cloud layer, for of A collection of n-dimensional vectors; , For the The output probability coefficient corresponding to the cloud thickness is: for of A collection of n-dimensional vectors;

[0036] Step S542, defining the extreme learning model according to formula (2):

[0037] (2);

[0038] in, , is the first connection weight between the hidden layer and the output layer of the extreme learning model; is the number of nodes in the hidden layer; is the output value of the hidden layer, ; ; is the activation function; is the input layer of the extreme learning model and the hidden layer The second connection weight of the nodes; is the hidden layer The threshold of nodes;

[0039] Step S543, randomly generating the second connection weight and the threshold value by a nonlinear mapping function in a feature space based on the extreme learning model;

[0040] Step S544, obtaining the output value of the hidden layer by calculating the randomly generated second connection weight and the threshold;

[0041] Step S545: according to formula (3), the optimal solution of the first connection weight is obtained by minimizing the approximate square error:

[0042] (3);

[0043] in, is the output matrix of the hidden layer; is the target matrix of the training sample; is the optimal solution; The output matrix The Moore-Penrose generalized inverse matrix of ;

[0044] Step S546, obtaining the optimal solution and the second connection weight and threshold corresponding to the optimal solution and substituting them into the extreme learning model to obtain the interpolation model.

[0045] As a further improvement of the present application, step S7, obtaining the real-time cloud thickness or future cloud thickness of the preset area and substituting it into the output probability function to obtain the real-time output probability or future output probability, then includes:

[0046] Step S10, determining whether the real-time output probability is less than a preset output threshold, if the real-time output probability is less than the preset output threshold, executing step S20;

[0047] Step S20, switching the photovoltaic array to an off-grid state.

[0048] As a further improvement of the present application, step S7, obtaining the real-time cloud thickness or future cloud thickness of the preset area and substituting it into the output probability function to obtain the real-time output probability or future output probability, then includes:

[0049] Step S100, determining whether the future output probability is less than the preset output threshold, if the future output probability is less than the preset output threshold, executing step S200;

[0050] Step S200, obtaining the occurrence timestamp of the future output probability;

[0051] Step S300, when the occurrence timestamp arrives, switching the photovoltaic array to an off-grid state.

[0052] To achieve the above objectives, this application also provides the following technical solutions:

[0053] A power output probability prediction device for a photovoltaic power station, the power output probability prediction device is applied to the power output probability prediction method as described above, and the power output probability prediction device comprises:

[0054] A historical cloud thickness acquisition module, used to acquire a plurality of historical cloud thicknesses of the preset area at the preset time interval according to a preset strategy;

[0055] A historical output acquisition module, used for acquiring, based on the photovoltaic array, a first historical output under the historical maximum cloud thickness, a second historical output under unobstructed illumination, and a plurality of third historical outputs under any cloud thickness;

[0056] A historical output interval construction module, configured to construct a historical output interval with the first historical output as a minimum value and the second historical output as a maximum value, and put all third historical outputs into the historical output interval;

[0057] An output probability coefficient definition module is used to normalize the historical output interval to obtain a normalized interval between [0, 1], and define all data in the normalized interval as an output probability coefficient;

[0058] A filled normalized interval acquisition module is used to interpolate and fill all output probability coefficients in the normalized interval through a machine learning model to obtain a filled normalized interval;

[0059] An output probability function acquisition module is used to perform linear fitting with the cloud thickness as an independent variable and all output probability coefficients of the filled normalized interval as dependent variables to obtain an output probability function;

[0060] The output probability acquisition module is used to obtain the real-time cloud thickness or future cloud thickness of the preset area and substitute it into the output probability function to obtain the real-time output probability or future output probability.

[0061] To achieve the above objectives, this application also provides the following technical solutions:

[0062] An electronic device includes a processor and a memory coupled to the processor, wherein the memory stores program instructions executable by the processor; when the processor executes the program instructions stored in the memory, the output probability prediction method as described above is implemented.

[0063] To achieve the above objectives, this application also provides the following technical solutions:

[0064] A storage medium stores program instructions, which, when executed by a processor, can implement the above-mentioned output probability prediction method.

[0065] The present application obtains several historical cloud thicknesses in a preset area at preset strategy intervals and preset time periods; obtains the first historical output under the historical maximum cloud thickness, the second historical output under unobstructed vision, and several third historical outputs under arbitrary cloud thickness based on the photovoltaic array; constructs a historical output interval with the first historical output as the minimum value and the second historical output as the maximum value, and puts all third historical outputs into the historical output interval; normalizes the historical output interval to obtain a normalized interval based on [0,1], and defines all data in the normalized interval as an output probability coefficient; interpolates and fills all output probability coefficients in the normalized interval through a machine learning model to obtain a filled normalized interval; performs linear fitting with cloud thickness as the independent variable and all output probability coefficients in the filled normalized interval as the dependent variable to obtain an output probability function; obtains the real-time cloud thickness or future cloud thickness in the preset area and substitutes it into the output probability function to obtain the real-time output probability or future output probability. The present application utilizes the most important factor that affects the output of photovoltaic arrays (large-area light obstruction caused by clouds) and directly matches and benchmarks the cloud parameters with the existing output. Compared with the existing technology, it omits the intermediate steps of model establishment, conversion, iteration, etc., directly matches the actual cloud results with the actual processing results, and completes the subsequent prediction process in the form of results against results. There is no need to first analyze the cause of the previous result and then derive the second result from the cause. The intermediate links are omitted, thereby avoiding the errors caused by the intermediate links. The present application also interpolates and fills in the missing output probability coefficients, so that subsequent predictions can be directly searched without the need for model calculations, making the output probability prediction accurate and concise. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A schematic diagram of the steps of an embodiment of a method for predicting the output probability of a photovoltaic power station according to the present application;

[0067] Figure 2 This is a structural schematic diagram of an embodiment of an output probability prediction device for a photovoltaic power station of the present application;

[0068] Figure 3 This is a schematic diagram of the structure of an embodiment of the electronic device of the present application;

[0069] Figure 4 This is a schematic diagram of the structure of an embodiment of the storage medium of the present application. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0071] The terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of this application are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0072] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0073] like Figure 1 As shown, this embodiment provides an embodiment of a method for predicting the output probability of a photovoltaic power station, and the photovoltaic power station includes a photovoltaic array installed in a preset area.

[0074] Specifically, the output probability prediction method includes the following steps:

[0075] Step S1, obtaining several historical cloud thicknesses of a preset area at preset time intervals through a preset strategy.

[0076] Preferably, the preset time period can be set according to the speed of change of the cloud morphology, such as 1min, 5min, 10min, 15min, 30min, etc.

[0077] Preferably, the preset strategy can detect the cloud thickness through the weather radar and specific measuring instruments in the prior art.

[0078] Among them, the weather radar is a device that can detect weather systems within a few hundred kilometers around the radar station, and has functions such as detecting cloud thickness. Through the radar's detection of cloud thickness, the echo signal strength of the cloud can be obtained, which depends on the thickness and size of the cloud. Measuring instruments, such as cloud height meters, can not only measure and identify cloud height, cloud thickness, and cloud quantity, but also reflect vertical visibility. These instruments emit high-speed laser pulse sequences and automatically collect data, use FPGA (field programmable gate array) to store discrete data, and process it through embedded processors.

[0079] It should be noted that the acquisition of cloud thickness is a mature existing technology. The acquisition of cloud thickness in this embodiment is a conventional application, so the detailed acquisition process of cloud thickness is not repeated in this embodiment.

[0080] Step S2, based on the photovoltaic array, obtain a first historical output under the historical maximum cloud thickness, a second historical output under unobstructed direct sunlight, and a plurality of third historical outputs under arbitrary cloud thicknesses.

[0081] Preferably, the output refers to the amount of electricity that can be output by a photovoltaic power station or array, and this data can be directly obtained without difficulty.

[0082] Step S3, constructing a historical output interval with the first historical output as the minimum value and the second historical output as the maximum value, and putting all third historical outputs into the historical output interval.

[0083] Step S4, normalizing the historical output interval to obtain a normalized interval between [0, 1], and defining all data in the normalized interval as an output probability coefficient.

[0084] Preferably, the design intention of this embodiment is to define the output probability coefficient when there is no obstruction and full sunlight as 1.

[0085] Step S5, interpolating and filling all output probability coefficients in the normalized interval through a machine learning model to obtain a filled normalized interval.

[0086] Step S6, performing linear fitting with the cloud thickness as the independent variable and all output probability coefficients in the filled normalized interval as the dependent variable to obtain the output probability function.

[0087] Step S7, obtaining the real-time cloud thickness or future cloud thickness of the preset area and substituting it into the power output probability function to obtain the real-time power output probability or future power output probability.

[0088] Furthermore, in step S4, the historical output interval is normalized to obtain a normalized interval between [0, 1], and all data in the normalized interval are defined as output probability coefficients, including:

[0089] Step S41, normalize all historical outputs according to formula (1):

[0090] (1).

[0091] in, is the normalized value of all historical outputs, The value of all historical contributions, is the average of all historical outputs, The standard deviation of all historical contributions.

[0092] Preferably, this embodiment prefers the normalization method of zero-mean normalization (Z-score normalization), which gives the mean and standard deviation of the original data to standardize the data, and the processed data conforms to the standard normal distribution, that is, the mean is 0 and the standard deviation is 1. For the normalization method, this embodiment can also use batch normalization. Compared with simple normalization, in the previous neural network training, only the input layer data is normalized, but not in the middle layer. Although the data set of the input node is normalized, the data distribution of the input data after matrix multiplication is likely to change greatly, and as the number of network layers of machine learning continues to deepen, the change in data distribution will become greater and greater. Therefore, the normalization processing performed by batch normalization in the middle layer of machine learning makes the training effect better.

[0093] Step S42: Integrate all normalized historical output values ​​to obtain a normalized interval between [0, 1].

[0094] Step S43, defining all data in the normalized interval as output probability coefficients.

[0095] Further, in step S5, all output probability coefficients in the normalized interval are interpolated and filled by a machine learning model to obtain a filled normalized interval, including:

[0096] Step S51, obtaining data missing points of all output probability coefficients in the normalized interval.

[0097] Step S52, taking the data missing point as the midpoint and at least one preset unit thickness as the cutting length to cut off a normalized interval, so as to form at least one sample to be interpolated.

[0098] Step S53, taking two adjacent data missing points as end points and at least two preset unit thicknesses as cut-off lengths to cut off a normalized interval to form at least one training sample, and all training samples do not contain data missing points.

[0099] Step S54, defining an extreme learning model according to the data dimension of the training samples and training the extreme learning model through the training samples to obtain an interpolation model.

[0100] Preferably, the extreme learning model is a feedforward neural network that can be used to train a single hidden layer. The extreme learning machine is different from the traditional single hidden layer feedforward neural network training algorithm. The extreme learning machine randomly selects the input layer weights and hidden layer biases, and the output layer weights are calculated and analyzed based on the Moore-Penrose (MP) generalized inverse matrix theory by minimizing the loss function composed of the training error term and the regularization term of the output layer weight norm. Even if the extreme learning machine randomly generates hidden layer nodes, the extreme learning machine still maintains the general approximation ability of a single hidden layer feedforward neural network.

[0101] Step S55: output the samples to be interpolated to the interpolation model to obtain the historical output interpolation at the data missing point.

[0102] Preferably, a sample to be interpolated may contain only one data missing point. In the actual simulation operation process, if a sample to be interpolated contains several data missing points, it is easy to cause excessive computing power burden.

[0103] Step S56, all historical output interpolation values ​​are filled into the normalized interval to obtain a filled normalized interval.

[0104] Furthermore, in step S51, data missing points of all output probability coefficients in the normalized interval are obtained, including:

[0105] Step S511, generating coordinate points in a preset rectangular coordinate system with the cloud thickness as the horizontal coordinate and the output probability coefficient as the vertical coordinate.

[0106] Step S512 , taking each coordinate point as a starting point, extending in a positive direction or a negative direction of the horizontal axis of the preset rectangular coordinate system by at most a preset unit thickness to form a single-side extension area.

[0107] Step S513, determining whether there are other coordinate points in the one-side extension area of ​​the current coordinate point. If there are no other coordinate points in the one-side extension area of ​​the current coordinate point, executing step S514.

[0108] Step S514: define the farthest end of the one-side extension area of ​​the current coordinate point from the current coordinate point as a data missing point.

[0109] Further, step S54, defining an extreme learning model according to the data dimension of the training samples and training the extreme learning model through the training samples to obtain an interpolation model, includes:

[0110] Step S541, define the training sample as , the number of training samples is .

[0111] in, , For the The thickness of the cloud layer, for of A collection of n-dimensional vectors; , For the The output probability coefficient corresponding to the cloud thickness is: for of A collection of dimensional vectors.

[0112] Step S542, define the extreme learning model according to formula (2):

[0113] (2).

[0114] in, , is the first connection weight between the hidden layer and the output layer of the extreme learning model; is the number of nodes in the hidden layer; is the output value of the hidden layer, ; ; is the activation function; is the input layer and the hidden layer of the extreme learning model The second connection weight of the nodes; The hidden layer The threshold of nodes.

[0115] Preferably, It can be set to activation functions such as Sigmoid function and Gaussian function.

[0116] Preferably, if the Sigmoid function is selected, then It can be expressed as:

[0117] , here .

[0118] It should be noted that the above The formula is only used to explain the principle of the Sigmoid function, and its symbolic meaning is not interoperable with other parts of the embodiment.

[0119] Step S543: randomly generate the second connection weight and threshold value through a nonlinear mapping function in the feature space based on the extreme learning model.

[0120] Step S544, obtaining the output value of the hidden layer through the randomly generated second connection weight and threshold calculation.

[0121] Step S545, according to formula (3), the optimal solution of the first connection weight is obtained by minimizing the approximate square error:

[0122] (3).

[0123] in, is the output matrix of the hidden layer; is the target matrix of the training samples; is the optimal solution; The output matrix The Moore-Penrose generalized inverse matrix of .

[0124] Step S546, obtaining the optimal solution and the second connection weight and threshold corresponding to the optimal solution and substituting them into the extreme learning model to obtain the interpolation model.

[0125] Preferably, the Moore-Penrose generalized inverse matrix can be calculated by singular value decomposition .

[0126] It can be understood that the training of a single hidden layer feedforward neural network by an extreme learning machine is divided into two main stages:

[0127] ① Random feature mapping:

[0128] The hidden layer parameters are randomly initialized, and then some nonlinear mapping is used as the activation function to map the input data to a new feature space (called the extreme learning machine feature space). In simple terms, the weights and biases on the hidden layer nodes of the extreme learning machine are randomly generated. The random feature mapping stage is different from many existing learning algorithms, such as support vector machines that use kernel functions for feature mapping, restricted Boltzmann machines in deep neural networks, and autoencoders / autodecoders for feature learning. The nonlinear mapping function in the extreme learning machine can be any nonlinear piecewise continuous function. In the extreme learning machine, the hidden layer node parameters are randomly generated according to any continuous probability distribution and are independent of the training data.

[0129] ②Linear parameter solution:

[0130] After the first stage, the second connection weight and threshold are randomly generated. In the second stage, only the first connection weight needs to be solved. In order to obtain the first connection weight with good effect on the training sample set, it is necessary to ensure that its training error is minimized. The target matrix of the training samples Minimize the square error as the objective function for evaluating the training error, and then find the solution with the smallest objective function is the optimal solution.

[0131] Preferably, the eigenspace is a vector space that represents a special property of a matrix. The eigenspace is a vector space consisting of all eigenvectors of the matrix. For example, for an n×n matrix A, if there exists a scalar λ and a non-zero vector x such that Ax=λx, then x is the eigenvector of A, and the vector space where the vector x is located is the eigenspace of A.

[0132] It should be noted that the meanings of the symbols in the above vector space related explanations are not interoperable with other parts of the embodiment.

[0133] Further, in step S7, the real-time cloud thickness or future cloud thickness of the preset area is obtained and substituted into the output probability function to obtain the real-time output probability or future output probability, and then, the following steps are included:

[0134] Step S10, determining whether the real-time output probability is less than a preset output threshold. If the real-time output probability is less than the preset output threshold, executing step S20.

[0135] Step S20, switching the photovoltaic array to an off-grid state.

[0136] Further, in step S7, the real-time cloud thickness or future cloud thickness of the preset area is obtained and substituted into the output probability function to obtain the real-time output probability or future output probability, and then, the following steps are included:

[0137] Step S100, determining whether the future power output probability is less than a preset power output threshold, if the future power output probability is less than the preset power output threshold, executing step S200.

[0138] Step S200, obtaining the occurrence timestamp of the future output probability.

[0139] Step S300, when a timestamp occurs, the photovoltaic array is switched to an off-grid state.

[0140] This embodiment obtains several historical cloud thicknesses in a preset area at preset strategy intervals and preset time periods; obtains the first historical output under the historical maximum cloud thickness, the second historical output under unobstructed vision, and several third historical outputs under any cloud thickness based on the photovoltaic array; constructs a historical output interval with the first historical output as the minimum value and the second historical output as the maximum value, and puts all the third historical outputs into the historical output interval; normalizes the historical output interval to obtain a normalized interval based on [0,1], and defines all data in the normalized interval as an output probability coefficient; interpolates and fills all output probability coefficients in the normalized interval through a machine learning model to obtain a filled normalized interval; linear fitting is performed with cloud thickness as the independent variable and all output probability coefficients in the filled normalized interval as the dependent variable to obtain an output probability function; obtains the real-time cloud thickness or future cloud thickness in the preset area and substitutes it into the output probability function to obtain the real-time output probability or the future output probability. This embodiment utilizes the most important factor that affects the output of the photovoltaic array (large-area light blocking caused by clouds) and directly matches and benchmarks the cloud parameters with the existing output. Compared with the existing technology, it omits the intermediate steps of model establishment, conversion, iteration, etc., directly matches the actual cloud results with the actual processing results, and completes the subsequent prediction process in the form of results against results. There is no need to first analyze the cause of the previous result and then derive the second result from the cause. The intermediate links are omitted, thereby avoiding the errors caused by the intermediate links. This embodiment also interpolates and fills the missing output probability coefficients, so that subsequent predictions can be directly searched without the need for model calculations, making the prediction of the output probability accurate and concise.

[0141] like Figure 2 As shown, this embodiment provides an embodiment of an output probability prediction device for a photovoltaic power station. In this embodiment, the output probability prediction device is applied to the output probability prediction method in the above embodiment.

[0142] Specifically, the output probability prediction device includes a historical cloud thickness acquisition module 1, a historical output acquisition module 2, a historical output interval construction module 3, an output probability coefficient definition module 4, a filled normalized interval acquisition module 5, an output probability function acquisition module 6, and an output probability acquisition module 7, which are electrically connected in sequence.

[0143] Among them, the historical cloud thickness acquisition module 1 is used to obtain several historical cloud thicknesses in a preset area at a preset time period through a preset strategy interval; the historical output acquisition module 2 is used to obtain the first historical output under the historical maximum cloud thickness, the second historical output under unobstructed vision, and several third historical outputs under any cloud thickness based on the photovoltaic array; the historical output interval construction module 3 is used to construct a historical output interval with the first historical output as the minimum value and the second historical output as the maximum value, and put all the third historical outputs into the historical output interval; the output probability coefficient definition module 4 is used to normalize the historical output interval to obtain Based on the normalized interval between [0,1], all data in the normalized interval are defined as output probability coefficients; the filled normalized interval acquisition module 5 is used to interpolate and fill all output probability coefficients in the normalized interval through a machine learning model to obtain the filled normalized interval; the output probability function acquisition module 6 is used to perform linear fitting with cloud thickness as the independent variable and all output probability coefficients in the filled normalized interval as the dependent variable to obtain the output probability function; the output probability acquisition module 7 is used to obtain the real-time cloud thickness or future cloud thickness of the preset area and substitute it into the output probability function to obtain the real-time output probability or the future output probability.

[0144] Furthermore, the output probability coefficient definition module 4 specifically includes a first output probability coefficient definition submodule, a second output probability coefficient definition submodule, and a third output probability coefficient definition submodule, which are electrically connected in sequence; the first output probability coefficient definition submodule is electrically connected to the historical output interval construction module, and the third output probability coefficient definition submodule is electrically connected to the filled normalized interval acquisition module.

[0145] Among them, the first output probability coefficient definition submodule is used to normalize all historical outputs according to formula (1):

[0146] (1).

[0147] in, is the normalized value of all historical outputs, The value of all historical contributions, is the average of all historical outputs, The standard deviation of all historical contributions.

[0148] The second output probability coefficient definition submodule is used to integrate all normalized values ​​of historical outputs to obtain a normalized interval between [0, 1].

[0149] The third output probability coefficient definition submodule is used to define all data in the normalized interval as output probability coefficients.

[0150] Furthermore, the filled normalized interval acquisition module 5 specifically includes a first filled normalized interval acquisition submodule, a second filled normalized interval acquisition submodule, a third filled normalized interval acquisition submodule, a fourth filled normalized interval acquisition submodule, a fifth filled normalized interval acquisition submodule, and a sixth filled normalized interval acquisition submodule, which are electrically connected in sequence; the first filled normalized interval acquisition submodule is electrically connected to the third output probability coefficient definition submodule, and the sixth filled normalized interval acquisition submodule is electrically connected to the output probability function acquisition module.

[0151] Among them, the first filled normalized interval acquisition submodule is used to obtain the data missing points of all output probability coefficients in the normalized interval; the second filled normalized interval acquisition submodule is used to intercept the normalized interval with the data missing point as the midpoint and at least one preset unit thickness as the interception length to form at least one sample to be interpolated; the third filled normalized interval acquisition submodule is used to intercept the normalized interval with two adjacent data missing points as the two end points and at least two preset unit thicknesses as the interception length to form at least one training sample, and all training samples do not contain data missing points; the fourth filled normalized interval acquisition submodule is used to define the extreme learning model according to the data dimension of the training sample and train the extreme learning model through the training sample to obtain the interpolation model; the fifth filled normalized interval acquisition submodule is used to output the sample to be interpolated to the interpolation model to obtain the historical output interpolation located at the data missing point; the sixth filled normalized interval acquisition submodule is used to fill all historical output interpolations into the normalized interval to obtain the filled normalized interval.

[0152] Furthermore, the first filled normalized interval acquisition submodule specifically includes a first filled normalized interval acquisition unit, a second filled normalized interval acquisition unit, a third filled normalized interval acquisition unit, and a fourth filled normalized interval acquisition unit which are electrically connected in sequence; the first filled normalized interval acquisition unit is electrically connected to the third output probability coefficient definition submodule, and the fourth filled normalized interval acquisition unit is electrically connected to the second filled normalized interval acquisition submodule.

[0153] Among them, the first filled normalized interval acquisition unit is used to generate coordinate points in a preset rectangular coordinate system with cloud thickness as the horizontal coordinate and output probability coefficient as the vertical coordinate; the second filled normalized interval acquisition unit is used to take each coordinate point as the starting point, and extend to the positive or negative direction of the horizontal axis of the preset rectangular coordinate system by at most one preset unit thickness to form a unilateral extension area; the third filled normalized interval acquisition unit is used to determine whether there are other coordinate points in the unilateral extension area of ​​the current coordinate point; the fourth filled normalized interval acquisition unit is used to define the farthest end of the unilateral extension area of ​​the current coordinate point from the current coordinate point as a data missing point if there are no other coordinate points in the unilateral extension area of ​​the current coordinate point.

[0154] Further, the fourth filled normalized interval acquisition submodule specifically includes a fifth filled normalized interval acquisition unit, a sixth filled normalized interval acquisition unit, a seventh filled normalized interval acquisition unit, an eighth filled normalized interval acquisition unit, a ninth filled normalized interval acquisition unit, and a tenth filled normalized interval acquisition unit, which are electrically connected in sequence; the fifth filled normalized interval acquisition unit is electrically connected to the third filled normalized interval acquisition submodule, and the tenth filled normalized interval acquisition unit is electrically connected to the fifth filled normalized interval acquisition submodule.

[0155] The fifth filled normalized interval acquisition unit is used to define the training sample as , the number of training samples is .

[0156] in, , For the The thickness of the cloud layer, for of A collection of n-dimensional vectors; , For the The output probability coefficient corresponding to the cloud thickness is: for of A collection of dimensional vectors.

[0157] The sixth filled normalized interval acquisition unit is used to define the extreme learning model according to formula (2):

[0158] (2).

[0159] in, , is the first connection weight between the hidden layer and the output layer of the extreme learning model; is the number of nodes in the hidden layer; is the output value of the hidden layer, ; ; is the activation function; is the input layer and the hidden layer of the extreme learning model The second connection weight of the nodes; The hidden layer The threshold of nodes.

[0160] The seventh filled normalized interval acquisition unit is used to randomly generate the second connection weight and the threshold value through a nonlinear mapping function in the feature space based on the extreme learning model.

[0161] The eighth filled normalized interval acquisition unit is used to obtain the output value of the hidden layer through the randomly generated second connection weight and threshold calculation.

[0162] The ninth filled normalized interval acquisition unit is used to solve the optimal solution of the first connection weight by minimizing the approximate square error according to formula (3):

[0163] (3).

[0164] in, is the output matrix of the hidden layer; is the target matrix of the training samples; is the optimal solution; The output matrix The Moore-Penrose generalized inverse matrix of .

[0165] The tenth filled normalized interval acquisition unit is used to acquire the optimal solution and the second connection weight and threshold corresponding to the optimal solution and substitute them into the extreme learning model to obtain the interpolation model.

[0166] Furthermore, the output probability prediction device also includes a real-time output probability judgment module and a photovoltaic array state switching module which are electrically connected in sequence; the real-time output probability judgment module is electrically connected to the output probability acquisition module 7 .

[0167] Among them, the real-time output probability judgment module is used to judge whether the real-time output probability is less than the preset output threshold; the photovoltaic array state switching module is used to switch the photovoltaic array to the off-grid state if the real-time output probability is less than the preset output threshold.

[0168] Furthermore, a timestamp acquisition module is electrically connected between the real-time output probability judgment module and the photovoltaic array state switching module.

[0169] Moreover, the real-time output probability judgment module is also used to judge whether the future output probability is less than the preset output threshold; the occurrence timestamp acquisition module is used to obtain the occurrence timestamp of the future output probability if the future output probability is less than the preset output threshold; the photovoltaic array state switching module is also used to switch the photovoltaic array to the off-grid state when the occurrence timestamp arrives.

[0170] It should be noted that this embodiment is an apparatus embodiment based on the above method embodiment. The preferences, extensions, limitations, examples, and principle descriptions of this embodiment may refer to the above embodiments, and will not be described in detail in this embodiment.

[0171] This embodiment obtains several historical cloud thicknesses in a preset area at preset strategy intervals and preset time periods; obtains the first historical output under the historical maximum cloud thickness, the second historical output under unobstructed vision, and several third historical outputs under any cloud thickness based on the photovoltaic array; constructs a historical output interval with the first historical output as the minimum value and the second historical output as the maximum value, and puts all the third historical outputs into the historical output interval; normalizes the historical output interval to obtain a normalized interval based on [0,1], and defines all data in the normalized interval as an output probability coefficient; interpolates and fills all output probability coefficients in the normalized interval through a machine learning model to obtain a filled normalized interval; linear fitting is performed with cloud thickness as the independent variable and all output probability coefficients in the filled normalized interval as the dependent variable to obtain an output probability function; obtains the real-time cloud thickness or future cloud thickness in the preset area and substitutes it into the output probability function to obtain the real-time output probability or the future output probability. This embodiment utilizes the most important factor that affects the output of the photovoltaic array (large-area light blocking caused by clouds) and directly matches and benchmarks the cloud parameters with the existing output. Compared with the existing technology, it omits the intermediate steps of model establishment, conversion, iteration, etc., directly matches the actual cloud results with the actual processing results, and completes the subsequent prediction process in the form of results against results. There is no need to first analyze the cause of the previous result and then derive the second result from the cause. The intermediate links are omitted, thereby avoiding the errors caused by the intermediate links. This embodiment also interpolates and fills the missing output probability coefficients, so that subsequent predictions can be directly searched without the need for model calculations, making the prediction of the output probability accurate and concise.

[0172] Figure 3 Schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 3 As shown, the electronic device 8 includes a processor 81 and a memory 82 coupled to the processor 81 .

[0173] The memory 82 stores program instructions for implementing a method for predicting output probability of a photovoltaic power station according to any one of the above embodiments.

[0174] The processor 81 is used to execute the program instructions stored in the memory 82 to perform output probability prediction of the photovoltaic power station.

[0175] The processor 81 may also be referred to as a CPU (Central Processing Unit). The processor 81 may be an integrated circuit chip having the ability to process signals. The processor 81 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0176] Further, Figure 4 This is a schematic diagram of the structure of a storage medium according to an embodiment of the present application. Figure 4 The storage medium 9 of the embodiment of the present application stores program instructions 91 that can implement all the above methods, wherein the program instructions 91 can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0177] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0178] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. The above is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of this application, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A method for predicting the output probability of a photovoltaic power station, wherein the photovoltaic power station comprises a photovoltaic array installed in a preset area, characterized in that: The output probability prediction method comprises: Step S1, obtaining a plurality of historical cloud thicknesses of the preset area at preset time intervals through a preset strategy; Step S2, based on the photovoltaic array, obtaining a first historical output under the historical maximum cloud thickness, a second historical output under unobstructed illumination, and a plurality of third historical outputs under any cloud thickness; Step S3, constructing a historical output interval with the first historical output as the minimum value and the second historical output as the maximum value, and placing all third historical outputs into the historical output interval; Step S4, normalizing the historical output interval to obtain a normalized interval between [0, 1], and defining all data in the normalized interval as an output probability coefficient; Step S5, interpolating and filling all output probability coefficients in the normalized interval through a machine learning model to obtain a filled normalized interval; Step S6, performing linear fitting with the cloud thickness as an independent variable and all output probability coefficients in the filled normalized interval as dependent variables to obtain an output probability function; Step S7, obtaining the real-time cloud thickness or future cloud thickness of the preset area and substituting it into the output probability function to obtain the real-time output probability or future output probability; The step S5, interpolating and filling all output probability coefficients in the normalized interval by a machine learning model to obtain a filled normalized interval, includes: Step S51, obtaining data missing points of all output probability coefficients in the normalized interval; Step S52, taking the data missing point as the midpoint and at least one preset unit thickness as the cutting length to cut the normalized interval, so as to form at least one sample to be interpolated; Step S53, taking two adjacent data missing points as end points and at least two preset unit thicknesses as cut-off lengths to cut off the normalized interval to form at least one training sample, wherein all training samples do not contain data missing points; Step S54, defining an extreme learning model according to the data dimension of the training samples and training the extreme learning model through the training samples to obtain an interpolation model; Step S55, outputting the sample to be interpolated to the interpolation model to obtain the historical output interpolation at the data missing point; Step S56, filling all historical output interpolation values ​​into the normalized interval to obtain the filled normalized interval; The step S51, obtaining data missing points of all output probability coefficients in the normalized interval, includes: Step S511, generating a coordinate point in a preset rectangular coordinate system with the cloud thickness as the abscissa and the output probability coefficient as the ordinate; Step S512, taking each coordinate point as a starting point, extending in the positive direction or the negative direction of the horizontal axis of the preset rectangular coordinate system by at most one preset unit thickness to form a unilateral extension area; Step S513, determining whether there are other coordinate points in the one-side extension area of ​​the current coordinate point, if there are no other coordinate points in the one-side extension area of ​​the current coordinate point, executing step S514; Step S514: define the farthest end of the one-side extension area of ​​the current coordinate point from the current coordinate point as the data missing point.

2. The output probability prediction method according to claim 1, characterized in that: Step S4, normalizing the historical output interval to obtain a normalized interval between [0, 1], and defining all data in the normalized interval as an output probability coefficient, including: Step S41, normalize all historical outputs according to formula (1): （1）； in, is the normalized value of all historical outputs, The value of all historical contributions, is the average of all historical outputs, The standard deviation of all historical contributions; Step S42, integrating all normalized values ​​of historical output to obtain a normalized interval between [0, 1]; Step S43: defining all data in the normalized interval as the output probability coefficient.

3. The output probability prediction method according to claim 2, characterized in that: Step S54, defining an extreme learning model according to the data dimension of the training sample, and training the extreme learning model through the training sample to obtain an interpolation model, including: Step S541, defining the training sample as , the number of training samples is ; in, , For the The thickness of the cloud layer, for of A collection of n-dimensional vectors; , For the The output probability coefficient corresponding to the cloud thickness is: for of A collection of n-dimensional vectors; Step S542, defining the extreme learning model according to formula (2): （2）； in, , is the first connection weight between the hidden layer and the output layer of the extreme learning model; is the number of nodes in the hidden layer; is the output value of the hidden layer, ; ; is the activation function; is the input layer of the extreme learning model and the hidden layer The second connection weight of the nodes; is the hidden layer The threshold of nodes; Step S543, randomly generating the second connection weight and the threshold value by a nonlinear mapping function in a feature space based on the extreme learning model; Step S544, obtaining the output value of the hidden layer by calculating the randomly generated second connection weight and the threshold; Step S545: according to formula (3), the optimal solution of the first connection weight is obtained by minimizing the approximate square error: （3）； in, is the output matrix of the hidden layer; is the target matrix of the training sample; is the optimal solution; The output matrix The Moore-Penrose generalized inverse matrix of ; Step S546, obtaining the optimal solution and the second connection weight and threshold corresponding to the optimal solution and substituting them into the extreme learning model to obtain the interpolation model.

4. The output probability prediction method according to claim 1, characterized in that: Step S7, obtaining the real-time cloud thickness or future cloud thickness of the preset area and substituting it into the output probability function to obtain the real-time output probability or future output probability, and then comprising: Step S10, determining whether the real-time output probability is less than a preset output threshold, if the real-time output probability is less than the preset output threshold, executing step S20; Step S20, switching the photovoltaic array to an off-grid state.

5. The output probability prediction method according to claim 4, characterized in that: Step S7, obtaining the real-time cloud thickness or future cloud thickness of the preset area and substituting it into the output probability function to obtain the real-time output probability or future output probability, and then comprising: Step S100, determining whether the future output probability is less than the preset output threshold, if the future output probability is less than the preset output threshold, executing step S200; Step S200, obtaining the occurrence timestamp of the future output probability; Step S300, when the occurrence timestamp arrives, switching the photovoltaic array to an off-grid state.

6. A photovoltaic power station output probability prediction device, the output probability prediction device is applied to the output probability prediction method according to any one of claims 1 to 5, characterized in that: The output probability prediction device comprises: A historical cloud thickness acquisition module, used to acquire a plurality of historical cloud thicknesses of the preset area at the preset time interval according to a preset strategy; A historical output acquisition module, used for acquiring, based on the photovoltaic array, a first historical output under the historical maximum cloud thickness, a second historical output under unobstructed illumination, and a plurality of third historical outputs under any cloud thickness; A historical output interval construction module, configured to construct a historical output interval with the first historical output as a minimum value and the second historical output as a maximum value, and put all third historical outputs into the historical output interval; An output probability coefficient definition module is used to normalize the historical output interval to obtain a normalized interval between [0, 1], and define all data in the normalized interval as an output probability coefficient; A filled normalized interval acquisition module is used to interpolate and fill all output probability coefficients in the normalized interval through a machine learning model to obtain a filled normalized interval; An output probability function acquisition module is used to perform linear fitting with the cloud thickness as an independent variable and all output probability coefficients of the filled normalized interval as dependent variables to obtain an output probability function; The output probability acquisition module is used to obtain the real-time cloud thickness or future cloud thickness of the preset area and substitute it into the output probability function to obtain the real-time output probability or future output probability.

7. An electronic device, characterized in that: It includes a processor and a memory coupled to the processor, wherein the memory stores program instructions that can be executed by the processor; when the processor executes the program instructions stored in the memory, the output probability prediction method as described in any one of claims 1 to 5 is implemented.

8. A storage medium, characterized in that: The storage medium stores program instructions, and when the program instructions are executed by the processor, the output probability prediction method according to any one of claims 1 to 5 can be implemented.

Images