Photovoltaic power ultra-short-term prediction system and method

Abstract

The present invention provides a photovoltaic power ultra-short-term prediction system and method, and relates to the technical field of photovoltaic power prediction. The present invention collects historical environmental time series data and historical photovoltaic power time series data of the location of a photovoltaic station to be predicted through a data acquisition module, classifies photovoltaic panels through a clustering algorithm according to the maximum average power generation of each photovoltaic panel, constructs a photovoltaic power ultra-short-term prediction model based on a fuzzy neural network through a model construction module, uses the collected historical environmental time series data of the photovoltaic area as the input of the prediction model, and uses the historical power time series data as the output of the prediction model, trains the photovoltaic power ultra-short-term prediction model, and globally optimizes the center value and width value of the prediction model through a model optimization module, obtains environmental data at future times through meteorological forecasts, and obtains photovoltaic power prediction values ​​at future times through the trained prediction model.

Description

Technical Field

[0001] The invention relates to the technical field of photovoltaic power prediction, and in particular to a photovoltaic power ultra-short-term prediction system and method. Background Art

[0002] With the rapid development of the economy, energy demand is increasing. Photovoltaic power generation has the advantages of being green, environmentally friendly and renewable. This new energy utilization method has shown steady growth worldwide. However, the power output of photovoltaic power generation systems is affected by many factors and is highly volatile. Especially when large-scale photovoltaic power generation is connected to the grid, the safe operation and power quality of the power system will be greatly affected. Therefore, photovoltaic power prediction is of great significance to grid planning and stable operation.

[0003] In the existing technology, neural networks are often used for ultra-short-term prediction of photovoltaic power, that is, meteorological data of the location of the photovoltaic station is obtained, the neural network model is trained, and then the current meteorological data is input according to the trained neural network model to obtain photovoltaic power data, or the photovoltaic power is predicted under different weather conditions. However, the neural network prediction model used in the existing technology is too simple and lacks the ability to optimize itself.

[0004] However, the existing technologies are all based on the prediction of the entire photovoltaic station, taking all the photovoltaic panels of the photovoltaic station as a whole. However, each photovoltaic panel has certain differences. Since there are many photovoltaic panels in the photovoltaic station, the differences between the photovoltaic panels will be further magnified. At the same time, with the increase in usage time, the aging of the photovoltaic panels will further aggravate the differences between the photovoltaic panels. Therefore, the existing technology will produce certain errors in the overall photovoltaic power prediction. Therefore, there is an urgent need for a photovoltaic power ultra-short-term prediction system that takes into account the differences between photovoltaic panels and can optimize the prediction system.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a photovoltaic power ultra-short-term prediction system and method to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A photovoltaic power ultra-short-term prediction system and method, the specific steps include:

[0009] A data acquisition module, wherein the data acquisition module is used to collect historical environmental time series data and historical photovoltaic power time series data of the location of the photovoltaic station to be predicted, and based on the photovoltaic power data of the same environmental data at the same time, the maximum power generation of each photovoltaic panel for multiple days is counted, and the average maximum power generation of the photovoltaic panel is calculated. The environmental data includes solar radiation, solar altitude angle, temperature value, humidity value and wind speed value. The photovoltaic panels are classified using a clustering algorithm according to the average maximum power generation, one of which is selected as a benchmark group, and the power generation rate ratio between each group is calculated;

[0010] A model building module, wherein the model building module is used to build a photovoltaic power ultra-short-term prediction model based on a fuzzy neural network, using the collected historical environmental time series data of the photovoltaic area as the input of the prediction model, and the historical power time series data as the output of the prediction model, to train the photovoltaic power ultra-short-term prediction model, obtain environmental data at a future time t through meteorological forecasts, and input the data into the input end of the trained photovoltaic power ultra-short-term prediction model, and obtain the ultra-short-term prediction value of photovoltaic power at a future time t;

[0011] A model optimization module is used to establish an optimization model, construct an optimization model, set optimization model parameters, perform global optimization on the center value and width value of the prediction model, and construct the error function of the output of the prediction model into a fitness function, determine the optimal values ​​of the center value and width value of the prediction model according to the fitness, and improve the prediction accuracy of the model;

[0012] A power prediction module calculates the ultra-short-term predicted power of photovoltaic panels of the benchmark group at a future moment according to the prediction model, calculates the ultra-short-term predicted power of photovoltaic power generation of other groups of photovoltaic panels by power generation rate ratio, and calculates the ultra-short-term predicted photovoltaic power of the entire photovoltaic station according to all photovoltaic groupings and the number of photovoltaic panels in each group.

[0013] In this embodiment, the calculation formula of the average maximum power generation is:

[0014]

[0015] Among them, P is the average maximum power generation, P n , max , is the maximum power on the nth day, and d is the number of days when the maximum power is collected.

[0016] In this embodiment, the specific logic of classifying photovoltaic panels by using the clustering algorithm is:

[0017] The maximum power generation of each photovoltaic panel is taken as a data node, and K data nodes are randomly selected as the initial centroids. K represents the number of clustered categories, and K is a positive integer. The i-th initial centroid is expressed as: C i[P(i)], P(i) represents the maximum power of the i-th data node as the initial centroid, i is a positive integer, and i = 1, 2, ..., K;

[0018] The power generation of each photovoltaic panel is used as the feature vector of the data node. For each feature vector of each photovoltaic panel, its distance to each initial centroid is calculated and assigned to the cluster represented by the nearest centroid. The formula for calculating the distance to the initial centroid is:

[0019]

[0020] Where d(i, j) represents the distance between the i-th initial centroid and the j-th data node, and P(j) represents the maximum power of the photovoltaic panel corresponding to the j-th data node;

[0021] For each cluster, after each clustering is completed, the mean of all points in the cluster is recalculated and used as the feature data of the new centroid. The update formula of the feature data of the i-th centroid is:

[0022]

[0023] Where m represents the number of data nodes assigned to the i-th centroid, P(i) old represents the maximum power of the photovoltaic panel corresponding to the data node assigned to the i-th centroid, P(i) new represents the maximum power of the updated centroid;

[0024] Based on the updated feature vector of the centroid, clustering is performed again until the change of all centroid positions is less than the threshold, then the clustering is considered stable and the clustering is terminated.

[0025] In this embodiment, one of the categories is selected as the reference group, and the formula for calculating the power generation rate ratio between each group is:

[0026]

[0027] Among them, α i is the power generation ratio of the i-th group, P sta Maximum power of the reference group, P i is the maximum power of the ith photovoltaic panel grouped by clustering.

[0028] In this embodiment, the photovoltaic power ultra-short-term prediction model based on fuzzy neural network consists of an input layer, a fuzzy layer, a calculation layer, and an output layer. The specific logic is:

[0029] The first layer is the input layer, which represents the input parameters. This layer has k nodes x1, x2, x3, …, x k represents k factors related to photovoltaic power generation, and the formula is:

[0030] x j =(x1,x2,x3,…,x k )

[0031] Among them, x j The input environmental parameters of the photovoltaic station;

[0032] The second layer is the fuzzification layer, which uses the Gaussian function as the membership function to calculate the input variable x j The membership degree is:

[0033]

[0034] in, The input variable x j The membership degree, m ji and σ ji Fuzzy Sets The center value and width value of , j = 1, 2, 3, ..., k, i = 1, 2, 3, ..., n, n is the number of fuzzy classifications of the input;

[0035] The third layer is the fuzzy rule calculation layer, which performs normalization operations on the fuzzified parameters. The formula is:

[0036]

[0037] Among them, ω i is the membership product of the input parameters;

[0038] The fourth layer is the output layer, which completes the center average defuzzification operation and outputs the result. The formula is:

[0039]

[0040] Among them, y i is the power output value, are the fuzzy system parameters.

[0041] In this embodiment, the optimization model is constructed based on the sparrow search optimization algorithm. The sparrow population consists of discoverers, joiners and a certain number of scouts. The specific optimization logic is:

[0042] The finder's position update formula is as follows:

[0043]

[0044] Where t is the current iteration number, j is the dimension of the parameter to be optimized, j = 1, 2, ..., d, d is the total number of parameters to be optimized, X i,j is the position information of the i-th sparrow in the j-th dimension; itermax is the maximum number of iterations; α is a random number, α∈(0,1), R2 and ST are the alarm value and safety threshold; L is a 1×d dimensional matrix; Q is a random number that conforms to the normal distribution;

[0045] The position update formula of the joiner is as follows:

[0046]

[0047] Where: X P is the optimal position of the current discoverer; X worst is the current global worst position; A is a 1×d matrix, whose elements are randomly assigned 1 or -1, and ni is the size of the sparrow population;

[0048] The scout position update formula is as follows:

[0049]

[0050] Among them, X best is the global optimal position, β is the iteration step length, f i is the fitness value of the current sparrow individual, f g and f w is the current global optimal and worst fitness value, K is a random number, K∈[-1,1], ε is a constant to prevent the denominator from being 0;

[0051] The root mean square error is used to calculate the fitness, and the calculation formula is as follows:

[0052]

[0053] Among them, n is the number of samples, x i is the actual value of photovoltaic power, y i is the predicted value of photovoltaic power.

[0054] In this embodiment, the ultra-short-term predicted power of the reference group of photovoltaic panels obtained by the prediction model is calculated by the formula of the ultra-short-term predicted power of photovoltaic power generation of other groups of photovoltaic panels through the power generation rate ratio as follows:

[0055] P i,pre,t =α i *P sta,pre,t

[0056] Among them, α i is the power generation ratio of the ith group of photovoltaic panels, P i,pre,t The predicted power of the ith photovoltaic panel at the next time t, P sta,pre,t is the predicted power of the benchmark group PV panels at time t in the future.

[0057] In this embodiment, the calculation formula for the ultra-short-term predicted photovoltaic power of the entire photovoltaic station is as follows:

[0058]

[0059] Among them, P z,pre is the ultra-short-term predicted power of the photovoltaic station at time t, P i,pre,t The ultra-short-term predicted power of the ith group of photovoltaic panels at time t, l i The number of photovoltaic panels in the i-th group, i is the i-th group of photovoltaic panels, k is the number of photovoltaic panels, and i and k are positive integers.

[0060] The present invention further provides a photovoltaic power ultra-short-term prediction method, which is executed by the above-mentioned photovoltaic power ultra-short-term prediction system, and the photovoltaic power ultra-short-term prediction method comprises the following steps:

[0061] Step 1: Obtain the historical environmental time series data and historical photovoltaic power time series data of the location of the photovoltaic station, and classify each photovoltaic panel by clustering method according to the maximum photovoltaic power of the photovoltaic panel, take one of the categories as the benchmark group, and calculate the power generation rate ratio between each group;

[0062] Step 2: Construct a photovoltaic power ultra-short-term prediction model based on fuzzy neural network, use the collected historical environmental time series data of the photovoltaic area as the input of the prediction model, and the historical power time series data as the output of the prediction model, and train the photovoltaic power ultra-short-term prediction model;

[0063] Step 3: Establish an optimization model, set the optimization model parameters, perform global optimization on the center value and width value of the prediction model, and construct the error function of the output of the prediction model into a fitness function, and determine the optimal values ​​of the center value and width value of the prediction model according to the fitness;

[0064] Step 4: Obtain environmental data at a future time t, input it into the input end of the trained photovoltaic power ultra-short-term prediction model, and obtain the photovoltaic power ultra-short-term prediction value at the future time t;

[0065] Step 5: Based on the ultra-short-term predicted power of the photovoltaic panels of the benchmark group at future times obtained by the prediction model, the ultra-short-term predicted power of photovoltaic power generation of photovoltaic panels in other groups is calculated through the power generation rate ratio. Based on all photovoltaic groupings and the number of photovoltaic panels in each group, the ultra-short-term predicted photovoltaic power of the entire photovoltaic station is calculated.

[0066] Compared with the prior art, the beneficial effects of the present invention are:

[0067] The present invention collects historical environmental time series data and historical photovoltaic power time series data of the location of the photovoltaic station to be predicted through a data acquisition module, classifies photovoltaic panels through a clustering algorithm according to the maximum average power generation of each photovoltaic panel, and constructs a photovoltaic power ultra-short-term prediction model based on a fuzzy neural network through a model construction module. The collected historical environmental time series data of the photovoltaic area is used as the input of the prediction model, and the historical power time series data is used as the output of the prediction model. The photovoltaic power ultra-short-term prediction model is trained, and the center value and width value of the prediction model are globally optimized through a model optimization module. The environmental data at future times are obtained through meteorological forecasts and input into the trained prediction model to obtain the photovoltaic power prediction value at future times. The power of the entire photovoltaic station is predicted through a power prediction module:

[0068] The present invention groups photovoltaic panels according to the maximum power of each photovoltaic panel and the aging characteristics of the photovoltaic panels, calculates the difference in power generation efficiency between each group of photovoltaic panels, and then predicts the total power generation power of the group according to the number of photovoltaic panels in each group and the power generation power of the photovoltaic panels in each group. The present invention also optimizes the photovoltaic power ultra-short-term prediction model through an optimization model to further improve the accuracy of the photovoltaic power ultra-short-term prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a schematic diagram of the overall system structure of the present invention;

[0070] Figure 2 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0071] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0072] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0073] Example:

[0074] See also Figure 1 , the present invention provides a technical solution:

[0075] A photovoltaic power ultra-short-term prediction system includes a data acquisition module, a model building module, a model optimization module and a power prediction module, wherein:

[0076] The data acquisition module is used to collect historical environmental time series data and historical photovoltaic power time series data of the location of the photovoltaic station to be predicted. According to the photovoltaic power data of the same environmental data at the same time, the maximum power generation power of each photovoltaic panel for multiple days is counted, and the average maximum power generation power of the photovoltaic panels is calculated. The environmental data includes solar radiation, solar altitude angle, temperature value, humidity value and wind speed value. The photovoltaic panels are classified using a clustering algorithm according to the average maximum power generation power, one of which is selected as the benchmark group, and the power generation rate ratio between each group is calculated.

[0077] Through historical meteorological data, the historical environmental time series data of solar radiation, solar altitude angle, temperature, humidity and wind speed at the location of the photovoltaic station are obtained respectively. The intensity of solar radiation directly determines the energy absorbed per unit area on the photovoltaic panel. After receiving photons, the photovoltaic panel will convert them into electrons and generate current. Therefore, the power generation efficiency of the photovoltaic panel is positively correlated with the intensity of solar radiation. The solar altitude angle directly affects the amount of sunlight scattered on the solar panel. The photoelectric conversion efficiency of the photovoltaic panel will decrease at high temperature. This is because the electrons and holes in the photovoltaic material are activated by heat, making it difficult for the electrons to be released from the semiconductor material, thereby reducing the generation of current. A high humidity environment may cause dust or other pollutants to accumulate on the surface of the photovoltaic panel. These pollutants or dust will reduce the transmittance of light, thereby reducing the power generation efficiency of the photovoltaic panel. A moderate wind speed helps to dissipate heat and clean the photovoltaic panel, thereby improving its power generation efficiency and long-term stability to a certain extent.

[0078] As time goes by, photovoltaic panels will be affected by many factors during use, including changes in environmental conditions, temperature, humidity, light intensity, and aging and degradation of the internal materials of photovoltaic panels. The efficiency of photovoltaic panels will decrease, which means that under the same lighting conditions, the proportion of photovoltaic panels that can be converted into electrical energy may decrease. The decrease in efficiency will cause the current and voltage generated under the same lighting to no longer be the optimal combination, thereby affecting the location of the maximum power generation point and the maximum power generation value. Therefore, the maximum power generation of photovoltaic panels can be one of the important indicators for evaluating their aging characteristics and long-term performance degradation. It can reflect the gradual changes and losses of photovoltaic panels in actual operation. The maximum power generation of photovoltaic panels is classified by clustering method, and the maximum power generation is similarly divided into a group. By predicting the photovoltaic power of photovoltaic panels in different groups, the accuracy of power prediction is greatly improved.

[0079] In this embodiment, the calculation formula of the average maximum power generation is:

[0080]

[0081] Among them, P is the average maximum power generation, P n , max , is the maximum power on the nth day, and d is the number of days when the maximum power is collected.

[0082] In this embodiment, the specific logic of classifying photovoltaic panels by using the clustering algorithm is:

[0083] The maximum power generation of each photovoltaic panel is taken as a data node, and K data nodes are randomly selected as the initial centroids. K represents the number of clustered categories, and K is a positive integer. The i-th initial centroid is expressed as: C i [P(i)], P(i) represents the maximum power of the i-th data node as the initial centroid, i is a positive integer, and i = 1, 2, ..., K;

[0084] The power generation of each photovoltaic panel is used as the feature vector of the data node. For each feature vector of each photovoltaic panel, its distance to each initial centroid is calculated and assigned to the cluster represented by the nearest centroid. The formula for calculating the distance to the initial centroid is:

[0085]

[0086] Where d(i, j) represents the distance between the i-th initial centroid and the j-th data node, and P(j) represents the maximum power of the photovoltaic panel corresponding to the j-th data node;

[0087] For each cluster, after each clustering is completed, the mean of all points in the cluster is recalculated and used as the feature data of the new centroid. The update formula of the feature data of the i-th centroid is:

[0088]

[0089] Where m represents the number of data nodes assigned to the i-th centroid, P(i) old represents the maximum power of the photovoltaic panel corresponding to the data node assigned to the i-th centroid, P(i) new represents the maximum power of the updated centroid;

[0090] Based on the updated feature vector of the centroid, clustering is performed again until the change of all centroid positions is less than the threshold, then the clustering is considered stable and the clustering is terminated.

[0091] In this embodiment, one of the categories is selected as the reference group, and the formula for calculating the power generation rate ratio between each group is:

[0092]

[0093] Among them, α i is the power generation ratio of the i-th group, P sta Maximum power of the reference group, P i is the maximum power of the ith photovoltaic panel grouped by clustering.

[0094] The model building module is used to build a photovoltaic power ultra-short-term prediction model based on a fuzzy neural network. The collected historical environmental time series data of the photovoltaic area is used as the input of the prediction model, and the historical power time series data is used as the output of the prediction model. The photovoltaic power ultra-short-term prediction model is trained, and the environmental data at a future time t is obtained through meteorological forecasts. The data is input into the input end of the trained photovoltaic power ultra-short-term prediction model to obtain the ultra-short-term prediction value of the photovoltaic power at the future time t.

[0095] Fuzzy neural network is a hybrid computing model that combines fuzzy logic and neural network. It combines the fuzzy system's ability to handle uncertainty and ambiguity with the advantages of neural network in complex pattern recognition and data modeling, so it can handle some problems that are difficult to handle with traditional methods. Photovoltaic power generation is affected by many factors such as weather conditions, light intensity, and solar altitude angle. These factors are uncertain and ambiguous. Fuzzy neural network can effectively process these complex and fuzzy input data and output relatively accurate prediction results, which makes it have an advantage in complex system prediction.

[0096] In this embodiment, the photovoltaic power ultra-short-term prediction model based on fuzzy neural network consists of an input layer, a fuzzy layer, a calculation layer, and an output layer. The specific logic is:

[0097] The first layer is the input layer, which represents the input parameters. This layer has k nodes x1, x2, x3, …, x k represents k factors related to photovoltaic power generation, and the formula is:

[0098] x j =(x1,x2,x3,…,x k )

[0099] Among them, x j The input environmental parameters of the photovoltaic station;

[0100] The second layer is the fuzzification layer, which uses the Gaussian function as the membership function to calculate the input variable x j The membership degree is:

[0101]

[0102] in, The input variable x j The membership degree, m ji and σ ji Fuzzy Sets The center value and width value of j =

[0103] 1,2,3,…,k, i = 1,2,3,…,n, n is the number of fuzzy classifications of the input;

[0104] The third layer is the fuzzy rule calculation layer, which performs normalization operations on the fuzzified parameters. The formula is:

[0105]

[0106] Among them, ω i is the membership product of the input parameters;

[0107] The fourth layer is the output layer, which completes the center average defuzzification operation and outputs the result. The formula is:

[0108]

[0109] Among them, y i is the photovoltaic power output value, are the fuzzy system parameters.

[0110] The model optimization module is used to establish an optimization model. By constructing the optimization model and setting the optimization model parameters, the center value and width value of the prediction model are globally optimized, and the error function of the output of the prediction model is constructed into a fitness function. The optimal values ​​of the center value and width value of the prediction model are determined according to the fitness, thereby improving the model prediction accuracy.

[0111] The center value and width of the fuzzy neural network are usually generated by random initialization, generating a random array with an initial value limited to a certain range, and correcting the center value and width through the error back propagation algorithm. However, this method has a large randomness, which can easily lead to local optimal problems in the training parameters of the fuzzy neural network. The sparrow search optimization algorithm model is used to perform global optimization on the center value and width of the fuzzy neural network, which can avoid the network from falling into the local optimal problem and improve the prediction accuracy of the model.

[0112] In this embodiment, the optimization model is constructed based on the sparrow search optimization algorithm. The sparrow population consists of discoverers, joiners and a certain number of scouts. The specific optimization logic is:

[0113] The finder's position update formula is as follows:

[0114]

[0115] Where t is the current iteration number, j is the dimension of the parameter to be optimized, j = 1, 2, ..., d, d is the total number of parameters to be optimized, X i,j is the position information of the i-th sparrow in the j-th dimension, iter max is the maximum number of iterations, α is a random number, α∈(0,1), R2 and ST are the alarm value and safety threshold, L is a 1×d dimensional matrix, and Q is a random number that conforms to the normal distribution;

[0116] The position update formula of the joiner is as follows:

[0117]

[0118] Where: X P is the optimal position of the current discoverer; X worst is the current global worst position; A is a 1×d matrix, whose elements are randomly assigned 1 or -1, and ni is the size of the sparrow population;

[0119] The scout position update formula is as follows:

[0120]

[0121] Among them, X best is the global optimal position, β is the iteration step length, f i is the fitness value of the current sparrow individual, f g and f w is the current global optimal and worst fitness value, K is a random number, K∈[-1,1], ε is a constant to prevent the denominator from being 0;

[0122] The root mean square error is used to calculate the fitness of the optimization model, and the calculation formula is as follows:

[0123]

[0124] Among them, n is the number of samples, x i is the actual value of photovoltaic power, y i is the predicted value of photovoltaic power.

[0125] The fitness value reflects the quality or performance of the current individual solution. Generally, the higher the fitness value, the better the individual solution and the closer it is to the optimal solution of the problem. The lower the fitness value, the worse the individual solution is and the more optimization is needed. During the calculation process, the individual center value and width value are continuously updated through an iterative optimization process to gradually improve the fitness value and eventually reach the optimal solution or a state close to the optimal solution of the problem.

[0126] A group of sparrow individuals are randomly initialized, each of which corresponds to a data set of center and width values ​​in the fuzzy neural network. According to the current combination of center and width values ​​and the fitness of the fuzzy neural network, the prediction accuracy of the fuzzy neural network is evaluated, and it is determined whether the current center and width values ​​are optimal. Then, according to the strategy of the sparrow search optimization algorithm, the position of each sparrow individual is updated, and the evaluation and update steps are repeated until the set maximum number of iterations is reached or the fitness threshold is reached. During the search process, the global optimal combination of center and width values ​​is recorded and retained as the optimal parameter setting of the fuzzy neural network.

[0127] The power prediction module calculates the ultra-short-term predicted power of the photovoltaic panels of the benchmark group at future times according to the prediction model, calculates the ultra-short-term predicted power of photovoltaic power generation of other groups of photovoltaic panels through the power generation rate ratio, and calculates the ultra-short-term predicted photovoltaic power of the entire photovoltaic station according to all photovoltaic groupings and the number of photovoltaic panels in each group.

[0128] In this embodiment, the ultra-short-term predicted power of the reference group of photovoltaic panels obtained by the prediction model is calculated by the formula of the ultra-short-term predicted power of photovoltaic power generation of other groups of photovoltaic panels through the power generation rate ratio as follows:

[0129] P i,pre,t =α i *P sta,pre,t

[0130] Among them, α i is the power generation ratio of the ith group of photovoltaic panels, P i,pre,t The predicted power of the ith photovoltaic panel at the next time t, P sta,pre,t is the predicted power of the benchmark group PV panels at time t in the future.

[0131] In this embodiment, the calculation formula for the ultra-short-term predicted photovoltaic power of the entire photovoltaic station is as follows:

[0132]

[0133] Among them, P z,pre is the ultra-short-term predicted power of the photovoltaic station at time t, P i,pre,t The ultra-short-term predicted power of the ith group of photovoltaic panels at time t, l i The number of photovoltaic panels in the i-th group, i is the i-th group of photovoltaic panels, k is the number of photovoltaic panels, and i and k are positive integers.

[0134] See also Figure 2 The present invention also provides a photovoltaic power ultra-short-term prediction method, which is executed by the above-mentioned photovoltaic power ultra-short-term prediction system, and the photovoltaic power ultra-short-term prediction method comprises the following steps:

[0135] Step 1: Obtain the historical environmental time series data and historical photovoltaic power time series data of the location of the photovoltaic station, and classify each photovoltaic panel by clustering method according to the maximum photovoltaic power of the photovoltaic panel, take one of the categories as the benchmark group, and calculate the power generation rate ratio between each group;

[0136] Step 2: Construct a photovoltaic power ultra-short-term prediction model based on fuzzy neural network, use the collected historical environmental time series data of the photovoltaic area as the input of the prediction model, and the historical power time series data as the output of the prediction model, and train the photovoltaic power ultra-short-term prediction model;

[0137] Step 3: Establish an optimization model, set the optimization model parameters, perform global optimization on the center value and width value of the prediction model, and construct the error function of the output of the prediction model into a fitness function, and determine the optimal values ​​of the center value and width value of the prediction model according to the fitness;

[0138] Step 4: Obtain environmental data at a future time t, input it into the input end of the trained photovoltaic power ultra-short-term prediction model, and obtain the photovoltaic power ultra-short-term prediction value at the future time t;

[0139] Step 5: Based on the ultra-short-term predicted power of the photovoltaic panels of the benchmark group at future times obtained by the prediction model, the ultra-short-term predicted power of photovoltaic power generation of photovoltaic panels in other groups is calculated through the power generation rate ratio. Based on all photovoltaic groupings and the number of photovoltaic panels in each group, the ultra-short-term predicted photovoltaic power of the entire photovoltaic station is calculated.

[0140] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0141] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0142] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0143] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A photovoltaic power ultra-short-term prediction system, characterized in that: The specific steps include: A data acquisition module, wherein the data acquisition module is used to collect historical environmental time series data and historical photovoltaic power time series data of the location of the photovoltaic station to be predicted, and based on the photovoltaic power data of the same environmental data at the same time, the maximum power generation of each photovoltaic panel for multiple days is counted, and the average maximum power generation of the photovoltaic panel is calculated. The environmental data includes solar radiation, solar altitude angle, temperature value, humidity value and wind speed value. The photovoltaic panels are classified using a clustering algorithm according to the average maximum power generation, one of which is selected as a benchmark group, and the power generation rate ratio between each group is calculated; The model building module is used to build a photovoltaic power ultra-short-term prediction model based on a fuzzy neural network, using the collected historical environmental time series data of the photovoltaic area as the input of the prediction model and the historical power time series data as the output of the prediction model to train the photovoltaic power ultra-short-term prediction model and obtain the future time through weather forecast. The environmental data at that time is input into the input end of the trained photovoltaic power ultra-short-term prediction model to obtain the future Ultra-short-term prediction value of photovoltaic power at time ; A model optimization module is used to establish an optimization model, construct an optimization model, set optimization model parameters, perform global optimization on the center value and width value of the prediction model, and construct the error function of the output of the prediction model into a fitness function, determine the optimal values ​​of the center value and width value of the prediction model according to the fitness, and improve the prediction accuracy of the model; A power prediction module, which calculates the ultra-short-term predicted power of photovoltaic panels of the benchmark group at a future moment according to the prediction model, calculates the ultra-short-term predicted power of photovoltaic power generation of photovoltaic panels of other groups by power generation rate ratio, and calculates the ultra-short-term predicted power of photovoltaic power generation of the entire photovoltaic station according to all photovoltaic groupings and the number of photovoltaic panels in each group; The photovoltaic power ultra-short-term prediction model based on fuzzy neural network consists of input layer, fuzzy layer, calculation layer and output layer. The specific logic is: The first layer is the input layer, which represents the input parameters. Nodes represent Factors related to photovoltaic power generation, the formula is: in, The input environmental parameters of the photovoltaic station; The second layer is the fuzzification layer, which uses the Gaussian function as the membership function to calculate the input variable The membership degree is: in, For input variables The membership degree of and Fuzzy Sets The center and width values ​​of , , is the number of fuzzy classifications of the input; The third layer is the fuzzy rule calculation layer, which performs normalization operations on the fuzzified parameters. The formula is: in, is the membership product of the input parameters; The fourth layer is the output layer, which completes the center average defuzzification operation and outputs the result. The formula is: in, is the power output value, is the fuzzy system parameter; One of the selected groups is the reference group, and the formula for calculating the power generation ratio between each group is: in, For the The power generation efficiency ratio of the photovoltaic panels is The maximum power of the benchmark group, For clustering group The maximum power of the photovoltaic panels; The formula for calculating the ultra-short-term predicted power of photovoltaic panels of the benchmark group obtained by the prediction model and the ultra-short-term predicted power of photovoltaic power generation of other groups of photovoltaic panels by the power generation rate ratio is as follows: in, For the The power generation efficiency ratio of the photovoltaic panels is No. Photovoltaic panels for the future The predicted power at Benchmarking photovoltaic panels for the future The predicted power at .

2. A photovoltaic power ultra-short-term prediction system according to claim 1, characterized in that: The calculation formula for the average maximum power generation is: in, is the average maximum power generation, For the Maximum power per day, is the number of days when the maximum power is collected.

3. A photovoltaic power ultra-short-term prediction system according to claim 1, characterized in that: The specific logic of classifying photovoltaic panels through clustering algorithm is: The maximum power generation of each photovoltaic panel is used as a data node and randomly selected data nodes as the initial centroid, represents the number of categories after clustering, is a positive integer, where The initial centroid is expressed as: , Indicates The maximum power of the data nodes as the initial centroid, is a positive integer, and ; The power generation of each photovoltaic panel is used as the feature vector of the data node. For each feature vector of each photovoltaic panel, its distance to each initial centroid is calculated and assigned to the cluster represented by the nearest centroid. The formula for calculating the distance to the initial centroid is: in, Indicates The initial centroid and The distance between data nodes, Indicates The maximum power of the photovoltaic panel corresponding to each data node; For each cluster, after each clustering is completed, the mean of all points in the cluster is recalculated and used as the feature data of the new centroid. The update formula of the centroid feature data is: in, Indicates that the The number of data nodes of the centroid, Indicates that the The maximum power of the photovoltaic panel corresponding to the data node of the centroid, represents the maximum power of the updated centroid; Based on the updated feature vector of the centroid, clustering is performed again until the change of all centroid positions is less than the threshold, then the clustering is considered stable and the clustering is terminated.

4. A photovoltaic power ultra-short-term prediction system according to claim 1, characterized in that: The optimization model is based on the sparrow search optimization algorithm. The sparrow population consists of discoverers, joiners and a certain number of scouts. The specific optimization logic is: The finder's position update formula is as follows: in, is the current iteration number, is the dimension of the parameter to be optimized, , is the total number of parameters to be optimized, For the A sparrow in Location information in the dimension; is the maximum number of iterations; is a random number, , and are alarm values ​​and safety thresholds; for Dimensional matrix; is a random number that conforms to the normal distribution; The position update formula of the joiner is as follows: Where: is the optimal position of the current discoverer; is the current worst position in the world; for A matrix whose elements are randomly assigned 1 or -1. is the sparrow population size; The scout position update formula is as follows: in, is the global optimal position, is the iteration step length, is the fitness value of the current sparrow individual, and are the current global optimal and worst fitness values, is a random number, , is a constant to prevent the denominator from being 0; The root mean square error is used to calculate the fitness, and the calculation formula is as follows: in, is the number of samples, is the actual value of photovoltaic power, is the predicted value of photovoltaic power.

5. The photovoltaic power ultra-short-term prediction system according to claim 1, characterized in that: The calculation formula for the ultra-short-term predicted photovoltaic power of the entire photovoltaic station is as follows: in, Ultra-short-term The predicted power is No. Ultra-short-term photovoltaic panels When the power is predicted, No. The number of photovoltaic panels in a group, For the Photovoltaic panels, is the number of photovoltaic panels, and Is a positive integer.

6. A photovoltaic power ultra-short-term prediction method, characterized by: The photovoltaic power ultra-short-term prediction method is executed by the photovoltaic power ultra-short-term prediction system according to any one of claims 1 to 5, and the photovoltaic power ultra-short-term prediction method comprises the following steps: Step 1: Obtain the historical environmental time series data and historical photovoltaic power time series data of the location of the photovoltaic station, and classify each photovoltaic panel by clustering method according to the maximum photovoltaic power of the photovoltaic panel, take one of the categories as the benchmark group, and calculate the power generation rate ratio between each group; Step 2: Construct a photovoltaic power ultra-short-term prediction model based on fuzzy neural network, use the collected historical environmental time series data of the photovoltaic area as the input of the prediction model, and the historical power time series data as the output of the prediction model, and train the photovoltaic power ultra-short-term prediction model; Step 3: Establish an optimization model, set the optimization model parameters, perform global optimization on the center value and width value of the prediction model, and construct the error function of the output of the prediction model into a fitness function, and determine the optimal values ​​of the center value and width value of the prediction model according to the fitness; Step 4: Get the future moment The environmental data at that time is input into the input end of the trained photovoltaic power ultra-short-term prediction model to obtain the future Ultra-short-term prediction value of photovoltaic power at time ; Step 5: Based on the ultra-short-term predicted power of the photovoltaic panels of the benchmark group at future times obtained by the prediction model, the ultra-short-term predicted power of photovoltaic power generation of photovoltaic panels in other groups is calculated through the power generation rate ratio. Based on all photovoltaic groupings and the number of photovoltaic panels in each group, the ultra-short-term predicted photovoltaic power of the entire photovoltaic station is calculated.

Images