Method for predicting output power of optical storage power generation system

Abstract

The invention relates to an optical storage power generation system output power prediction method, which comprises the following steps of: selecting factors with relatively strong relevance with output power, and constructing a photovoltaic array output model; selecting similar days; constructing a BP neural network prediction model through parameters of an input layer, a hidden layer and an output layer; the BP neural network prediction model is trained, the prediction speed and prediction precision of the BP neural network prediction model are improved, and the BP neural network prediction model has good precision and certain practicability and feasibility.

Description

Technical Field

[0001] The present invention belongs to the technical field of power control of photovoltaic power generation systems, and in particular relates to a method for predicting output power of photovoltaic power generation systems. Background Art

[0002] In recent years, photovoltaic power generation has developed rapidly due to its clean energy characteristics. However, due to the influence of solar radiation intensity and weather factors, the output power of photovoltaic power generation systems is volatile and intermittent. The grid-connected operation of large-scale photovoltaic power generation systems will affect the safe, stable and economic operation of the power system, and have an adverse impact on the safe and stable operation of the power grid. Therefore, in order to overcome the problem that the output power of the photovoltaic system fluctuates due to changes in external environmental factors, it is necessary to find an accurate method for predicting the output power of the photovoltaic power generation system. This is the key to improving the operating efficiency of photovoltaic power stations and ensuring the safe and stable operation of photovoltaic power generation after grid connection. It is conducive to the dispatching management personnel to adjust the dispatching plan in time, reduce the negative impact of the new energy power generation system after grid connection, and thus improve the efficiency and stability of the power grid. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide a method for predicting the output power of a photovoltaic power generation system. The accurate prediction of the output power of the photovoltaic power generation system is beneficial to the power system dispatching department to adjust the dispatching plan in a timely manner, thereby effectively reducing the adverse effects of the access of the photovoltaic power generation system to the power grid, and has high practicality and feasibility.

[0004] The technical solution adopted by the present invention is: a method for predicting the output power of a photovoltaic power generation system, comprising the following steps:

[0005] S1: Select factors with strong correlation with output power and construct a photovoltaic array output model;

[0006] S2: Select similar days;

[0007] S3: Construct the BP neural network prediction model through the input layer, hidden layer and output layer parameters;

[0008] S4: Train the BP neural network prediction model to improve the prediction speed and prediction accuracy of the BP neural network prediction model.

[0009] Specifically, in step S1, the photovoltaic power output power per unit area can be calculated according to the following formula:

[0010] P s =ηSI[1-0.005(t 0 +25)]

[0011] Where: η is the photoelectric conversion efficiency of the photovoltaic array; S is the total area of ​​the photovoltaic array; I is the solar radiation intensity received by the photovoltaic array; t 0 is the operating temperature of the solar panel.

[0012] Specifically, in step S2, in order to select a date with the same weather type and season type as the predicted day and the closest temperature from the historical records of the photovoltaic power generation system, the output power sequence and temperature information are used as part of the prediction model input, and a similar day selection algorithm is proposed. The algorithm steps are as follows:

[0013] 1) Select n historical records that are consistent with the weather type and season type of the predicted day to form a sample set D;

[0014] 2) Calculate the Euclidean distance d between the predicted day and the historical records of the temperature in the sample set D i , d i The calculation formula is as follows:

[0015]

[0016] Where: Y 1 , Y 2 , Y 3 are the maximum temperature, minimum temperature and average temperature of the forecast day respectively; X i1 , X i2 , X i3 are the maximum temperature, minimum temperature and average temperature values ​​of the i-th record in the sample set D respectively;

[0017] 3) Set the temperature Euclidean distance set {d 1 ,d 2 ,…d n}Sort in ascending order according to the value size, and the date corresponding to the minimum value is the similar day corresponding to the predicted day.

[0018] Specifically, in step S3, the BP neural network prediction model includes an input layer, a hidden layer and an output layer;

[0019] Selection of input layer variables: The input variables are similar daily output power series, temperature information, and predicted daily temperature information. There are 20 input variables in total.

[0020] Selection of output layer variables: Since the output power model of the photovoltaic power generation system predicts the average output power per hour from 6:00 to 19:00 on the prediction day, there are 14 output variables in total;

[0021] Selection of hidden layer variables: According to Kolmogorov theorem, if the number of input variables is n, then

[0022] The number of neurons in the hidden layer can generally be 2n+1, or the number of neurons in the hidden layer can be determined by repeated trials.

[0023] Specifically, in step S4, the generalization ability of the BP neural network is the main performance indicator. When the BP neural network is trained using samples, a part of the input variable data is used as a training set and the other part is used as a test set.

[0024] Beneficial effects of the invention: The invention proposes a method for predicting the output power of a photovoltaic power generation system. By analyzing the factors that affect the output of the photovoltaic power generation system, the historical output power sequence is selected, and the relevant meteorological information is used as the input of the prediction model. In order to improve the prediction accuracy of the model under different weather conditions, a similar day selection algorithm and a method for determining training samples based on the meteorological information of the prediction day are proposed, which can effectively improve the prediction accuracy of the model under various weather conditions; the model directly predicts the output power, avoids complex modeling, and simplifies the prediction process. The BP neural network prediction model of the invention has good accuracy and has certain practicality and feasibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a flow chart of the steps of the present invention;

[0026] Figure 2 It is a structural diagram of the BP neural network prediction model in the present invention;

[0027] Figure 3 This is a graph showing the prediction results of array output power for 7 consecutive days in the present invention;

[0028] Figure 4 This is a prediction error curve diagram of Model I in the present invention;

[0029] Figure 5 This is the prediction error curve of model II in the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention, and are specifically described in combination with the embodiments below.

[0031] like Figure 1-5 As shown, the present invention comprises the following steps:

[0032] S1: Select factors with strong correlation with output power and construct a photovoltaic array output model. The specific process is as follows:

[0033] In the step S1, the output power of the photovoltaic power source per unit area can be calculated according to the following formula:

[0034] P s = ηSI[1 - 0.005(t 0 + 25)]

[0035] In the formula: η is the photoelectric conversion efficiency of the photovoltaic array; S is the total area of the photovoltaic array; I is the solar radiation intensity received by the photovoltaic array; t 0 is the operating temperature of the solar panel.

[0036] S2: Select similar days. The specific process is as follows: In the step S2, in order to select the date with the same weather type and season type as the prediction day and the closest temperature from the historical records of the photovoltaic power generation system, and use its output power sequence and temperature information as part of the input of the prediction model, a similar day selection algorithm is proposed. The algorithm steps are as follows:

[0037] 1) Select n historical records with the same weather type and season type as the prediction day to form a sample set D;

[0038] 2) Calculate the Euclidean distance d i of the temperature between the prediction day and the historical records in the sample set D, and d i The calculation formula is as follows:

[0039]

[0040] In the formula: Y 1 , Y 2 , Y 3 are the maximum temperature, minimum temperature and average temperature values of the prediction day respectively; X i1 , X i2 , X i3 are the maximum temperature, minimum temperature and average temperature values of the i-th record in the sample set D respectively;

[0041] 3) Arrange the set of Euclidean distances of temperatures {d 1 , d 2 , … d n} in ascending order according to the values, and the date corresponding to the minimum value is the similar day corresponding to the prediction day.

[0042] The determination method of the training samples depends on the similar day selection algorithm. The historical records of the first k dates arranged in ascending order of the Euclidean distance of temperatures form the training samples. The neural network can be trained using the training samples, and then the trained neural network can be used for output power prediction (1 < k ≤ n, and the value of k can be specified according to actual needs.

[0043] S3: Construct a BP neural network prediction model through the input layer, hidden layer and output layer parameters, which is specifically expressed as follows:

[0044] In the step S3, Figure 2 is a structural diagram of a BP neural network prediction model, wherein the BP neural network prediction model includes an input layer, a hidden layer and an output layer;

[0045] Selection of input layer variables: The input variables are similar daily output power series, temperature information, and predicted daily temperature information. There are 20 input variables in total, as shown in Table 1:

[0046] Input variables Enter the variable name <![CDATA[x 1 -x 14 ]]> Average hourly output power between 6:00 and 19:00 on similar days <![CDATA[x 15 -x 17 ]]> Highest, lowest and average temperatures on similar days <![CDATA[x 18 -x 20 ]]> The maximum temperature, minimum temperature and average temperature of the forecast day

[0047] Table 1: Prediction model input variables

[0048] Selection of output layer variables: Since the output power model of the photovoltaic power generation system predicts the average output power per hour from 6:00 to 19:00 on the prediction day, there are 14 output variables in total;

[0049] Selection of hidden layer variables: According to Kolmogorov theorem, if the number of input variables is n, then

[0050] The number of neurons in the hidden layer can generally be 2n+1, or the number of neurons in the hidden layer can be determined by repeated trials.

[0051] S4: Train the BP neural network prediction model to improve the prediction speed and prediction accuracy of the BP neural network prediction model, as shown below:

[0052] In step S4, the generalization ability of the BP neural network is the main performance indicator. When the BP neural network is trained using samples, a part of the input variable data is used as a training set and the other part is used as a test set.

[0053] The training and prediction steps of the BP neural network prediction model are:

[0054] 1) Preprocess the input meteorological data. Remove individual singular observation data, and normalize the input data to avoid neuron saturation. When normalizing, obtain the range of each variable based on the historical meteorological data of the forecast area.

[0055]

[0056] In the formula, x ik is the original data of the kth eigenvector on the i-th day; x imax 、x imin are the maximum and minimum values ​​of the original data respectively; x' ik The normalized data

[0057] 2) When making predictions, the days with correlation function value r≥0.80 are selected as similar days, and the meteorological data of the first five similar days are used as input variables of the prediction model to predict the array output power on the prediction day, which is called prediction model II. In order to facilitate intuitive comparison and analysis, the meteorological data of the five days before the prediction day are used as input, and prediction is made using the BP neural network model, which is called prediction model I.

[0058] Specific case verification: Taking the measured output power value of a photovoltaic array in a certain place in my country and the local meteorological data as an example, the power data value is sampled once every 1 hour, and the meteorological data at the corresponding time point is used as the input variable of the model. All meteorological and power data are divided into 2 groups, and the observation data one month before the prediction day is used as the training set. The network is trained multiple times; finally, the array output power for one week is predicted. According to the principle of selecting similar days, the meteorological data of 5 days before the prediction day are selected as input variables. Taking the first prediction day as an example, the correlation coefficients between each similar day and the prediction day are shown in Table 1. Similarly, 5 similar days can be selected as the prediction model input for given different prediction days.

[0059] Date serial number 1 2 3 4 5 Correlation coefficient / r 0.84 0.86 0.81 0.83 0.85

[0060] Table 2: Correlation coefficients between similar days and predicted days

[0061] Prediction results and error analysis:

[0062] The meteorological data of the five days before the forecast date is used as input, and the forecast model I is established according to the BP neural network structure; the forecast model II is constructed according to the principle of similar day selection, and the array output power forecast results for the selected seven consecutive days are as follows: Figure 3 shown.

[0063] When conducting error analysis, selecting reasonable and scientific error indicators is of great significance for evaluating the prediction effect. The error curves of the two models are as follows: Figure 4 , Figure 5 As shown. It can be seen that compared with model I, the prediction error of model II is improved, and the stability of the prediction results is also significantly improved. MAPE And the root mean square error e RMSE Analyze and compare the prediction results of Model I and Model II:

[0064]

[0065]

[0066] Where: P i is the measured value of the photovoltaic array output power; is the predicted value of output power; N is the number of predicted samples.

[0067] Prediction error of prediction model I and II MAPE 、e RMSE See Table 3.

[0068] Prediction Model <![CDATA[e MAPE ]]> <![CDATA[e RMSE ]]> Prediction Model I 20.4 0.6262 Prediction Model II 15.7 0.4925

[0069] It can be seen that the similar day prediction model II can effectively improve the prediction accuracy of the model.

[0070] The present invention proposes a method for predicting the output power of a photovoltaic power generation system. By analyzing various factors that affect the output of the photovoltaic power generation system, a historical output power sequence is selected, and relevant meteorological information is used as the input of a prediction model. In order to improve the prediction accuracy of the model under different weather conditions, a similar day selection algorithm and a method for determining training samples based on the meteorological information of the prediction day are proposed, which can effectively improve the prediction accuracy of the model under various weather conditions. The model directly predicts the output power, avoids complex modeling, and simplifies the prediction process. The BP neural network prediction model of the present invention has good accuracy and has certain practicality and feasibility.

[0071] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations within the meaning and scope of the elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for predicting the output power of a photovoltaic power generation system, characterized in that: The steps include: S1: Select factors with strong correlation with output power and construct a photovoltaic array output model; S2: Select similar days; S3: Construct the BP neural network prediction model through the input layer, hidden layer and output layer parameters; S4: Train the BP neural network prediction model to improve the prediction speed and prediction accuracy of the BP neural network prediction model.

2. The method for predicting output power of a photovoltaic power generation system according to claim 1, characterized in that: In step S1, the photovoltaic power output power per unit area can be calculated according to the following formula: P s =ηSI[1-0.005(t0+25)] Where: η is the photoelectric conversion efficiency of the photovoltaic array; S is the total area of ​​the photovoltaic array; I is the solar radiation intensity received by the photovoltaic array; t0 is the operating temperature of the solar panel.

3. The method for predicting output power of a photovoltaic power generation system according to claim 1, characterized in that: In step S2, in order to select a date with the same weather type and season type as the predicted day and the closest temperature from the historical records of the photovoltaic power generation system, the output power sequence and temperature information are used as part of the prediction model input, and a similar day selection algorithm is proposed. The algorithm steps are as follows: 1) Select n historical records that are consistent with the weather type and season type of the predicted day to form a sample set D; 2) Calculate the Euclidean distance d between the predicted day and the historical records of the temperature in the sample set D i , d i The calculation formula is as follows: Where: Y1, Y2, Y3 are the highest temperature, lowest temperature and average temperature of the forecast day respectively; X i1 , X i2 , X i3 are the maximum temperature, minimum temperature and average temperature values ​​of the i-th record in the sample set D respectively; 3) The temperature Euclidean distance set {d1, d2, …d n }Sort in ascending order according to the value. The date corresponding to the minimum value is the similar day corresponding to the predicted day.

4. The method for predicting output power of a photovoltaic power generation system according to claim 1, characterized in that: In step S3, the BP neural network prediction model includes an input layer, a hidden layer and an output layer; Selection of input layer variables: The input variables are similar daily output power series, temperature information, and predicted daily temperature information. There are 20 input variables in total. Selection of output layer variables: Since the output power model of the photovoltaic power generation system predicts the average output power per hour from 6:00 to 19:00 on the prediction day, there are 14 output variables in total; Selection of hidden layer variables: According to Kolmogorov theorem, if the number of input variables is n, the number of hidden layer neurons can generally be 2n+1. The number of hidden layer neurons can also be determined by repeated trial and error.

5. The method for predicting output power of a photovoltaic power generation system according to claim 1, characterized in that: In step S4, the generalization ability of the BP neural network is the main performance indicator. When the BP neural network is trained using samples, a part of the input variable data is used as a training set and the other part is used as a test set.

Images